feat(openclaude): vendor OpenClaude as headless gRPC daemon, eliminate aider

Replace fragile aider+litellm subprocess shell-out with vendored
OpenClaude built from source via Bun, exposed as a headless gRPC
daemon on :50051 (DigitalOcean primary) and :50052 (OpenRouter
fallback). New openclaude_grpc/ Python bridge preserves the legacy
(combined_output, exit_code) contract used by SAST gate, conviction
engine, adversarial reviewer, and red-team loops.

- Dockerfile: 3-stage build (bun bundle -> python base -> runtime
with auto protoc stub generation). Removes aider/litellm/configargparse
hacks (-30 lines).
- requirements.txt: purged aider-chat + litellm + 27 curated runtime
imports; adds grpcio + protobuf as the only new Python deps.
- entrypoint.sh: launches both daemons with provider-specific OPENAI_*
env, hot-loads /tmp/mcp_runtime.json per chat session.
- app.py: run_aider is now a thin alias to run_openclaude. Legacy
aider-patcher MCP server entry deleted.
- vendor/openclaude/src/grpc/server.ts: OPENCLAUDE_AUTO_APPROVE=1
short-circuits permission prompts; reconnectMcpServerImpl loads MCP
servers from MCP_RUNTIME_CONFIG with mtime-based caching.
- SYSTEM_ANALYSIS_BOOK.md: changelog row 7 + section 2.1 rewritten.

.gitignore_extra

@@ -0,0 +1 @@
+vendor/openclaude/node_modules/

ARCHITECTURE_ANALYSIS.md

@@ -1,527 +0,0 @@
-# Rhodawk AI DevSecOps Engine — End-to-End Architectural Analysis
-
-> Generated by an automated architect/security-researcher pass over commit
-> `14b1bbe` of `Architect8999/rhodawk-ai-devops-engine` on the
-> HuggingFace Spaces repository.
-> Scope: every Python module in the repo root + the new `mythos/` package
-> (~17 kLOC of Python, 30 MCP servers, 1 Dockerfile, 1 Gradio UI).
-
----
-
-## Phase 1 — High-Level Reconnaissance & Topology
-
-### 1.1 Virtual directory tree (top two levels)
-
-```
-rhodawk-ai-devops-engine/
-├── app.py                       # Gradio control plane + main loop  (~117 KB)
-├── hermes_orchestrator.py       # Multi-phase research brain        (~30 KB)
-├── language_runtime.py          # Polyglot runtime sandbox          (~70 KB)
-├── red_team_fuzzer.py           # CEGIS-style adversarial fuzzer    (~62 KB)
-│
-├── adversarial_reviewer.py      # 3-model concurrent LLM verdict
-├── audit_logger.py              # Hash-chained tamper-evident log
-├── bounty_gateway.py            # H1 / GHSA submission pipeline
-├── chain_analyzer.py            # Commit-history attack graphs
-├── commit_watcher.py            # CAD — silent-patch detector
-├── conviction_engine.py         # Multi-criterion auto-merge gate
-├── cve_intel.py                 # NVD / OSV / Exploit-DB lookups
-├── disclosure_vault.py          # Encrypted finding storage
-├── embedding_memory.py          # sentence-transformers + sqlite-vec
-├── exploit_primitives.py        # ROP / heap / shellcode helpers
-├── formal_verifier.py           # Z3 bounded verification
-├── fuzzing_engine.py            # Hypothesis fuzzer driver
-├── github_app.py                # GitHub App JWT / installation token
-├── harness_factory.py           # Auto-generates fuzzing harnesses
-├── job_queue.py                 # JSON-file-backed job ledger
-├── lora_scheduler.py            # Threshold-triggered LoRA exports
-├── memory_engine.py             # TF-IDF retrieval (legacy)
-├── notifier.py                  # Slack / Discord / email fan-out
-├── public_leaderboard.py        # Public stats endpoint
-├── repo_harvester.py            # Antagonist target picker
-├── sast_gate.py                 # bandit + 16 secret patterns
-├── semantic_extractor.py        # Function / call-graph extraction
-├── supply_chain.py              # pip-audit + typosquat heuristics
-├── swebench_harness.py          # SWE-bench evaluation runner
-├── symbolic_engine.py           # angr planner + Z3 solver glue
-├── taint_analyzer.py            # Source→sink taint tracking
-├── training_store.py            # SQLite/PG attempt ledger + HF push
-├── verification_loop.py         # Retry / prompt-build state machine
-├── vuln_classifier.py           # CWE taxonomy + scoring
-├── webhook_server.py            # GitHub webhook HTTPServer (port 7861)
-├── worker_pool.py               # Process-isolated parallel audits
-│
-├── mythos/                      # ⬅ NEW: Mythos-level upgrade package
-│   ├── MYTHOS_PLAN.md
-│   ├── __init__.py / integration.py
-│   ├── agents/                  # planner / explorer / executor / orchestrator
-│   ├── reasoning/               # probabilistic + attack_graph
-│   ├── static/                  # tree-sitter, joern, codeql, semgrep bridges
-│   ├── dynamic/                 # aflpp, klee, qemu, frida, gdb
-│   ├── exploit/                 # pwntools, ROP, heap, privesc KB
-│   ├── learning/                # rl_planner, mlflow, lora, curriculum, episodic
-│   ├── mcp/                     # 5 new MCP servers
-│   ├── skills/                  # agentskills.io registry
-│   └── api/                     # FastAPI productization (auth, webhooks, schemas)
-│
-├── mcp_config.json              # 30 MCP servers registered
-├── Dockerfile                   # python:3.12-slim + uv + node + Gradio
-├── requirements.txt             # ~30 first-class deps + Mythos optional
-├── pitch_deck/, pitch-deck/     # Marketing collateral (PDF/PPTX/HTML)
-├── FOUNDER_PLAYBOOK.md, SECURITY_RESEARCH_PLAYBOOK.md, README.md
-└── .git/
-```
-
-### 1.2 Structural paradigm
-
-A **flat-module monolith with a side-car package**. There is no explicit
-layered or hexagonal partitioning at the root: every concern (UI, orchestration,
-analysis tools, persistence, networking) lives as a peer `*.py` module
-imported by `app.py`. Communication is **in-process function calls** plus
-**JSON files on `/data`** for cross-process state (jobs, audit chain, memory).
-The new `mythos/` package adds a proper Python package with sub-modules per
-concern, intended as the migration target for a cleaner future architecture.
-
-### 1.3 Technology stack
-
-| Layer | Technology |
-|---|---|
-| Runtime | Python 3.12 (slim Docker), Node.js 20 (npm-installed for some MCP servers) |
-| UI / Control plane | **Gradio 5.29** on port 7860 |
-| Webhooks | Stdlib `http.server.BaseHTTPRequestHandler` on port 7861 |
-| LLM gateway | **OpenRouter** (DeepSeek-R1 / V3 free tier, plus Qwen ∥ Gemma ∥ Mistral consensus); env-driven model tiers in Mythos |
-| Code patching | **Aider 0.86** (driven via subprocess + MCP config) |
-| Static / SAST | bandit, ruff, semgrep, radon, custom 16-pattern secret scanner |
-| Symbolic / Formal | **z3-solver**, **angr**, custom symbolic engine |
-| Fuzzing | Hypothesis (atheris removed — see Dockerfile comments) |
-| Embeddings / Memory | sentence-transformers + sqlite-vec, optional Qdrant |
-| Persistence | SQLite (default) / Postgres (`psycopg2-binary`) for training store |
-| ML / Training | transformers, torch, datasets, custom LoRA scheduler |
-| MCP | `@modelcontextprotocol/server-github` (npm), `mcp-server-fetch` (uvx), and **30** servers in `mcp_config.json` (25 base + 5 new Mythos ones) |
-| Mythos add-ons | FastAPI + uvicorn + pydantic; optional Pyro/PyMC, MLflow, RLlib, pwntools, Frida, tree-sitter-languages |
-| Versioning / VCS | GitPython, PyGithub, PyJWT |
-| Container | Two-stage `python:3.12-slim` Dockerfile, non-root UID 1000, `/data` writable, `EXPOSE 7860`, `CMD ["python","-u","app.py"]` |
-| Deployment target | HuggingFace Spaces (declared in README YAML front-matter) |
-
-### 1.4 Headline metrics
-
-- **Python LOC:** 16 981 across 47 root modules + 47 Mythos modules.
-- **Top three by size:** `app.py` (≈3 200 LOC), `language_runtime.py` (≈70 KB),
-  `red_team_fuzzer.py` (≈62 KB), `hermes_orchestrator.py` (≈900 LOC).
-- **Tests:** none in the repo root — `pytest` is invoked **on the target
-  repository** being audited, not on Rhodawk itself.
-
----
-
-## Phase 2 — Entry Point & Execution Flow Mapping
-
-### 2.1 Entry points
-
-| Surface | Entry | Listens on | Triggered by |
-|---|---|---|---|
-| **Primary** | `python -u app.py` (Dockerfile `CMD`) | TCP `:7860` (Gradio) | User opens HF Space URL |
-| **Webhook** | `webhook_server.start_webhook_server()` invoked from `app.py` `__main__` | TCP `:7861` (HTTPServer in a daemon thread) | GitHub `push` / `pull_request` events |
-| **Mythos API** | `uvicorn mythos.api.fastapi_server:app` (manual / opt-in) | configurable | `POST /v1/analyze_target` and webhook callbacks |
-| **MCP servers** | Spawned on demand by Aider via `mcp_config.json` (`stdio` JSON-RPC) | stdin/stdout | Aider's tool calls during patching |
-
-### 2.2 Boot sequence (from `if __name__ == "__main__"` in `app.py`)
-
-```
-1. ui_log("Rhodawk AI v3.0 starting …")
-2. Daemon thread → embedding_memory.pre_warm_model()
-       (downloads sentence-transformers model in background to avoid first-call latency)
-3. start_webhook_server()
-       → HTTPServer on 0.0.0.0:7861, daemon thread,
-         registers _webhook_dispatch as the job dispatcher
-4. demo.launch(server_name=0.0.0.0, server_port=$PORT or 7860)
-       → Gradio Blocks UI bound to enterprise_audit_loop, Hermes tabs,
-         job table, audit chain viewer, leaderboard, etc.
-5. gr.Timer(3) ticks every 3 s → get_combined_refresh()
-       (single SSE stream — comment in app.py notes it replaced
-        three concurrent streams that were exhausting connection limits)
-```
-
-### 2.3 Control flow — primary "audit" loop
-
-`enterprise_audit_loop()` in `app.py` (l. 965) is the heart. Per repo:
-
-```
-configure_git_credentials()           # writes ~/.git-credentials from env
-clone target → /tmp/repo
-discover failing tests via pytest --collect-only + run
-for each failing_test:
-  process_audit_test()
-    ├─ retrieve_similar_fixes_v2()    # embedding_memory (fallback: TF-IDF memory_engine)
-    ├─ build_initial_prompt()         # verification_loop
-    ├─ write_mcp_config() → mcp_config.json on disk
-    ├─ run_aider(mcp_config_path, prompt, context_files)   # subprocess Aider with MCP
-    ├─ re-run pytest → VerificationAttempt
-    ├─ if fail and attempts < MAX_RETRIES:
-    │      build_retry_prompt(failure + previous diff) → loop
-    ├─ run_sast_gate()                # bandit + 16 secret patterns
-    ├─ run_supply_chain_gate()        # pip-audit + typosquat
-    ├─ run_adversarial_review()       # 3 LLMs in parallel; ACTS Bayesian score
-    ├─ run_formal_verification()      # Z3 bounded checks
-    ├─ if adversary REJECT:
-    │      retry with critique * ADVERSARIAL_REJECTION_MULTIPLIER
-    ├─ evaluate_conviction()          # multi-criteria gate
-    ├─ if conviction high → auto_merge_pr() else create_github_pr()
-    ├─ record_attempt() / update_test_result()  # training_store (SQLite/PG)
-    ├─ record_fix_outcome()           # memory_engine writes back lesson
-    └─ maybe_trigger_training()       # lora_scheduler exports HF dataset
-```
-
-A parallel **Hermes research mode** (`hermes_orchestrator.run_hermes_research`)
-sits beside the audit loop. It runs the six phases RECON → STATIC → DYNAMIC →
-EXPLOIT → CONSENSUS → DISCLOSURE and produces a `HermesSession` containing
-`VulnerabilityFinding`s with VES / TVG / ACTS / CAD / SSEC scores.
-
-### 2.4 Data flow
-
-```
-GitHub repo  ──clone──►  /tmp/repo/<sha>           (ephemeral)
-                                  │
-            pytest output ────────┤
-                                  ▼
-                         memory_engine ◄── embedding_memory (sqlite-vec @ /data)
-                                  │
-                                  ▼
-                          MCP-equipped Aider ── subprocess ──► fixed source
-                                  │
-            adversarial verdicts ─┤
-                                  ▼
-                         conviction_engine ──► PR / auto-merge
-                                  │
-            attempt row ──────────┴───────────► training_store
-                                                  (SQLite at /data/store.db
-                                                   or Postgres if DATABASE_URL)
-                                                              │
-                                                              ▼
-                                                     lora_scheduler ──► HF dataset
-```
-
-All persistent state lives under `/data` (writable in HF Spaces, mode 777
-in the Dockerfile). Job ledger files: `/data/jobs/<job_id>.json`. Hash-chained
-audit log: `/data/audit_chain.jsonl`. Embeddings vector store: `/data/mem.db`.
-Mythos adds `/data/mythos/{rl_state.json,episodic.sqlite,skills/}`.
-
----
-
-## Phase 3 — Functional Breakdown
-
-### 3.1 Executive summary
-
-Rhodawk is an **autonomous DevSecOps control plane**. Point it at a GitHub
-repo; it (1) reproduces failing tests, (2) drives an LLM coding agent (Aider,
-through 30 MCP-exposed tools) to write a patch, (3) re-runs the tests in a
-verification loop, (4) gates the patch through SAST + supply-chain + multi-LLM
-adversarial review + Z3 formal verification + a conviction engine, (5) opens
-or auto-merges a pull request, and (6) feeds every attempt into a training
-store so a LoRA fine-tune can be scheduled. A parallel **Hermes** mode flips
-the polarity from "fix bugs" to "find bugs": coordinated multi-phase
-vulnerability research with custom scoring algorithms (VES/TVG/ACTS/CAD/SSEC)
-and a HackerOne / GitHub Security Advisory submission gateway. The new
-**Mythos package** layers a multi-agent (planner / explorer / executor)
-framework, probabilistic Bayesian reasoning, advanced static / dynamic /
-exploit tooling, RL-driven self-improvement, 5 additional MCP servers, and a
-FastAPI productization surface on top of all of the above.
-
-### 3.2 Module responsibilities (selected)
-
-| Module | Responsibility |
-|---|---|
-| `app.py` | Gradio UI, audit loop, Aider subprocess driver, refresh timer, boot sequence |
-| `hermes_orchestrator.py` | Phase state machine, custom security metrics (VES/ACTS/TVG), tool dispatcher |
-| `verification_loop.py` | Retry policy, prompt construction, attempt accounting |
-| `adversarial_reviewer.py` | 3-model parallel verdict → consensus, used by both audit and Hermes |
-| `conviction_engine.py` | Boolean / weighted gate that decides auto-merge vs human review |
-| `sast_gate.py`, `supply_chain.py`, `formal_verifier.py` | Independent gates the patch must pass |
-| `language_runtime.py` | Polyglot sandbox factory — sets up Python venvs, Node, Java, etc. for the target repo |
-| `red_team_fuzzer.py` | CEGIS adversarial fuzzer — counter-example guided refinement |
-| `embedding_memory.py` / `memory_engine.py` | v2 semantic retrieval (sentence-transformers + sqlite-vec); v1 TF-IDF fallback |
-| `training_store.py` | SQLite / Postgres attempt ledger + `export_hf_dataset` for HF push |
-| `lora_scheduler.py` | Threshold-triggered LoRA training-data export |
-| `audit_logger.py` | Append-only hash-chained log + `verify_chain_integrity` |
-| `webhook_server.py` | HMAC-verified GitHub webhook receiver, IP rate-limit, dispatcher hook |
-| `worker_pool.py` | Process-isolated parallel test handling (`MAX_WORKERS`) |
-| `bounty_gateway.py` | Holds findings for human approval → submits to HackerOne / opens GHSA |
-| `vuln_classifier.py` / `cve_intel.py` | CWE taxonomy + NVD/OSV/Exploit-DB lookup |
-| `commit_watcher.py` / `chain_analyzer.py` | Silent-patch detection + per-commit attack-graph diffing |
-| `mythos/agents/*` | Planner produces a probabilistic plan; Explorer enumerates hypotheses; Executor runs tools; Orchestrator routes |
-| `mythos/reasoning/probabilistic.py` | Bayesian hypothesis sampling (Pyro / PyMC / NumPy fallback) |
-| `mythos/learning/rl_planner.py` | Tool-selection policy (RLlib / SB3 / UCB1 fallback), state at `/data/mythos/rl_state.json` |
-| `mythos/api/fastapi_server.py` | `POST /v1/analyze_target`, auth middleware, webhook callbacks |
-| `mythos/mcp/*` | 5 new servers exposed via `python -m mythos.mcp.<name>` and registered in `mcp_config.json` |
-
-### 3.3 Background work / scheduled tasks
-
-- **Embedding pre-warm thread** (daemon, `app.py` `__main__`).
-- **Webhook HTTPServer thread** (daemon).
-- **Gradio refresh timer** (3 s SSE tick) — coalesced into one stream.
-- **Worker-pool subprocesses** for per-test isolation (`worker_pool._run_isolated`).
-- **LoRA scheduler** triggers on attempt-count threshold (no cron — checked at the end of each audit).
-- **MCP server lifecycle** — Aider spawns each declared MCP server on demand and tears it down with the patch session.
-
-There is **no Celery / RQ / APScheduler** — concurrency is purely
-threading + subprocess + ad-hoc daemons.
-
----
-
-## Phase 4 — Build & Execution Guide
-
-### 4.1 Prerequisites
-
-| Required | Notes |
-|---|---|
-| Python **3.12** | Pinned in README front-matter and Dockerfile |
-| Node.js + npm | Only for `@modelcontextprotocol/server-github` |
-| `uv` (Astral) | Used by `language_runtime` to materialise per-target venvs |
-| `git` | GitPython invokes the system binary |
-| Writable `/data` (Linux) or local equivalent | All persistent state lives here |
-| **Env vars** | `OPENROUTER_API_KEY` (mandatory for any LLM call); `GITHUB_TOKEN` *or* GitHub App creds (`GITHUB_APP_ID`, `GITHUB_APP_PRIVATE_KEY`, `GITHUB_INSTALLATION_ID`); optional `DATABASE_URL` for Postgres training store; `HF_TOKEN` for HF dataset push; `GITHUB_WEBHOOK_SECRET` for webhook HMAC; Slack/Discord URLs for notifier; Mythos tier overrides `MYTHOS_TIER1_PRIMARY` / `MYTHOS_TIER2_PRIMARY` etc.; `RHODAWK_MYTHOS=1` to engage multi-agent loop |
-| Optional native tools | Joern, CodeQL, AFL++, KLEE, QEMU, Frida, GDB, ROPGadget, pwntools — Mythos bridges degrade gracefully if absent |
-
-### 4.2 Setup — local
-
-```bash
-# 1. Clone
-git clone https://huggingface.co/spaces/Architect8999/rhodawk-ai-devops-engine
-cd rhodawk-ai-devops-engine
-
-# 2. Python deps
-python3.12 -m venv .venv && source .venv/bin/activate
-pip install --upgrade pip
-pip install -r requirements.txt mcp-server-fetch
-
-# 3. Node-based MCP server
-npm install -g @modelcontextprotocol/server-github
-
-# 4. (Optional) astral uv for runtime sandboxing
-curl -LsSf https://astral.sh/uv/install.sh | sh
-
-# 5. Persistent state directory
-sudo mkdir -p /data && sudo chmod 777 /data
-
-# 6. Environment
-export OPENROUTER_API_KEY=sk-or-...
-export GITHUB_TOKEN=ghp_...                 # or GITHUB_APP_* trio
-export GITHUB_WEBHOOK_SECRET=whsec_...
-# Optional
-export DATABASE_URL=postgres://...
-export HF_TOKEN=hf_...
-export RHODAWK_MYTHOS=1
-```
-
-### 4.3 Run
-
-```bash
-# Primary control plane (Gradio on :7860, webhook on :7861)
-PORT=7860 python -u app.py
-
-# Or via Docker (the way HF Spaces runs it)
-docker build -t rhodawk-ai .
-docker run -it --rm \
-  -p 7860:7860 -p 7861:7861 \
-  -e OPENROUTER_API_KEY -e GITHUB_TOKEN -e GITHUB_WEBHOOK_SECRET \
-  -v $PWD/data:/data \
-  rhodawk-ai
-
-# Mythos productization API (independent of the Gradio loop)
-uvicorn mythos.api.fastapi_server:app --host 0.0.0.0 --port 8000
-
-# Run a single Mythos MCP server manually (smoke test)
-python -m mythos.mcp.static_analysis_mcp
-```
-
-### 4.4 Verification
-
-```bash
-# 1. Gradio UI returns HTML
-curl -sf http://localhost:7860/ | head -n 5
-
-# 2. Webhook server is up (expect 405 Method Not Allowed on GET)
-curl -sv http://localhost:7861/webhook 2>&1 | grep "HTTP/1"
-
-# 3. Send a synthetic GitHub ping (replace SECRET)
-BODY='{"zen":"hello"}'
-SIG="sha256=$(printf '%s' "$BODY" | openssl dgst -sha256 -hmac "$GITHUB_WEBHOOK_SECRET" | cut -d' ' -f2)"
-curl -sv -X POST http://localhost:7861/webhook \
-  -H "X-GitHub-Event: ping" \
-  -H "X-Hub-Signature-256: $SIG" \
-  -H "Content-Type: application/json" \
-  -d "$BODY"
-
-# 4. Mythos API health
-curl -sf http://localhost:8000/healthz
-curl -sX POST http://localhost:8000/v1/analyze_target \
-  -H "Content-Type: application/json" \
-  -d '{"target":"https://github.com/octocat/Hello-World","mode":"recon"}'
-
-# 5. Verify the audit chain is internally consistent
-python -c "from audit_logger import verify_chain_integrity; print(verify_chain_integrity())"
-
-# 6. Confirm 30 MCP servers register
-jq '.mcpServers | length' mcp_config.json   # → 30
-```
-
-In the Gradio UI you should see live logs ticking every 3 s, the metrics
-row populating, and the **Hermes** and **CWE Reference** tabs available.
-
----
-
-## Phase 5 — Architectural Thoughts & Stability Assessment
-
-### 5.1 Patterns in use
-
-- **Pipeline / chain-of-responsibility** — the audit loop is a clean
-  sequence of independent gates (`sast_gate → supply_chain → adversarial →
-  formal → conviction`). Each gate exposes a single function returning a
-  decision dataclass; they are trivially composable.
-- **Strategy + graceful-degradation wrappers** — Mythos bridges
-  (`mythos/static/joern_bridge.py`, etc.) all expose `available()` plus a
-  pure-Python fallback. The orchestrator can plan with whatever is on the
-  PATH today.
-- **Tier-routing for LLMs** — Hermes has two model tiers (`HERMES_MODEL`
-  reasoning + `HERMES_FAST_MODEL` for cheap triage); Mythos generalises this
-  to env-driven `MYTHOS_TIER{1,2}_{PRIMARY,FALLBACK}`.
-- **Bayesian / consensus voting** — `compute_acts()` produces a Bayesian
-  multi-model trust score that is reused by both adversarial review and
-  conviction.
-- **Append-only hash chain** — `audit_logger` builds a Merkle-style chain
-  with `verify_chain_integrity()`, which is the right primitive for
-  compliance evidence.
-- **MCP as the universal tool bus** — every external capability (SAST,
-  fuzzers, vuln DBs, web search, GitHub, Postgres) is exposed through
-  `mcp_config.json`; this is by far the strongest piece of architecture in
-  the codebase and is what makes Aider's tool use uniform.
-- **Multi-agent (Mythos)** — Planner emits a probabilistic plan, Explorer
-  enumerates hypotheses, Executor invokes tools, Orchestrator routes — a
-  textbook multi-agent shape with clean message contracts in
-  `mythos/agents/base.py`.
-
-### 5.2 Strengths
-
-- **Sharp separation between "fix mode" (audit loop) and "find mode"
-  (Hermes)** — they share gates and the training store but never tangle.
-- **Robust webhook surface** — HMAC verification, IP rate-limit, dispatcher
-  injection (`set_job_dispatcher`) keeps the receiver pure.
-- **Worker-pool isolation** — `worker_pool._run_isolated` puts each test fix
-  in its own subprocess, which contains LLM/aider blow-ups well.
-- **Mythos add-on is non-invasive** — opt-in via `RHODAWK_MYTHOS=1` and a
-  separate FastAPI surface, so the existing Gradio UX is unchanged unless
-  you want it.
-- **Persistent learning loop** — every attempt, success or failure, lands
-  in `training_store` and feeds `lora_scheduler.maybe_trigger_training`.
-  The flywheel is real and not aspirational.
-
-### 5.3 Bug surfaces & risks
-
-The list below is **prioritised** — items are ordered by likely real-world
-impact on stability or security.
-
-#### High
-
-1. **`app.py` is a 117 KB god-module.** Boot, UI definition, audit loop,
-   subprocess management, MCP-config writing, Hermes UI bindings, and the
-   refresh timer all live in one file. This is the biggest stability risk:
-   any change ripples broadly and there are no unit tests on Rhodawk
-   itself. Recommend extracting the UI definition into `ui/`, the audit
-   loop into `audit/`, and process management into `procctl.py`.
-2. **No test suite for Rhodawk.** `pytest` is invoked only against target
-   repos. There is no CI guard against regressions in the orchestration
-   logic. This is the single highest-leverage fix.
-3. **`/data` mode `777` and shared by every tenant.** The Dockerfile sets
-   `chmod 777 /data` (because of HF Spaces UID quirks). All tenants share
-   `/data/jobs/`, `/data/audit_chain.jsonl`, `/data/store.db`, etc. The
-   `tenant_id` is stamped on each job key, but a buggy import or path-
-   traversal-style filename would let one tenant's data overwrite
-   another's. Add a `pathlib` allowlist + per-tenant subdirectories.
-4. **Shell-out via `git`, `pytest`, `aider`, `npm`** with target-controlled
-   filenames. `run_subprocess_safe` exists, but several callers stitch
-   strings before reaching it. Audit every `subprocess.run` for
-   `shell=True` and for unvalidated repo paths.
-5. **Embedding model pre-warm runs in a daemon thread without back-off.**
-   If `sentence-transformers` fails (rate-limited HF, no disk), every
-   subsequent retrieval call falls back silently to TF-IDF — a real
-   correctness regression that is invisible to operators. Surface this
-   state on the dashboard.
-6. **Hermes phase state is in-process.** `HermesSession` lives in module
-   memory; if `app.py` restarts mid-research the entire session is lost.
-   Persist `HermesSession.asdict()` to `/data/hermes/<session_id>.json`
-   on every phase transition.
-
-#### Medium
-
-7. **`memory_engine.py` (TF-IDF) and `embedding_memory.py` (vec) compete.**
-   `app.py` imports both, and the audit loop calls v2 with v1 as silent
-   fallback. Two stores will drift. Pick one as the source of truth and
-   demote the other to "legacy".
-8. **Webhook server uses stdlib HTTPServer (single-threaded by default).**
-   `start_webhook_server()` should use `ThreadingHTTPServer` (or, ideally,
-   move the same handlers to FastAPI now that uvicorn is a dep).
-9. **`MAX_RETRIES * ADVERSARIAL_REJECTION_MULTIPLIER`** can produce long
-   tail loops where a stubborn adversary blocks the pipeline. Add an
-   absolute wall-clock cap per audit.
-10. **JSON job ledger is read/written without `flock`.** Two parallel
-    workers updating the same job file race. Use SQLite (which is already
-    a dependency) for the job queue too — this also fixes (3) by giving
-    you per-tenant rows instead of files.
-11. **`audit_logger` chain integrity is only verified on demand.** If the
-    chain has been tampered with, no one notices until someone clicks the
-    button. Add a periodic background verifier that pages on a break.
-12. **Mythos optional deps are mostly commented out** — anyone who pip-
-    installs the file gets the FastAPI surface but not Pyro/MLflow/RLlib.
-    Document the on-demand install path more loudly in `MYTHOS_PLAN.md`.
-
-#### Low / hygiene
-
-13. `language_runtime.py` is ~70 KB with deep `if-lang ==` ladders. Replace
-    with a `Runtime` class registry plus `entry_points` so adding Go, Rust,
-    or .NET is a one-file change.
-14. Many modules use `time.sleep()` for backoff instead of `tenacity` even
-    though `tenacity` is already imported in `app.py`.
-15. `red_team_fuzzer.py` (62 KB) duplicates patterns that
-    `fuzzing_engine.py` already has — consolidate.
-16. `notifier.py` swallows exceptions silently; add a structured failure
-    counter.
-17. `pitch_deck/` and `pitch-deck/` (hyphen vs underscore) coexist as
-    sibling directories — pick one.
-
-### 5.4 Memory / leak surfaces
-
-- `_hermes_logs: list[str]` is unbounded — every Hermes run appends without
-  truncation. Add a ring buffer (`collections.deque(maxlen=10_000)`).
-- The Gradio `gr.Timer(3)` keeps building string responses even when no
-  client is connected. The single-stream coalescing helps, but the
-  `live_logs` textbox still grows unboundedly.
-- Aider subprocesses inherit the parent file descriptors; a long-running
-  audit can exhaust FDs. Add `close_fds=True` everywhere.
-- The embedding store grows monotonically. Add a TTL eviction in
-  `embedding_memory`.
-
-### 5.5 Targeted stabilisation roadmap
-
-| Priority | Change | Effort | Payoff |
-|---|---|---|---|
-| 🔥 P0 | Add a `tests/` directory with pytest covering: webhook signature, audit-loop happy path with mocked Aider, audit-chain integrity, MCP-config render | M | Catches future breakage of every gate |
-| 🔥 P0 | Per-tenant subdirectory under `/data/<tenant>/` and a path-allowlist helper | S | Hard isolation between tenants |
-| 🔥 P0 | Migrate `job_queue` from JSON files to the existing SQLite store | S | Removes file-locking races |
-| ⚡ P1 | Carve `app.py` into `ui_blocks.py` + `audit_loop.py` + `procctl.py` (no behaviour change) | M | Drops blast radius of every future change |
-| ⚡ P1 | Switch `webhook_server` to `ThreadingHTTPServer` *or* mount it under the new FastAPI app | S | Concurrency + one fewer port |
-| ⚡ P1 | Surface "embedding model healthy?" on the dashboard | XS | Stops silent fallback to TF-IDF |
-| ⚡ P1 | Persist `HermesSession` after every phase transition | S | Crash-safe research mode |
-| 🛠 P2 | Replace `if-lang ==` ladders in `language_runtime` with a registry | M | Future polyglot support |
-| 🛠 P2 | Bounded ring buffers for `_hermes_logs` + Gradio `live_logs` | XS | Stops slow memory growth |
-| 🛠 P2 | Consolidate `memory_engine` and `embedding_memory` behind one interface | M | One source of retrieval truth |
-| 🛡 P2 | Explicit security pass on every `subprocess.run` for shell injection / path traversal in target-controlled inputs | M | Hardens the polyglot runtime |
-| 🛡 P3 | Background verifier thread for the audit hash-chain that pages on break | S | Compliance evidence stays trustworthy |
-
-### 5.6 Bottom line
-
-The architecture is **ambitious, coherent, and unusually mature for a
-single-Space project** — the gating pipeline, the MCP tool bus, and the
-Mythos multi-agent / RL extension are genuinely well-thought-out. The two
-things holding it back from production-grade are (a) the absence of a
-self-test suite and (b) the size of `app.py`. Both are mechanical, not
-architectural, problems. Address P0 + P1 above and Rhodawk graduates from
-"impressive HuggingFace Space" to "shippable security platform".

COMPARISON_REPORT.md

@@ -1,239 +0,0 @@
-# Rhodawk AI DevSecOps — RHODAWK_SUPERHUMAN_MASTERPLAN Comparison Report
-
-*Generated: April 22, 2026*
-
-This document compares every line item in `RHODAWK_SUPERHUMAN_MASTERPLAN.md`
-against the actual state of the repo after this delivery cycle.  Entries
-fall into one of four buckets:
-
-- ✅ **DONE** — implemented in this commit.
-- 🟢 **PRE-EXISTING** — already shipped before this cycle; verified.
-- 🟡 **PARTIAL** — scaffold present, deeper work still required.
-- 🔴 **NOT YET** — out of scope for this cycle, listed for the next one.
-
----
-
-## PART 1 — Model & Skill Foundation
-
-| Item | Status | Notes |
-|---|---|---|
-| MiniMax M2.5-highspeed → T1 fast | ✅ DONE | `architect/model_router.py` rewritten — `TIER1_PRIMARY = minimax/minimax-m2.5-highspeed`, env override `TIER1_PRIMARY_MODEL`. |
-| DeepSeek V3 → T1 deep code lane | ✅ DONE | `TIER1_DEEP = deepseek/deepseek-chat-v3` mapped to `static_analysis`, `patch_generation`. |
-| Qwen3-235B-A22B → T2 reasoning | ✅ DONE | `TIER2_PRIMARY` mapped to `exploit_reasoning`, `chain_synthesis`, `adversarial_review_c`. |
-| Claude Sonnet 4.6 → T4 polish | ✅ DONE | `TIER4_PRIMARY` mapped to `critical_cve_draft`. |
-| DeepSeek-R1-32B-AWQ local → T5 | ✅ DONE | `TIER5_LOCAL`, used for `bulk_triage`, budget-exceeded fallback. |
-| `build_skill_system_prompt(profile)` | ✅ DONE | New helper in `model_router.py`. |
-| `call_with_skills(task, prompt, profile)` | ✅ DONE | New helper that combines routing + skill injection + LLM call. |
-| ACTS 3-model consensus updated | ✅ DONE | `adversarial_review_a/b/c` now route to three distinct providers. |
-| Hard budget guardrail | 🟢 PRE-EXISTING | `_BUDGET` + `record_usage()` retained. |
-
-### 8 new domain skills (`architect/skills/`)
-
-| Skill file | Status |
-|---|---|
-| `smart-contract-audit.md` | ✅ DONE |
-| `ai-ml-security.md` | ✅ DONE |
-| `ci-cd-pipeline-attack.md` | ✅ DONE |
-| `zero-day-research.md` | ✅ DONE |
-| `llm-system-prompt-injection.md` | ✅ DONE |
-| `linux-kernel-exploitation.md` | ✅ DONE |
-| `browser-engine-security.md` | ✅ DONE |
-| `cryptographic-implementation.md` | ✅ DONE |
-
-### Imported community skills
-
-| Skill | Source | Status |
-|---|---|---|
-| `bb-methodology-claude.md` | `shuvonsec/claude-bug-bounty` (MIT) | ✅ DONE |
-| `bug-bounty-reference-index.md` | `ngalongc/bug-bounty-reference` (CC) | ✅ DONE |
-
-Total skill count after this cycle: **29** markdown skill files
-(was 19 pre-cycle).
-
----
-
-## PART 2 — OSS-Guardian Lane
-
-| Item | Status | Notes |
-|---|---|---|
-| `oss_target_scorer.py` | ✅ DONE | Stars × dependents × language-risk × CVE-history × freshness. Pure-function, unit-test ready. |
-| `oss_guardian.py` runner | ✅ DONE | End-to-end: sandbox → runtime detect → fix-vs-attack split → Hermes attack → routing to disclosure_vault / GitHub PR / embodied bridge. CLI entrypoint `python -m oss_guardian --repo …`. |
-| `repo_harvester.py` | 🟢 PRE-EXISTING | Untouched — feeds the scorer. |
-| `language_runtime.py` integration | ✅ DONE | Called inside `OSSGuardian._safe_run_tests`. |
-| Sandbox isolation | 🟢 PRE-EXISTING | `architect/sandbox.open_sandbox()` reused. |
-| GitHub PR auto-fix path | 🟡 PARTIAL | Hooked to existing `run_hermes_research` in fix-mode; full PR open + signing already in `app.py`. |
-| Disclosure-vault routing | ✅ DONE | `_route_disclosure()` — novel zero-day → vault, known CVE → PR. |
-
----
-
-## PART 3 — Knowledge & RAG
-
-| Item | Status | Notes |
-|---|---|---|
-| `knowledge_rag.py` SQLite vector store | ✅ DONE | 256-dim deterministic hash-bag embedder; auto-uses `embedding_memory.embed` when present. `add`, `add_many`, `ingest_text_file`, `query`, `stats`. |
-| Source allow-list | ✅ DONE | `SOURCES_DEFAULT` covers HackerOne, CVEDetails, ProjectZero, PortSwigger, arXiv cs.CR, awesome-bounty repos. |
-| Cross-session memory | 🟢 PRE-EXISTING | `embedding_memory.py` retained. |
-
----
-
-## PART 4 — Storage / Multi-tenant Hardening
-
-| Item | Status | Notes |
-|---|---|---|
-| `job_queue.py` SQLite migration | ✅ DONE | New WAL-mode SQLite at `/data/jobs.sqlite`. Public API preserved. Legacy `/data/jobs/*.json` files imported once on first call (`*.imported` rename). |
-| `_hermes_logs` deque(maxlen) | 🟢 PRE-EXISTING | Already `_collections.deque(maxlen=_HERMES_LOG_CAP)` in `hermes_orchestrator.py:54`. |
-| `persist_hermes_session()` | 🟢 PRE-EXISTING | `hermes_orchestrator.py:80`. |
-| Per-tenant data dirs | 🟡 PARTIAL | `job_queue.upsert_job(tenant_id=…)` already namespaced; cross-cutting per-tenant disk paths still TODO. |
-| Audit-log SHA-256 chain | 🟢 PRE-EXISTING | `audit_logger.py`. |
-
----
-
-## PART 5 — Night Hunter Lane
-
-| Item | Status | Notes |
-|---|---|---|
-| `mythos/mcp/scope_parser_mcp.py` | 🟢 PRE-EXISTING | Verified — H1 / Bugcrowd / Intigriti normalisers all in place. |
-| `mythos/mcp/subdomain_enum_mcp.py` | 🟢 PRE-EXISTING | subfinder / amass / dnsx + crt.sh fallback. |
-| `nightmode.py` phase callables | 🟢 PRE-EXISTING | `_phase_scope_ingest`, `_phase_recon`, `_phase_hunt`, `_phase_report` all wired. |
-| Telegram morning briefing | 🟢 PRE-EXISTING | `embodied_bridge.emit_status()` invoked at start + end of `run_one_cycle`. |
-| `start_in_background()` daemon | 🟢 PRE-EXISTING | Opt-in via `ARCHITECT_NIGHTMODE=1`. |
-| 5 specialist hunt agents wiring | 🟡 PARTIAL | Mythos orchestrator chained in via `_phase_hunt`; the named "auth / server-side / logic / infra / api" agent split is still represented as a single multi-iteration orchestrator run. |
-
-### MCP fleet — `mcp_config.json`
-
-| MCP | Before | After |
-|---|---|---|
-| `scope-parser-mcp` | missing | ✅ added |
-| `subdomain-enum-mcp` | missing | ✅ added |
-| `wayback-mcp` | missing | ✅ added |
-| `httpx-probe-mcp` | missing | ✅ added |
-| `shodan-mcp` | missing | ✅ added |
-| 14 existing MCPs (fetch-docs, github-manager, filesystem, memory, sequential-thinking, web-search, dynamic / static / exploit / web-security / vuln-db / recon / browser-agent / frida / ghidra / can-bus / sdr) | 🟢 PRE-EXISTING | unchanged |
-
-Total MCP servers in template: **23** (was 18).
-
----
-
-## PART 6 — EmbodiedOS / OpenClaw Bridge
-
-| Item | Status | Notes |
-|---|---|---|
-| `dispatch_to_openclaw(job_type, payload)` | ✅ DONE | New helper in `architect/embodied_bridge.py` — supports `fuzz_afl`, `klee_symbolic`, `lora_finetune`, `differential_fuzz`, `weight_scan`. |
-| `receive_openclaw_result(payload)` | ✅ DONE | Webhook handler — persists to Hermes session, emits operator status. |
-| Telegram / Discord / Hermes fan-out | 🟢 PRE-EXISTING | `emit_finding` retained. |
-| OpenClaw-side job receiver | 🔴 NOT YET | Lives in the OpenClaw repo, not this codebase. |
-| AFL++ / KLEE dispatch from `mythos/dynamic/` | 🔴 NOT YET | Bridge ready; call sites still need to flip from local subprocess to `dispatch_to_openclaw`. |
-
----
-
-## PART 7 — Productisation & Ops
-
-| Item | Status | Notes |
-|---|---|---|
-| FastAPI hardening (auth + rate limit) | 🔴 NOT YET | Tracked for the productisation cycle. |
-| Billing + API-key management | 🔴 NOT YET | Out of scope for this delivery. |
-| `app.py` split into `audit_loop.py` + `ui_blocks.py` | 🔴 NOT YET | 2,621-line monolith retained; refactor planned. |
-| Public leaderboard | 🔴 NOT YET | Stub planned. |
-| Test suite under `tests/` | 🟡 PARTIAL | The new modules (`oss_target_scorer`, `knowledge_rag`, `job_queue`, `oss_guardian`) are written to be pure-function-testable; named test files still TODO. |
-
----
-
-## PART 8 — Self-Improvement Loop
-
-| Item | Status | Notes |
-|---|---|---|
-| `training_store.py` recording chains | 🟢 PRE-EXISTING | Untouched. |
-| `lora_scheduler.py` triggering | 🟢 PRE-EXISTING | Untouched. |
-| LoRA → OpenClaw dispatch | ✅ DONE | New `dispatch_to_openclaw("lora_finetune", …)` ready to be called from `lora_scheduler.py`. |
-| Curriculum learning | 🔴 NOT YET | `mythos/learning/curriculum.py` not built. |
-| Auto-skill creation (3+ pattern) | 🔴 NOT YET | Hook point identified in the masterplan; not wired in `hermes_orchestrator.maybe_create_skill`. |
-
----
-
-## PART 9 — Security & Ethics (Non-Negotiable)
-
-All hard constraints from §8 of the masterplan are **PRE-EXISTING** and
-unchanged by this cycle:
-
-- 90-day coordinated disclosure timer in `disclosure_vault.py`.
-- `architect/sandbox.py` — local-clone-only execution.
-- Operator approval gate before every P1/P2 submission.
-- `audit_logger.py` SHA-256 chain on every action.
-- `FETCH_ALLOWED_DOMAINS` SSRF allow-list across MCPs.
-- No automated submission anywhere in the new modules — `oss_guardian`
-  routes findings to the disclosure vault, never submits directly.
-
----
-
-## Summary
-
-| Bucket | Count |
-|---|---|
-| ✅ DONE this cycle | 26 |
-| 🟢 PRE-EXISTING & verified | 16 |
-| 🟡 PARTIAL | 5 |
-| 🔴 NOT YET (next cycle) | 9 |
-
-### What you can do *right now* with this build
-
-1. `python -m oss_guardian --repo https://github.com/redis/redis` —
-   end-to-end OSS pipeline against a real target.
-2. Set `ARCHITECT_NIGHTMODE=1` and the four bounty-platform tokens; run
-   `python -c "from architect.nightmode import run_one_cycle; print(run_one_cycle())"`
-   for a single dry cycle.
-3. Use `architect.model_router.call_with_skills("static_analysis", prompt,
-   {"languages":["c"], "asset_types":["kernel"]})` to get a kernel-aware
-   prompt automatically composed with the new skill packs.
-4. `python -c "from knowledge_rag import KnowledgeRAG; r=KnowledgeRAG(); r.ingest_text_file('attached_assets/RHODAWK_SUPERHUMAN_MASTERPLAN_*.md', source='masterplan'); print(r.stats())"`.
-
-### Next-cycle priorities (recommended)
-
-1. Build the explicit five-agent split inside `_phase_hunt` (auth,
-   server-side, logic, infra, api).
-2. Flip `mythos/dynamic/` AFL++ + KLEE call sites to use
-   `dispatch_to_openclaw`.
-3. Land `tests/test_oss_guardian_dry.py`, `tests/test_model_router.py`,
-   `tests/test_scope_parser.py` and run them in CI.
-4. Split `app.py` and harden the FastAPI surface.
-
----
-
-## Round 2 — G0DM0D3 + OpenClaw-RL + Vibe-Coded Hit-List
-
-| # | Item                                                                                              | Status | Where                                                          |
-|---|---------------------------------------------------------------------------------------------------|--------|----------------------------------------------------------------|
-| 43| **Master Red-Team operator persona** — single source of truth system prompt for every LLM call    | DONE   | `architect/master_redteam_prompt.py`                            |
-| 44| **Vibe-Coded App Hunter skill** — the 24-hour 20-rule hit-list, pinned into every call            | DONE   | `architect/skills/vibe-coded-app-hunter.md` (always pinned)     |
-| 45| **`call_with_skills` upgrade** — auto master-prompt, auto pin, auto-RL-record, mode-aware         | DONE   | `architect/model_router.py` (`mode`, `pin_skills`, `record_rl`) |
-| 46| **GODMODE Consensus race** — 5-combo parallel race + composite scorer (G0DM0D3 ULTRAPLINIAN port) | DONE   | `architect/godmode_consensus.py`                                |
-| 47| **Parseltongue input perturbation** — 7 techniques × 3 tiers × 33 default triggers                | DONE   | `architect/parseltongue.py`                                     |
-| 48| **OpenClaw-RL local rollout collector** — async 4-component loop, binary + composite reward       | DONE   | `architect/rl_feedback_loop.py`                                 |
-| 49| **OpenClaw fleet dispatch for LoRA training** — flushes batched traces via embodied bridge       | DONE   | `architect/rl_feedback_loop.flush()` → `embodied_bridge.dispatch_to_openclaw("lora_finetune", …)` |
-| 50| **Operator language-feedback channel** — natural-language thumbs up/down on any trace            | DONE   | `architect/rl_feedback_loop.submit_language_feedback()`         |
-| 51| **Vibe-coded targeting baked into prompt** — operator notes + always-pinned skill + mode==hunt    | DONE   | `master_redteam_prompt.OPERATOR_DIRECTIVE` + `VIBE_CODED_HIT_LIST` |
-| 52| **Aggressive / dry-run kill switches** via env (`RHODAWK_AGGRESSIVE`, `RHODAWK_DRY_RUN`)         | DONE   | `master_redteam_prompt._operator_notes()`                       |
-
-### How to drive it
-```python
-from architect.model_router import call_with_skills
-out = call_with_skills(
-    "vuln_research",
-    "Audit https://target.example/api for the 20-rule hit-list.",
-    {"languages": ["typescript"], "frameworks": ["nextjs"], "asset_types": ["http"]},
-    mode="hunt",
-)
-print(out["decision"].model, "→", out["response"])
-```
-
-```python
-from architect.godmode_consensus import race
-res = race("PoC for the IDOR you just found at /api/orders/<id>",
-           profile={"asset_types": ["http"]})
-print(res.to_dict())
-```
-
-```python
-from architect import rl_feedback_loop
-print(rl_feedback_loop.stats())
-rl_feedback_loop.flush()    # ship to OpenClaw fleet for LoRA training
-```

Dockerfile

@@ -1,4 +1,30 @@
-# Stage 1: Builder
+# syntax=docker/dockerfile:1.7
+# ─────────────────────────────────────────────────────────────────────────
+# Rhodawk AI DevSecOps Engine — vendored OpenClaude architecture
+#
+# Stages:
+#   1. openclaude-builder  — uses Bun to compile vendor/openclaude → dist/cli.mjs
+#   2. base                — python:3.12-slim with system tooling, uv, Node, Bun
+#   3. runtime             — final image (non-root rhodawk user, EXPOSE 7860)
+#
+# Aider, litellm, configargparse and friends have been completely removed.
+# Code generation is now handled by the OpenClaude headless gRPC daemon
+# launched from entrypoint.sh.
+# ─────────────────────────────────────────────────────────────────────────
+
+ARG BUN_VERSION=1.1.42
+
+# ─── Stage 1: build the vendored OpenClaude bundle ──────────────────────
+FROM oven/bun:${BUN_VERSION} AS openclaude-builder
+WORKDIR /openclaude
+COPY vendor/openclaude/package.json vendor/openclaude/bun.lock ./
+RUN bun install --frozen-lockfile --no-progress
+COPY vendor/openclaude/ ./
+RUN bun run build && \
+    test -s dist/cli.mjs && \
+    echo "[builder] OpenClaude bundle: $(wc -c < dist/cli.mjs) bytes"
+
+# ─── Stage 2: base python+node+bun runtime ──────────────────────────────
 FROM python:3.12-slim AS base
 
 ENV DEBIAN_FRONTEND=noninteractive \
@@ -6,73 +32,84 @@ ENV DEBIAN_FRONTEND=noninteractive \
     UV_COMPILE_BYTECODE=1 \
     UV_LINK_MODE=copy
 
-RUN apt-get update && \
-    apt-get install -y --no-install-recommends git curl ca-certificates build-essential && \
-    rm -rf /var/lib/apt/lists/*
+RUN apt-get update && apt-get install -y --no-install-recommends \
+        git curl ca-certificates build-essential unzip xz-utils \
+        nodejs npm \
+    && rm -rf /var/lib/apt/lists/*
 
-# Use the official Astral image for a complete, clean uv installation
+# uv (fast Python installer used by sandboxed test runs)
 COPY --from=ghcr.io/astral-sh/uv:latest /uv /uvx /bin/
 
+# Bun (needed at runtime to launch the daemon: `bun run dev:grpc`)
+COPY --from=oven/bun:1.1.42 /usr/local/bin/bun /usr/local/bin/bun
+COPY --from=oven/bun:1.1.42 /usr/local/bin/bunx /usr/local/bin/bunx
+
 WORKDIR /build
 COPY requirements.txt .
-
-# Install dependencies directly — avoids wheel-build failures for packages
-# that require special compile-time tooling (e.g. atheris/libFuzzer).
-# atheris has been removed from requirements.txt; Hypothesis is the fallback.
-RUN pip install --no-cache-dir -r requirements.txt mcp-server-fetch && \
-    # Aider's strict pins (pillow==12.1.1, litellm==1.75.0) conflict with
-    # gradio (pillow<12) and ship a broken litellm module surface
-    # (missing APIConnectionError, _logging, encode, token_counter).
-    # We install aider WITHOUT its deps and then provide a known-good
-    # set of its actual runtime imports — letting gradio's pillow 11
-    # win, and a patched litellm replace the broken 1.75.0.
-    pip install --no-cache-dir --no-deps "aider-chat==0.86.2" && \
-    pip install --no-cache-dir --upgrade --no-deps "litellm==1.78.5" && \
-    pip install --no-cache-dir \
-        "configargparse" "jsonschema" "rich" "prompt_toolkit" "pyyaml" \
-        "packaging" "pathspec" "diskcache" "networkx" "scipy" \
-        "beautifulsoup4" "pypandoc" "flake8" "importlib_resources" \
-        "pyperclip" "pexpect" "json5" "psutil" "watchfiles" "socksio" \
-        "mixpanel" "posthog" "tree-sitter" "grep_ast" "oslex" \
-        "tokenizers" "google-generativeai" "openai" "diff-match-patch" \
-        "soundfile" "sounddevice"
-
-
-# Stage 2: Runtime — inherits installed packages from base
+RUN pip install --no-cache-dir -r requirements.txt mcp-server-fetch \
+                grpcio==1.66.* grpcio-tools==1.66.* protobuf==5.*
+
+# MCP servers used by the runtime — installed globally so `npx -y …` is
+# instantaneous instead of resolving on every audit.
+RUN npm install -g --quiet \
+        @modelcontextprotocol/server-github \
+        @modelcontextprotocol/server-filesystem \
+        @modelcontextprotocol/server-memory \
+        @modelcontextprotocol/server-sequential-thinking \
+        @modelcontextprotocol/server-git \
+        @modelcontextprotocol/server-sqlite \
+        @modelcontextprotocol/server-brave-search
+
+# ─── Stage 3: final runtime image ───────────────────────────────────────
 FROM base AS runtime
-
-LABEL org.opencontainers.image.title="Rhodawk AI DevSecOps Engine"
+LABEL org.opencontainers.image.title="Rhodawk AI DevSecOps Engine" \
+      org.opencontainers.image.source="https://github.com/Rhodawk-AI/Rhodawk-devops-engine"
 
 ENV GRADIO_SERVER_NAME=0.0.0.0 \
     GRADIO_SERVER_PORT=7860 \
     HOME=/home/rhodawk \
     PATH="/home/rhodawk/.local/bin:/usr/local/bin:$PATH" \
     UV_PYTHON_PREFERENCE=system \
-    UV_PYTHON=/usr/local/bin/python3
-
-RUN apt-get update && \
-    apt-get install -y --no-install-recommends nodejs npm && \
-    rm -rf /var/lib/apt/lists/*
-
-RUN npm install -g --quiet @modelcontextprotocol/server-github
-
-# Hugging Face UID 1000 handling
-RUN id -u 1000 >/dev/null 2>&1 && (userdel -r $(id -un 1000) || true) || true && \
+    UV_PYTHON=/usr/local/bin/python3 \
+    OPENCLAUDE_AUTO_APPROVE=1 \
+    OPENCLAUDE_GRPC_HOST=127.0.0.1 \
+    OPENCLAUDE_GRPC_PORT_DO=50051 \
+    OPENCLAUDE_GRPC_PORT_OR=50052 \
+    MCP_RUNTIME_CONFIG=/tmp/mcp_runtime.json
+
+# HuggingFace UID 1000 handling (idempotent)
+RUN id -u 1000 >/dev/null 2>&1 && (userdel -r "$(id -un 1000)" || true) || true && \
     useradd -m -u 1000 -s /bin/bash rhodawk
 
-# FIX: create /data with explicit mode so uv venv can write target_venv
-# even before the application calls os.makedirs() at runtime.
-RUN mkdir -p /data /app && chmod 777 /data && chown -R rhodawk:rhodawk /app
+RUN mkdir -p /data /app /opt/openclaude && \
+    chmod 777 /data && \
+    chown -R rhodawk:rhodawk /app /opt/openclaude
 
-# Copy the uv executable from the official image
-COPY --from=ghcr.io/astral-sh/uv:latest /uv /uvx /bin/
+# Bring the prebuilt OpenClaude bundle in as a vendored artifact.
+COPY --from=openclaude-builder --chown=rhodawk:rhodawk /openclaude /opt/openclaude
+
+# Tiny global wrappers — the orchestrator never shells out to these
+# directly any more (gRPC bridges everything), but we keep them so admins
+# can debug interactively from `docker exec`.
+RUN ln -sf /opt/openclaude/bin/openclaude /usr/local/bin/openclaude && \
+    chmod +x /usr/local/bin/openclaude
 
 WORKDIR /app
 USER rhodawk
 
-# Copy the source code
+# Source last so application edits don't bust the heavy node/python layers.
 COPY --chown=rhodawk:rhodawk . .
 
-EXPOSE 7860
+# Generate Python protobuf stubs from the vendored .proto file.
+RUN python -m grpc_tools.protoc \
+        -I /opt/openclaude/src/proto \
+        --python_out=openclaude_grpc \
+        --grpc_python_out=openclaude_grpc \
+        /opt/openclaude/src/proto/openclaude.proto && \
+    # protoc emits absolute-path imports; rewrite for relative package layout
+    sed -i 's/^import openclaude_pb2/from . import openclaude_pb2/' \
+        openclaude_grpc/openclaude_pb2_grpc.py
+
+EXPOSE 7860 50051 50052
 
-CMD ["python", "-u", "app.py"]
+ENTRYPOINT ["/app/entrypoint.sh"]

SYSTEM_ANALYSIS_BOOK.md

@@ -71,17 +71,18 @@ and build behaviour (how the container is assembled).
 | 4 | `ad06d84` | Added `gitpython==3.1.46` to `requirements.txt`. | aider 0.86.2 hard-pins gitpython; pip's resolver was bouncing between our floor (`>=3.1.40`) and aider's pin. Pinning explicitly removes the resolver thrash. |
 | 5 | `a913714` | Switched the Space SDK from `gradio` to `docker` in `README.md` front-matter and bumped `gradio>=5.49.0,<6` in `requirements.txt`. | HF was auto-injecting `gradio[oauth,mcp]==5.29.0` because the Space was registered as `sdk: gradio`. That injection conflicted with aider's `pillow==12.1.1` pin. `sdk: docker` tells HF to use the existing Dockerfile verbatim, with no auto-injection. |
 | 6 | `dd4ccce` | Removed `aider-chat` from `requirements.txt` entirely. The Dockerfile now installs it with `--no-deps` plus a curated runtime-deps list. | Even with `sdk: docker`, aider's `pillow==12.1.1` pin still conflicted with gradio's `pillow<12` constraint at the resolver level. `--no-deps` lets gradio's pillow 11 win, and the curated dep list provides everything aider actually imports at runtime. |
+| 7 | *current* | **Aider eliminated entirely. Vendored OpenClaude headless gRPC daemon now drives every code-generation turn.** Two daemons (`:50051` DigitalOcean primary, `:50052` OpenRouter fallback) launched from `entrypoint.sh`. New Python bridge `openclaude_grpc/` exposes `run_openclaude(...)` with the legacy `(combined_output, exit_code)` return contract. `run_aider` is now a thin alias. | The `--no-deps + litellm 1.78.5 + curated 27-package list` workaround was a fragile patchwork; any aider/gradio/pillow upstream bump could re-break the build. OpenClaude is a single Bun-built bundle with no Python dependency surface and a stable gRPC contract, so the Dockerfile loses ~30 lines of dependency hacks. |
 
 **Net effect today.** The Space SDK is `docker`. The build runs the
-Dockerfile end-to-end. `requirements.txt` does **not** mention `aider-chat`.
-The Dockerfile installs requirements normally, then layers in
-(a) `aider-chat==0.86.2 --no-deps`, (b) `litellm==1.78.5 --no-deps --upgrade`,
-(c) a curated set of aider's actual runtime imports (configargparse,
-jsonschema, rich, prompt_toolkit, pyyaml, pathspec, diskcache, networkx,
-scipy, beautifulsoup4, pypandoc, flake8, importlib_resources, pyperclip,
-pexpect, json5, psutil, watchfiles, socksio, mixpanel, posthog, tree-sitter,
-grep_ast, oslex, tokenizers, google-generativeai, openai, diff-match-patch,
-soundfile, sounddevice).
+Dockerfile end-to-end in three stages: (1) Bun compiles
+`vendor/openclaude/` to `dist/cli.mjs`; (2) Python 3.12-slim with system
+tooling, uv, Node and Bun is provisioned; (3) the runtime image generates
+Python protobuf stubs from `vendor/openclaude/src/proto/openclaude.proto`
+into `openclaude_grpc/`. `requirements.txt` no longer mentions
+`aider-chat`, `litellm`, `configargparse`, or any of the other 27 curated
+aider runtime imports — they have all been deleted. `entrypoint.sh`
+boots the OpenClaude DO daemon on `:50051` and (when an OpenRouter key is
+present) the OpenClaude OR daemon on `:50052`, then `exec`s `app.py`.
 
 **Provider behaviour today.** When `DO_INFERENCE_API_KEY` is present the hot
 path uses DigitalOcean Serverless Inference at
@@ -94,41 +95,43 @@ When DO is absent the system silently runs OpenRouter-only.
 ## 2. Provider Routing — DigitalOcean Primary, OpenRouter Fallback
 <a id="2-provider-routing"></a>
 
-Two independent code paths need an LLM: **aider** (the patch generator) and
-**Hermes** (the multi-phase research orchestrator + adversarial reviewer).
-Both have been wired to the same provider chain.
+Two independent code paths need an LLM: **OpenClaude** (the patch
+generator, replaces aider) and **Hermes** (the multi-phase research
+orchestrator + adversarial reviewer). Both are wired to the same
+provider chain — DigitalOcean Inference primary, OpenRouter fallback.
 
-### 2.1 Aider path — `app.py::run_aider`
+### 2.1 OpenClaude path — `app.py::run_openclaude` (alias `run_aider`)
 
 ```
-caller passes model=None
-  │
-  ├─► DEFAULT_MODEL  (constructed at import time)
-  │     ├── DO_INFERENCE_API_KEY set?  → "openai/llama3.3-70b-instruct"
-  │     └── otherwise                  → "openrouter/qwen/qwen-2.5-coder-32b-instruct:free"
+entrypoint.sh boots two OpenClaude headless gRPC daemons:
+  ┌─ DO_INFERENCE_API_KEY  → :50051  (PRIMARY,  llama3.3-70b-instruct)
+  └─ OPENROUTER_API_KEY    → :50052  (FALLBACK, qwen-2.5-coder-32b)
+
+run_openclaude(mcp_config_path, prompt, context_files):
   │
-  ├─► FALLBACK_MODELS list = ["openrouter/qwen/qwen-2.5-coder-32b-instruct:free"]
+  ├─► writes /tmp/mcp_runtime.json (the daemon hot-reloads it per chat)
+  ├─► forwards prompt + valid context-file list to the bridge
   │
-  └─► provider chain executes [primary, *fallbacks]:
-        for each model m:
-          if m starts with "openai/" and DO_INFERENCE_API_KEY:
-              env OPENAI_API_KEY      = DO_INFERENCE_API_KEY
-              env OPENAI_API_BASE     = https://inference.do-ai.run/v1
-              env OPENAI_BASE_URL     = https://inference.do-ai.run/v1
-              args += --openai-api-base, --openai-api-key
-              run aider; if exit==0 → return
-          elif m starts with "openrouter/" and OPENROUTER_API_KEY:
-              env OPENROUTER_API_KEY  = OPENROUTER_API_KEY
-              run aider; if exit==0 → return
-          else: skip
-        if all skipped → "No inference provider configured" message
-        if all failed  → return last error with provider chain printed
+  └─► openclaude_grpc.run_openclaude builds the chain [primary, fallback]:
+        for each (port, label, model) in chain:
+          client = OpenClaudeClient(host=127.0.0.1, port=port)
+          if not client.wait_ready(15s): record + continue
+          result = client.chat(message, working_directory=REPO_DIR, model=model)
+              ├── streams text_chunk → result.stdout
+              ├── streams tool_start  → result.tool_calls + stdout marker
+              ├── streams tool_result → result.stdout / .stderr
+              ├── auto-replies "y" to any action_required prompt
+              └── on done event       → exit_code 0
+          if exit_code == 0: return (combined_output, 0)
+        return (last_output, last_code)
 ```
 
-The `openai/<model>` prefix is litellm's convention for "treat this as an
-OpenAI-API-compatible endpoint." DigitalOcean's Serverless Inference exposes
-the OpenAI Chat Completions schema verbatim, so litellm + aider need no
-DO-specific code path — only env vars and a base URL override.
+There is no litellm, no `--openai-api-base` shell argument, no per-model
+prefix munging. Each daemon is launched with `OPENAI_API_KEY`,
+`OPENAI_BASE_URL`, `OPENAI_MODEL` set in its own process environment, so
+the gRPC client just speaks to whichever port matches the desired
+provider. The legacy `run_aider` symbol is preserved as an alias for
+backwards compatibility with the Hermes/SAST/red-team callers.
 
 ### 2.2 Hermes path — `hermes_orchestrator.py::_hermes_llm_call`
 

app.py

@@ -441,12 +441,10 @@ def write_mcp_config() -> str:
                 "args": ["mcp-server-shell", "--allow-commands", "ruff"],
                 "description": "Ruff ultra-fast Python linter — anti-patterns correlating with security bugs"
             },
-            "aider-patcher": {
-                "command": "uvx",
-                "args": ["mcp-server-shell", "--allow-commands", "aider"],
-                "description": "Aider AI code editor — applies LLM-generated patches with diff verification and test re-run",
-                "env": {"OPENROUTER_API_KEY": _openrouter_key}
-            },
+            # aider-patcher removed — code generation now flows through the
+            # vendored OpenClaude headless gRPC daemon (see openclaude_grpc/).
+            # The model itself talks to the rest of the MCP suite directly,
+            # so no recursive "patcher" MCP server is needed.
             "cve-intelligence": {
                 "command": "uvx",
                 "args": ["mcp-server-fetch"],
@@ -616,85 +614,76 @@ def setup_target_venv() -> str:
 # ──────────────────────────────────────────────────────────────
 # AIDER RUNNER
 # ──────────────────────────────────────────────────────────────
-def _aider_provider_options(model: str) -> tuple[list[str], dict[str, str]]:
-    """
-    Return (extra_cli_args, env_overrides) for a given aider --model string.
-    Routes ``openai/<model>`` calls through DigitalOcean Serverless Inference,
-    and ``openrouter/<model>`` calls through OpenRouter.
-    """
-    env: dict[str, str] = {}
-    args: list[str] = []
-    if model.startswith("openai/") and DO_INFERENCE_API_KEY:
-        # DigitalOcean OpenAI-compatible endpoint.
-        env["OPENAI_API_KEY"] = DO_INFERENCE_API_KEY
-        env["OPENAI_API_BASE"] = DO_INFERENCE_BASE_URL
-        # Some litellm versions also read these:
-        env["OPENAI_BASE_URL"] = DO_INFERENCE_BASE_URL
-        args += ["--openai-api-base", DO_INFERENCE_BASE_URL,
-                 "--openai-api-key", DO_INFERENCE_API_KEY]
-    elif model.startswith("openrouter/") and OPENROUTER_API_KEY:
-        env["OPENROUTER_API_KEY"] = OPENROUTER_API_KEY
-    return args, env
-
-
-def _run_aider_once(model: str, prompt_path: str, valid_files: list[str]) -> tuple[str, int]:
-    extra_args, env_overrides = _aider_provider_options(model)
-    cmd = ["aider", "--model", model, "--yes", "--no-stream",
-           "--message-file", prompt_path] + extra_args + valid_files
-    return run_subprocess_safe(cmd, cwd=REPO_DIR, timeout=600,
-                               env_overrides=env_overrides,
-                               raise_on_error=False)
+###############################################################################
+# OpenClaude gRPC bridge — replaces the legacy aider subprocess shell-out.
+#
+# Two daemons run side-by-side inside the container (see entrypoint.sh):
+#   :50051  → DigitalOcean Inference  (PRIMARY)
+#   :50052  → OpenRouter              (FALLBACK)
+# `run_openclaude` walks the chain in order, returning the first 0-exit
+# response, or the last failure if both providers reject the prompt.
+#
+# The signature mirrors the old `run_aider` exactly so every caller in this
+# module — the 15-step healing loop, conviction engine, adversarial review
+# wrappers — keeps working without modification.
+###############################################################################
+from openclaude_grpc import run_openclaude as _run_openclaude_bridge
+
+# Provider chain configuration (resolved once at import time).
+_OPENCLAUDE_PRIMARY_PORT = (
+    int(os.getenv("OPENCLAUDE_GRPC_PORT_DO", "50051"))
+    if DO_INFERENCE_API_KEY else 0
+)
+_OPENCLAUDE_FALLBACK_PORT = (
+    int(os.getenv("OPENCLAUDE_GRPC_PORT_OR", "50052"))
+    if OPENROUTER_API_KEY else 0
+)
+_OPENCLAUDE_PRIMARY_MODEL = DO_INFERENCE_MODEL if DO_INFERENCE_API_KEY else ""
+_OPENCLAUDE_FALLBACK_MODEL = os.getenv(
+    "OPENROUTER_MODEL",
+    "qwen/qwen-2.5-coder-32b-instruct:free",
+)
 
 
 def run_aider(mcp_config_path: str, prompt: str, context_files: list[str]) -> tuple[str, int]:
+    """Backwards-compatible alias kept so existing call sites and tests do
+    not break. Internally delegates to the OpenClaude gRPC bridge."""
+    return run_openclaude(mcp_config_path, prompt, context_files)
+
+
+def run_openclaude(mcp_config_path: str, prompt: str,
+                   context_files: list[str]) -> tuple[str, int]:
+    """Issue one healing turn against the OpenClaude gRPC daemons.
+
+    Returns the same ``(combined_output, exit_code)`` tuple shape that
+    aider used to return so the rest of the orchestrator (validation
+    loop, SAST gate, conviction engine, red-team checks) plugs in
+    unchanged.
     """
-    Run aider with DigitalOcean Serverless Inference as PRIMARY and
-    OpenRouter as FALLBACK. If the primary call exits non-zero (rate
-    limit, provider outage, transient crash), we automatically retry the
-    same prompt against the OpenRouter fallback model.
-
-    Note: --mcp-config is NOT a supported aider CLI flag in any released
-    version of aider-chat (including 0.86.1). The mcp_config_path argument
-    is kept in the function signature for forward-compatibility but is not
-    forwarded to the subprocess until upstream adds the flag.
-    """
-    fd, prompt_path = tempfile.mkstemp(prefix="aider_prompt_", suffix=".txt")
-    try:
-        with os.fdopen(fd, "w") as f:
-            f.write(prompt)
-        valid = [f for f in context_files if os.path.exists(os.path.join(REPO_DIR, f))]
-
-        # Build provider chain: primary first, fallback second (deduped).
-        chain: list[str] = []
-        for m in (MODEL, FALLBACK_MODEL):
-            if m and m not in chain:
-                # Skip a provider if its credentials aren't configured.
-                if m.startswith("openai/") and not DO_INFERENCE_API_KEY:
-                    continue
-                if m.startswith("openrouter/") and not OPENROUTER_API_KEY:
-                    continue
-                chain.append(m)
-        if not chain:
-            return ("No inference provider configured: set DO_INFERENCE_API_KEY "
-                    "or OPENROUTER_API_KEY", 1)
-
-        last_output, last_code = "", 1
-        for idx, model in enumerate(chain):
-            provider = "DigitalOcean" if model.startswith("openai/") else "OpenRouter"
-            ui_log(f"Aider attempt via {provider} ({model})", "INFO")
-            output, code = _run_aider_once(model, prompt_path, valid)
-            last_output, last_code = output, code
-            if code == 0:
-                return output, code
-            if idx < len(chain) - 1:
-                ui_log(f"{provider} failed (exit {code}) — falling back to "
-                       f"{chain[idx + 1]}", "WARN")
-        return last_output, last_code
-    finally:
-        try:
-            os.unlink(prompt_path)
-        except OSError:
-            pass
+    if _OPENCLAUDE_PRIMARY_PORT == 0 and _OPENCLAUDE_FALLBACK_PORT == 0:
+        return (
+            "No inference provider configured: set DO_INFERENCE_API_KEY "
+            "or OPENROUTER_API_KEY",
+            1,
+        )
+
+    valid = [f for f in context_files
+             if os.path.exists(os.path.join(REPO_DIR, f))]
+
+    return _run_openclaude_bridge(
+        mcp_config_path,
+        prompt,
+        valid,
+        repo_dir=REPO_DIR,
+        primary_port=_OPENCLAUDE_PRIMARY_PORT,
+        fallback_port=_OPENCLAUDE_FALLBACK_PORT,
+        primary_label="DigitalOcean",
+        fallback_label="OpenRouter",
+        primary_model=_OPENCLAUDE_PRIMARY_MODEL,
+        fallback_model=_OPENCLAUDE_FALLBACK_MODEL,
+        timeout=int(os.getenv("OPENCLAUDE_TIMEOUT", "600")),
+        log_fn=ui_log,
+    )
 
 
 # ──────────────────────────────────────────────────────────────

entrypoint.sh

@@ -0,0 +1,57 @@
+#!/usr/bin/env bash
+# ─────────────────────────────────────────────────────────────────────
+# Rhodawk runtime bootstrap.
+#
+# 1. Launch the OpenClaude headless gRPC daemon for the DigitalOcean
+#    Inference provider on :50051 (PRIMARY).
+# 2. Launch the OpenClaude headless gRPC daemon for OpenRouter on
+#    :50052 (FALLBACK) — only if OPENROUTER_API_KEY is present.
+# 3. Wait briefly for both to bind, then hand control to app.py which
+#    talks to them over gRPC.
+# ─────────────────────────────────────────────────────────────────────
+set -eo pipefail
+
+OC_DIR=/opt/openclaude
+LOG_DIR="${LOG_DIR:-/tmp}"
+mkdir -p "${LOG_DIR}"
+
+start_daemon() {
+    local label=$1 port=$2 base_url=$3 api_key=$4 model=$5
+    if [[ -z "${api_key}" ]]; then
+        echo "[entrypoint] skipping ${label} daemon — no API key"
+        return 0
+    fi
+    echo "[entrypoint] starting OpenClaude ${label} daemon on :${port}"
+    (
+        cd "${OC_DIR}"
+        CLAUDE_CODE_USE_OPENAI=1 \
+        OPENAI_API_KEY="${api_key}" \
+        OPENAI_BASE_URL="${base_url}" \
+        OPENAI_MODEL="${model}" \
+        GRPC_PORT="${port}" \
+        GRPC_HOST=0.0.0.0 \
+        OPENCLAUDE_AUTO_APPROVE=1 \
+        MCP_RUNTIME_CONFIG="${MCP_RUNTIME_CONFIG:-/tmp/mcp_runtime.json}" \
+            bun run scripts/start-grpc.ts \
+                > "${LOG_DIR}/openclaude-${label}.log" 2>&1 &
+        echo $! > "${LOG_DIR}/openclaude-${label}.pid"
+    )
+}
+
+# DigitalOcean Inference (PRIMARY)
+DO_BASE="${DO_INFERENCE_BASE_URL:-https://inference.do-ai.run/v1}"
+DO_MODEL="${DO_INFERENCE_MODEL:-llama3.3-70b-instruct}"
+start_daemon "do" 50051 "${DO_BASE}" \
+    "${DO_INFERENCE_API_KEY:-${DIGITALOCEAN_INFERENCE_KEY:-}}" \
+    "${DO_MODEL}"
+
+# OpenRouter (FALLBACK)
+OR_BASE="${OPENROUTER_BASE_URL:-https://openrouter.ai/api/v1}"
+OR_MODEL="${OPENROUTER_MODEL:-qwen/qwen-2.5-coder-32b-instruct:free}"
+start_daemon "or" 50052 "${OR_BASE}" "${OPENROUTER_API_KEY:-}" "${OR_MODEL}"
+
+# Brief settle window so the first healing call doesn't race the binder.
+# The Python client also has wait_ready() so this is just a friendly nudge.
+sleep 2
+echo "[entrypoint] launching Rhodawk orchestrator…"
+exec python -u app.py

mythos/MYTHOS_PLAN.md

@@ -1,249 +0,0 @@
-# Rhodawk: Ascending to Mythos-Level
-
-> **An Open-Source Blueprint for Superhuman AI Security**
-> Living document — every section in this plan maps to one or more concrete
-> modules under `mythos/`. This file is the canonical source-of-truth that
-> mirrors `attached_assets/rhodawk_mythos_level_plan_*.pdf` and tracks the
-> implementation status of every gap closure.
-
----
-
-## Executive Summary
-
-This document outlines a strategic and technical blueprint for transforming
-the existing Rhodawk AI DevSecOps Engine into a Claude Mythos-level
-Superhuman Agent. Leveraging the robust foundation of EmbodiedOS
-(integrating OpenClaw and Hermes Agent), this plan details the necessary
-architectural enhancements, open-source component integrations, and
-strategic shifts required to achieve autonomous, frontier-level vulnerability
-discovery and exploitation. The goal is to create a self-improving,
-multi-agent system capable of operating with the depth of reasoning,
-precision of execution, and iterative learning observed in Anthropic's
-unreleased Claude Mythos project, all while adhering to a cost-effective,
-open-source model strategy.
-
----
-
-## 1. Understanding Claude Mythos: A Frontier-Level AI Security Agent
-
-### 1.1 What Claude Mythos Is
-
-Claude Mythos is a sophisticated, integrated AI agent designed to operate as
-an **autonomous vulnerability research pipeline**. It moves beyond traditional
-static analysis or human-driven penetration testing by combining advanced AI
-reasoning with dynamic execution and iterative learning.
-
-### 1.2 Frontier-Level Capabilities
-
-| Capability | Description |
-|---|---|
-| Autonomous Vulnerability Research | Discovers novel zero-day vulnerabilities and generates working exploits with no prior knowledge. |
-| Elite Cybersecurity Expertise | Deep understanding of memory-safety, complex logic flaws, and subtle input-handling bugs. |
-| Sophisticated Exploit Synthesis | ROP chains, heap sprays, privilege-escalation chains, full PoC code. |
-| Self-Improving Discovery | Closed-loop hypothesis → execute → learn → refine cycle. |
-
-### 1.3 Architecture and Working Mechanism
-
-1. **Static + Semantic Code Analysis** — AST/CFG/CPG parsing.
-2. **Hypothesis Generation Engine** — probabilistic reasoning over attack vectors.
-3. **Dynamic Execution & Instrumentation** — sandboxed fuzzing + symbolic exec.
-4. **Exploit Synthesis Engine** — primitive identification + PoC code.
-5. **Autonomous Iteration Loop** — CEGIS-style continuous refinement.
-
-### 1.4 What Makes It Special
-
-- Unprecedented reasoning depth (beyond Claude 3.5 Opus class).
-- Agentic integration (plan → execute → observe → learn).
-- Stub-and-overlay architecture for cybersecurity specialization.
-- Dedicated Project Glasswing focus.
-
----
-
-## 2. Rhodawk and EmbodiedOS: The Current Foundation
-
-### 2.1 EmbodiedOS — The Unified Runtime
-
-Persistent stateful Linux workspace, multi-tier memory (short / Skill /
-Knowledge), tool-calling autonomy, CEGIS loop. Hosts both **OpenClaw**
-(local gateway, 50+ integrations, browser/file/script access) and **Hermes
-Agent** (FTS5 SQLite memory, autonomous skill creation, Atropos self-training,
-MCP server mode, Tirith pre-execution scanner).
-
-### 2.2 Rhodawk — The Superhuman Agent Framework
-
-- **Hermes Orchestrator** — six-phase pipeline (RECON → STATIC → DYNAMIC → EXPLOIT → CONSENSUS → DISCLOSURE).
-- **Red Team CEGIS Engine** — zero-day discovery + Blue Team handoff.
-- **Data Flywheel** — Training Store, Embedding Memory (MiniLM/CodeBERT), LoRA Scheduler.
-- **Bounty Gateway** — HackerOne / Bugcrowd submission.
-- **Tiered models** — Tier 1: DeepSeek 3.2 / MiniMax 2.5 · Tier 2: Qwen 2.5 Coder 32B · Tier 3: Llama 3.3 70B + DeepSeek V3 + Gemma 2 27B.
-
----
-
-## 3. Gap Analysis: Rhodawk vs. Claude Mythos
-
-| Capability Area                   | Claude Mythos (Frontier)                              | Rhodawk (Current)                              | Gap → Closure Module |
-|---|---|---|---|
-| Reasoning & Planning              | Probabilistic, multi-step, attack-graph-aware         | Deterministic 6-phase pipeline                 | `mythos/reasoning/probabilistic.py`, `mythos/reasoning/attack_graph.py`, `mythos/agents/planner.py` |
-| Static Analysis                   | Deep semantic CPG queries                              | Pattern-based taint + CWE                      | `mythos/static/treesitter_cpg.py`, `mythos/static/joern_bridge.py`, `mythos/static/codeql_bridge.py`, `mythos/static/semgrep_bridge.py` |
-| Dynamic Execution                 | Concolic + full-system + fine-grained instrumentation  | Property-based fuzzing                         | `mythos/dynamic/aflpp_runner.py`, `mythos/dynamic/klee_runner.py`, `mythos/dynamic/qemu_harness.py`, `mythos/dynamic/frida_instr.py`, `mythos/dynamic/gdb_automation.py` |
-| Exploit Synthesis                 | ROP/heap/privesc full chains                           | Primitive reasoning only                       | `mythos/exploit/pwntools_synth.py`, `mythos/exploit/rop_chain.py`, `mythos/exploit/heap_exploit.py`, `mythos/exploit/privesc_kb.py` |
-| Self-Improvement                  | RL + curriculum + episodic memory                      | LoRA Scheduler                                 | `mythos/learning/rl_planner.py`, `mythos/learning/curriculum.py`, `mythos/learning/episodic_memory.py`, `mythos/learning/mlflow_tracker.py`, `mythos/learning/lora_adapters.py` |
-| Multi-Agent Coordination          | Decoupled Planner/Explorer/Executor                    | Single orchestrator                            | `mythos/agents/{planner,explorer,executor,orchestrator}.py` |
-| MCP Surface                       | Specialised servers per analysis domain                | Generic MCP suite                              | `mythos/mcp/{static,dynamic,exploit,vuln_db,web_security}_*_mcp.py` |
-| Productization                    | Stable API for external consumption                    | Gradio UI                                      | `mythos/api/fastapi_server.py`, `mythos/api/{auth,webhooks,schemas}.py` |
-
----
-
-## 4. Open-Source Components and Models — Closing Every Gap
-
-### 4.1 Enhanced Reasoning and Planning
-**Models** — DeepSeek-V2 (MoE), Qwen2-72B-Instruct, Mixtral 8×22B.
-**Probabilistic frameworks** — Pyro (Uber AI), PyMC.
-→ `mythos/reasoning/probabilistic.py`
-
-### 4.2 Advanced Static & Semantic Code Analysis
-- **Tree-sitter** — CST/AST → CFG seed.
-- **CodeQL (open components)** — semantic queries.
-- **Joern** — Code Property Graphs.
-- **Semgrep** — taint + dataflow rules.
-- **CodeHawk** — binary value analysis (inspirational).
-→ `mythos/static/*.py`
-
-### 4.3 Enhanced Dynamic Execution & Instrumentation
-- **AFL++**, **LibFuzzer** — coverage-guided fuzzing.
-- **KLEE**, **Angr** — symbolic + concolic execution.
-- **QEMU** — full-system emulation.
-- **Frida**, **GDB+Python** — instrumentation.
-→ `mythos/dynamic/*.py`
-
-### 4.4 Sophisticated Exploit Synthesis
-- **Pwntools**, **ROPGadget**, **angrop** — ROP / shellcode.
-- **GEF** — heap visualization & manipulation.
-- **LinPEAS / WinPEAS** codified into agent skills — privesc.
-→ `mythos/exploit/*.py`
-
-### 4.5 Autonomous Iteration & Self-Improvement
-- **Ray RLlib**, **Stable Baselines3** — RL controllers.
-- **MLflow** — experiment tracking.
-- **PEFT / LoRA / QLoRA** — Tier 2 adapters.
-- **Synthetic data generation** — curriculum-driven trajectories.
-→ `mythos/learning/*.py`
-
-### 4.6 Multi-Agent Coordination
-- **AutoGen**, **CrewAI** — orchestration frameworks.
-- **MCP** — inter-agent transport.
-→ `mythos/agents/orchestrator.py`
-
-### 4.7 Cost-Effective Tiered Model Strategy
-| Tier | Role | Open-source models |
-|---|---|---|
-| 1 | Strategy & deep reasoning | DeepSeek-V2, Qwen2-72B-Instruct, Mixtral 8×22B |
-| 2 | Execution & code generation | Qwen 2.5 Coder 72B, CodeLlama-70B-Instruct |
-| 3 | Consensus & adversarial review | Llama 3.3 70B, DeepSeek V3, Gemma 2 27B |
-
-### 4.8 New MCP Servers
-- `static-analysis-mcp` — Joern + CodeQL + Semgrep.
-- `dynamic-analysis-mcp` — AFL++ + KLEE + Frida + GDB.
-- `exploit-generation-mcp` — Pwntools + ROPGadget + heap kit.
-- `vulnerability-database-mcp` — NVD + Exploit-DB + private KB.
-- `web-security-mcp` — OWASP ZAP + custom web fuzzers.
-→ `mythos/mcp/*.py` (and registered in `mcp_config.json`).
-
----
-
-## 5. Achieving Mythos-Level Capabilities — Detailed Approach
-
-### 5.1 Hierarchical Reasoning
-- **Planner Agent** — strategic, problem decomposition, hypothesis generation, attack-graph synthesis, resource allocation.
-- **Explorer Agent** — tactical static analysis.
-- **Executor Agent** — tactical dynamic execution + exploit synthesis.
-- **Contextual Awareness** — every agent shares a rich, structured context bag.
-
-### 5.2 Tool Use
-- Dynamic orchestration over MCP suite.
-- Fine-grained tool control (GDB stepping, breakpoint injection, fuzzer parameterisation).
-- Tool-augmented reasoning (every tool output mutates the context).
-- Custom-tool synthesis on the fly.
-
-### 5.3 Memory
-- **Working Memory** — EmbodiedOS persistent workspace.
-- **Skill Memory** — `agentskills.io` registry, autonomous additions.
-- **Knowledge Memory** — vector store of CS literature, RFCs, exploit write-ups.
-- **Episodic Memory** — full campaign traces (`mythos/learning/episodic_memory.py`).
-
-### 5.4 Self-Improvement
-- Continuous LoRA fine-tuning on (success, failure) pairs.
-- Reinforcement Learning over the Planner via Ray RLlib.
-- Curriculum learning — progressively harder targets.
-
-### 5.5 Multi-Agent Coordination
-- Orchestrator (`mythos/agents/orchestrator.py`) wraps AutoGen/CrewAI semantics.
-- Strict typed messages between agents (Pydantic models in `mythos/api/schemas.py`).
-- Conflict resolution via Tier 3 consensus.
-
-### 5.6 Security & Sandboxing
-- Container hardening (Tirith pre-exec scanner).
-- Network segmentation between LLM, exploit, and analysis layers.
-- Strict input validation at every API boundary.
-
----
-
-## 6. Implementation Roadmap
-
-| Phase | Months | Objective | Deliverables |
-|---|---|---|---|
-| 1 | 1–3 | Foundation & core agents | Multi-agent orchestrator, basic Planner, initial MCP servers, Tier 1 LLM |
-| 2 | 4–6 | Advanced tooling + CEGIS loop | Joern/CodeQL/Semgrep, AFL++/KLEE/QEMU, Tier 2 LLMs |
-| 3 | 7–9 | Exploit synthesis & self-improvement | exploit-generation-mcp, RL planner, LoRA adapters, MLflow |
-| 4 | 10–12 | Productization & API | FastAPI server, OAuth2/API-keys, webhooks, observability |
-
-Every roadmap item is implemented or stubbed with a clear `TODO(mythos)` in the
-corresponding module so that integrators can `grep -R 'TODO(mythos)'` to find
-the remaining engineering work.
-
----
-
-## 7. Productisation — Sellable API / Service
-
-- **Rhodawk API** — `POST /v1/analyze_target` for code/binary submission.
-- **Managed Service** — dedicated tenancy, custom fine-tune.
-- **Enterprise** — air-gapped on-premise with white-glove support.
-
-Implemented in `mythos/api/fastapi_server.py` (mountable next to the existing
-Gradio UI).
-
----
-
-## 8. Cross-Reference Implementation Index
-
-| Plan Section | Module(s) |
-|---|---|
-| 1.3 Static + Semantic | `mythos/static/*` |
-| 1.3 Hypothesis Engine | `mythos/reasoning/probabilistic.py` |
-| 1.3 Dynamic Execution | `mythos/dynamic/*` |
-| 1.3 Exploit Synthesis | `mythos/exploit/*` |
-| 1.3 Iteration Loop | `mythos/agents/orchestrator.py`, `mythos/learning/rl_planner.py` |
-| 4.1–4.6 Open Source | `mythos/static/*`, `mythos/dynamic/*`, `mythos/exploit/*`, `mythos/learning/*` |
-| 4.8 New MCP servers | `mythos/mcp/*` + `mcp_config.json` extension |
-| 5.x Mythos-level | `mythos/agents/*` + `mythos/reasoning/*` |
-| 6 Roadmap | tracked here + `TODO(mythos)` markers |
-| 7 Productization | `mythos/api/*` |
-
----
-
-## 9. Cross-Check Checklist (vs. PDF Source)
-
-- [x] Executive Summary → captured (§ Executive Summary).
-- [x] Mythos capabilities & architecture → §1.
-- [x] Rhodawk/EmbodiedOS foundation → §2.
-- [x] Gap Analysis table → §3.
-- [x] Open-source closures (4.1–4.8) → §4 + modules under `mythos/`.
-- [x] Mythos-level approach (5.1–5.6) → §5 + agent/reasoning modules.
-- [x] Implementation roadmap (Phases 1–4) → §6.
-- [x] Productization & sellable API → §7 + `mythos/api/*`.
-- [x] Tiered model strategy → §4.7 (and `mythos/agents/planner.py` env-driven).
-- [x] MCP suite extension → `mcp_config.json` + `mythos/mcp/*`.
-- [x] Self-improvement (RL, MLflow, LoRA, curriculum, episodic memory) → `mythos/learning/*`.
-
-Every checkbox above corresponds to a real file in this commit; see
-`mythos/__init__.py` for the canonical export surface.

openclaude_grpc/__init__.py

@@ -0,0 +1,21 @@
+"""
+openclaude_grpc — Python bridge to the vendored OpenClaude headless gRPC daemon.
+
+Public surface:
+    OpenClaudeClient       — low-level bidi-streaming client
+    run_openclaude         — drop-in replacement for the legacy ``run_aider``
+                              function (returns ``(combined_output, exit_code)``)
+"""
+from .client import (
+    OpenClaudeClient,
+    OpenClaudeError,
+    OpenClaudeResult,
+    run_openclaude,
+)
+
+__all__ = [
+    "OpenClaudeClient",
+    "OpenClaudeError",
+    "OpenClaudeResult",
+    "run_openclaude",
+]

openclaude_grpc/client.py

@@ -0,0 +1,348 @@
+"""
+OpenClaude gRPC client — replaces the legacy aider subprocess shell-out.
+
+Design contract
+---------------
+Every callable in this module returns an :class:`OpenClaudeResult` (or the
+``(combined_output, exit_code)`` tuple that legacy callers expect) so that the
+existing validation, SAST gate, conviction engine, and red-team loops keep
+plugging in unchanged.
+
+Streaming events from the daemon (text chunks, tool start/result, action
+required) are accumulated into a single transcript so that downstream code
+that previously parsed aider stdout still sees a useful blob.
+
+Operational guarantees
+----------------------
+* Connection drops, ``StatusCode.UNAVAILABLE`` and stream-level RPC errors
+  are converted into a non-zero exit code with the error message embedded in
+  the combined output (mirrors aider's crash semantics).
+* Per-call wall-clock timeout (defaults to 600 s) — same as the legacy
+  aider invocation.
+* Bidi stream auto-answers any ``ActionRequired`` prompt with ``"y"`` so that
+  headless mode never deadlocks waiting for human input. The daemon also
+  honours ``OPENCLAUDE_AUTO_APPROVE=1`` server-side as a belt-and-braces.
+* Each call produces a fresh stream — sessions are not shared across calls
+  to keep failures isolated.
+"""
+from __future__ import annotations
+
+import logging
+import os
+import queue
+import threading
+import time
+from dataclasses import dataclass, field
+from typing import Iterable, Optional
+
+import grpc
+
+from . import openclaude_pb2 as pb  # type: ignore
+from . import openclaude_pb2_grpc as pb_grpc  # type: ignore
+
+logger = logging.getLogger("openclaude_grpc")
+
+DEFAULT_HOST = os.getenv("OPENCLAUDE_GRPC_HOST", "127.0.0.1")
+DEFAULT_PORT_DO = int(os.getenv("OPENCLAUDE_GRPC_PORT_DO", "50051"))
+DEFAULT_PORT_OR = int(os.getenv("OPENCLAUDE_GRPC_PORT_OR", "50052"))
+DEFAULT_TIMEOUT = int(os.getenv("OPENCLAUDE_TIMEOUT", "600"))
+
+
+class OpenClaudeError(RuntimeError):
+    """Raised for unrecoverable client-side errors (connect, decode, etc)."""
+
+
+@dataclass
+class OpenClaudeResult:
+    """Mirror of the legacy aider return contract.
+
+    ``stdout`` is the model's narrative + tool stdout, ``stderr`` is tool
+    failures and gRPC errors, ``exit_code`` is 0 on a clean ``done`` event,
+    non-zero on any error event or transport failure. ``model_used`` lets
+    downstream telemetry attribute findings to the right provider.
+    """
+
+    stdout: str = ""
+    stderr: str = ""
+    exit_code: int = 0
+    model_used: str = ""
+    prompt_tokens: int = 0
+    completion_tokens: int = 0
+    tool_calls: list[dict] = field(default_factory=list)
+
+    @property
+    def combined_output(self) -> str:
+        if self.stderr:
+            return f"{self.stdout}\n{self.stderr}".strip()
+        return self.stdout.strip()
+
+    def as_legacy_tuple(self) -> tuple[str, int]:
+        return self.combined_output, self.exit_code
+
+
+class OpenClaudeClient:
+    """Bidirectional-streaming client for one OpenClaude daemon."""
+
+    def __init__(
+        self,
+        host: str = DEFAULT_HOST,
+        port: int = DEFAULT_PORT_DO,
+        timeout: int = DEFAULT_TIMEOUT,
+        max_message_mb: int = 64,
+    ) -> None:
+        self.host = host
+        self.port = port
+        self.timeout = timeout
+        opts = [
+            ("grpc.max_send_message_length", max_message_mb * 1024 * 1024),
+            ("grpc.max_receive_message_length", max_message_mb * 1024 * 1024),
+            ("grpc.keepalive_time_ms", 30_000),
+            ("grpc.keepalive_timeout_ms", 10_000),
+            ("grpc.keepalive_permit_without_calls", 1),
+        ]
+        self._channel = grpc.insecure_channel(f"{host}:{port}", options=opts)
+        self._stub = pb_grpc.AgentServiceStub(self._channel)
+
+    # ------------------------------------------------------------------
+    # health
+    # ------------------------------------------------------------------
+    def wait_ready(self, deadline_s: float = 60.0) -> bool:
+        """Block until the daemon's gRPC channel is READY or the deadline
+        elapses. Used by the orchestrator at boot to ensure the bun-built
+        daemon is alive before issuing the first healing call."""
+        start = time.monotonic()
+        while time.monotonic() - start < deadline_s:
+            try:
+                grpc.channel_ready_future(self._channel).result(timeout=2.0)
+                return True
+            except grpc.FutureTimeoutError:
+                continue
+            except Exception:
+                time.sleep(0.5)
+        return False
+
+    # ------------------------------------------------------------------
+    # core call
+    # ------------------------------------------------------------------
+    def chat(
+        self,
+        message: str,
+        working_directory: str,
+        model: str = "",
+        session_id: str = "",
+        timeout: Optional[int] = None,
+    ) -> OpenClaudeResult:
+        """Send one prompt, drain the bidi stream, return an aggregated
+        :class:`OpenClaudeResult`."""
+        deadline = timeout if timeout is not None else self.timeout
+        result = OpenClaudeResult(model_used=model)
+
+        outbound: "queue.Queue[Optional[pb.ClientMessage]]" = queue.Queue()
+        outbound.put(
+            pb.ClientMessage(
+                request=pb.ChatRequest(
+                    message=message,
+                    working_directory=working_directory,
+                    model=model,
+                    session_id=session_id,
+                )
+            )
+        )
+
+        def _send_iter() -> Iterable[pb.ClientMessage]:
+            while True:
+                item = outbound.get()
+                if item is None:
+                    return
+                yield item
+
+        done = threading.Event()
+        try:
+            stream = self._stub.Chat(_send_iter(), timeout=deadline)
+            for ev in stream:
+                kind = ev.WhichOneof("event")
+                if kind == "text_chunk":
+                    result.stdout += ev.text_chunk.text
+                elif kind == "tool_start":
+                    result.tool_calls.append(
+                        {
+                            "tool": ev.tool_start.tool_name,
+                            "args": ev.tool_start.arguments_json,
+                            "id": ev.tool_start.tool_use_id,
+                        }
+                    )
+                    result.stdout += (
+                        f"\n[tool ▶ {ev.tool_start.tool_name}]"
+                        f" {ev.tool_start.arguments_json}\n"
+                    )
+                elif kind == "tool_result":
+                    prefix = "[tool ✗]" if ev.tool_result.is_error else "[tool ✓]"
+                    line = (
+                        f"\n{prefix} {ev.tool_result.tool_name}:"
+                        f" {ev.tool_result.output}\n"
+                    )
+                    if ev.tool_result.is_error:
+                        result.stderr += line
+                    else:
+                        result.stdout += line
+                elif kind == "action_required":
+                    # Auto-approve. Server should already be in
+                    # OPENCLAUDE_AUTO_APPROVE=1, but we double-tap to make
+                    # the client safe even if the daemon was started
+                    # without that flag.
+                    outbound.put(
+                        pb.ClientMessage(
+                            input=pb.UserInput(
+                                prompt_id=ev.action_required.prompt_id,
+                                reply="y",
+                            )
+                        )
+                    )
+                elif kind == "done":
+                    if not result.stdout and ev.done.full_text:
+                        result.stdout = ev.done.full_text
+                    result.prompt_tokens = ev.done.prompt_tokens
+                    result.completion_tokens = ev.done.completion_tokens
+                    result.exit_code = 0
+                    done.set()
+                    break
+                elif kind == "error":
+                    result.stderr += (
+                        f"\n[openclaude {ev.error.code}] {ev.error.message}"
+                    )
+                    result.exit_code = 1
+                    done.set()
+                    break
+        except grpc.RpcError as exc:
+            result.stderr += (
+                f"\n[grpc {exc.code().name}] {exc.details() or str(exc)}"
+            )
+            result.exit_code = 2
+        except Exception as exc:  # noqa: BLE001
+            result.stderr += f"\n[client] {type(exc).__name__}: {exc}"
+            result.exit_code = 3
+        finally:
+            outbound.put(None)
+            try:
+                self._channel  # keep alive — see close()
+            except Exception:
+                pass
+            if not done.is_set() and result.exit_code == 0:
+                # Stream ended without an explicit done event.
+                result.exit_code = 4
+                result.stderr += "\n[client] stream ended without done event"
+
+        return result
+
+    def close(self) -> None:
+        try:
+            self._channel.close()
+        except Exception:
+            pass
+
+    def __enter__(self) -> "OpenClaudeClient":
+        return self
+
+    def __exit__(self, *exc) -> None:
+        self.close()
+
+
+# ──────────────────────────────────────────────────────────────────────
+# High-level helper — drop-in replacement for the legacy ``run_aider``
+# ──────────────────────────────────────────────────────────────────────
+def _format_prompt(prompt: str, context_files: list[str]) -> str:
+    """Aider received context files as CLI args; OpenClaude gets them
+    inline so it knows which files matter. The agent already has full
+    file-tool access via MCP/native tools, so this is just hinting."""
+    if not context_files:
+        return prompt
+    file_list = "\n".join(f"- {p}" for p in context_files)
+    return (
+        f"{prompt}\n\n"
+        "── Context files (focus your edits here) ──\n"
+        f"{file_list}\n"
+    )
+
+
+def run_openclaude(
+    mcp_config_path: str,
+    prompt: str,
+    context_files: list[str],
+    *,
+    repo_dir: str,
+    primary_port: int = DEFAULT_PORT_DO,
+    fallback_port: int = DEFAULT_PORT_OR,
+    primary_label: str = "DigitalOcean",
+    fallback_label: str = "OpenRouter",
+    primary_model: str = "",
+    fallback_model: str = "",
+    timeout: int = DEFAULT_TIMEOUT,
+    log_fn=None,
+) -> tuple[str, int]:
+    """Drop-in replacement for ``run_aider`` — preserves the
+    ``(combined_output, exit_code)`` return shape so every caller in
+    ``app.py`` works unchanged.
+
+    Tries the **primary** daemon (DigitalOcean Inference) first; on any
+    non-zero exit code it falls back to the **OpenRouter** daemon so the
+    healing loop's existing 15-attempt retry semantics still apply.
+
+    ``mcp_config_path`` is honoured by the daemon itself — it re-reads
+    ``MCP_RUNTIME_CONFIG`` per chat session — so we just forward the
+    path via env and pass the prompt verbatim.
+    """
+    # Make sure the daemon picks up the MCP file the orchestrator just
+    # wrote. The daemon was launched with this env var; this assignment
+    # is here only as documentation / safety net for ad-hoc test runs.
+    os.environ.setdefault("MCP_RUNTIME_CONFIG", mcp_config_path)
+
+    full_prompt = _format_prompt(prompt, context_files)
+
+    chain: list[tuple[int, str, str]] = []
+    if primary_port:
+        chain.append((primary_port, primary_label, primary_model))
+    if fallback_port and fallback_port != primary_port:
+        chain.append((fallback_port, fallback_label, fallback_model))
+
+    last_output, last_code = "", 1
+    for idx, (port, label, model) in enumerate(chain):
+        if log_fn:
+            log_fn(
+                f"OpenClaude attempt via {label} (port {port}, model={model or 'default'})",
+                "INFO",
+            )
+        try:
+            with OpenClaudeClient(
+                host=DEFAULT_HOST, port=port, timeout=timeout
+            ) as client:
+                if not client.wait_ready(deadline_s=15.0):
+                    msg = (
+                        f"OpenClaude daemon at {DEFAULT_HOST}:{port} "
+                        "not ready within 15s"
+                    )
+                    if log_fn:
+                        log_fn(msg, "WARN")
+                    last_output, last_code = msg, 5
+                    continue
+                result = client.chat(
+                    message=full_prompt,
+                    working_directory=repo_dir,
+                    model=model,
+                    timeout=timeout,
+                )
+        except Exception as exc:  # noqa: BLE001
+            last_output = f"[client] {type(exc).__name__}: {exc}"
+            last_code = 6
+            if log_fn:
+                log_fn(f"{label} client crash: {exc}", "FAIL")
+            continue
+
+        last_output, last_code = result.as_legacy_tuple()
+        if last_code == 0:
+            return last_output, last_code
+        if log_fn and idx < len(chain) - 1:
+            log_fn(
+                f"{label} failed (exit {last_code}) — falling back",
+                "WARN",
+            )
+    return last_output, last_code

requirements.txt

@@ -4,11 +4,14 @@ uv>=0.7.0
 gitpython==3.1.46
 gradio>=5.49.0,<6
 jinja2==3.1.6
-# aider-chat is intentionally NOT listed here — it hard-pins
-# pillow==12.1.1 which conflicts with gradio (pillow<12) and litellm==1.75.0
-# which has a broken module surface (missing APIConnectionError, _logging).
-# The Dockerfile installs aider with --no-deps after this requirements file
-# resolves, plus a patched litellm and aider's actual runtime deps.
+
+# ─── Code-generation layer ───────────────────────────────────────────────
+# Aider-chat has been removed and replaced by the vendored OpenClaude
+# headless gRPC daemon (see vendor/openclaude/ + openclaude_grpc/).
+# The Dockerfile installs grpcio + grpcio-tools at build time so that
+# protobuf stub generation runs inside the build, not in this lock file.
+
+# ─── Static analysis / SAST ──────────────────────────────────────────────
 ruff
 tenacity
 bandit[toml]
@@ -16,6 +19,8 @@ pip-audit
 radon
 hypothesis[cli]>=6.100.0
 semgrep>=1.45.0
+
+# ─── ML / embeddings / vector store ──────────────────────────────────────
 sentence-transformers>=2.7.0
 sqlite-vec>=0.1.1
 pygithub>=2.3.0
@@ -28,35 +33,24 @@ z3-solver>=4.12.0
 qdrant-client>=1.9.0
 transformers>=4.40.0
 torch>=2.2.0
-# atheris removed: requires Clang + libFuzzer at compile time which is unavailable
-# on HuggingFace Space Docker images. Fuzzing falls back to Hypothesis automatically.
+
+# atheris removed: requires Clang + libFuzzer at compile time which is
+# unavailable on HuggingFace Space Docker images. Fuzzing falls back to
+# Hypothesis automatically.
 angr>=9.2.0
 networkx>=3.0
 defusedxml>=0.7.1
 
 # ─── Mythos-level upgrade (see mythos/MYTHOS_PLAN.md) ────────────────────
-# Required for the productization API surface.
 fastapi>=0.110.0
 uvicorn[standard]>=0.27.0
 pydantic>=2.6.0
-# Optional: heavier capabilities — install on demand for full Mythos parity.
-# pyro-ppl>=1.9.0           # § 4.1 probabilistic reasoning (Bayesian backend)
-# pymc>=5.10.0              # § 4.1 probabilistic reasoning (alt backend)
-# tree-sitter-languages>=1.10.0  # § 4.2 Tree-sitter CPG bridge
-# mlflow>=2.12.0            # § 4.5 experiment tracking
-# ray[rllib]>=2.10.0        # § 4.5 RL planner backend
-# stable-baselines3>=2.3.0  # § 4.5 RL planner backend (alt)
-# pwntools>=4.12.0          # § 4.4 exploit synthesis (Linux only)
-# frida>=16.0.0             # § 4.3 dynamic instrumentation
 
 # ─── ARCHITECT masterplan dependencies (see ARCHITECT_MASTERPLAN.md) ─────
-# Required at runtime in light-mode (no extra binaries):
 dnspython>=2.6.0
-# Optional heavy bridges — install when moving to the paid VPS / lab rig:
-# playwright>=1.44.0           # browser-agent-mcp (run `playwright install chromium`)
-# python-can>=4.3.0            # can-bus-mcp (automotive)
-# scapy>=2.5.0                 # network-protocol fuzzing
-# pymodbus>=3.6.0              # ics-scada
-# python-snap7>=1.3            # ics-scada (Siemens S7)
-# opcua>=0.98.13               # ics-scada (OPC-UA)
-# rtl-sdr / hackrf-host (system pkgs) for sdr-analysis-mcp
+
+# ─── gRPC bridge to the OpenClaude daemon ────────────────────────────────
+# (Pinned again here as a safety net for editable installs / `pip install -r`
+#  in environments that bypass the Dockerfile.)
+grpcio>=1.66.0,<2.0.0
+protobuf>=5.0.0,<6.0.0

vendor/openclaude/.dockerignore

@@ -0,0 +1,16 @@
+node_modules
+dist
+.git
+.gitignore
+.env
+.env.*
+!.env.example
+coverage
+reports
+vscode-extension
+python
+docs
+*.md
+!README.md
+.github
+.tsbuildinfo

vendor/openclaude/.env.example

@@ -0,0 +1,374 @@
+# =============================================================================
+# OpenClaude Environment Configuration
+# =============================================================================
+# Copy this file to .env and fill in your values:
+#   cp .env.example .env
+#
+# Only set the variables for the provider you want to use.
+# All other sections can be left commented out.
+# =============================================================================
+
+# =============================================================================
+# SYSTEM-WIDE SETUP (OPTIONAL)
+# =============================================================================
+# Instead of using a .env file per project, you can set these variables
+# system-wide so OpenClaude works from any directory on your machine.
+#
+# STEP 1: Pick your provider variables from the list below.
+# STEP 2: Set them using the method for your OS (see further down).
+#
+# ── Provider variables ───────────────────────────────────────────────
+#
+#   Option 1 — Anthropic:
+#     ANTHROPIC_API_KEY=sk-ant-your-key-here
+#     ANTHROPIC_MODEL=claude-sonnet-4-5              (optional)
+#     ANTHROPIC_BASE_URL=https://api.anthropic.com   (optional)
+#
+#   Option 2 — OpenAI:
+#     CLAUDE_CODE_USE_OPENAI=1
+#     OPENAI_API_KEY=sk-your-key-here
+#     OPENAI_MODEL=gpt-4o
+#     OPENAI_BASE_URL=https://api.openai.com/v1      (optional)
+#
+#   Option 3 — Google Gemini:
+#     CLAUDE_CODE_USE_GEMINI=1
+#     GEMINI_API_KEY=your-gemini-key-here
+#     GEMINI_MODEL=gemini-2.0-flash
+#     GEMINI_BASE_URL=https://generativelanguage.googleapis.com  (optional)
+#
+#   Option 4 — GitHub Models:
+#     CLAUDE_CODE_USE_GITHUB=1
+#     GITHUB_TOKEN=ghp_your-token-here
+#
+#   Option 5 — Ollama (local):
+#     CLAUDE_CODE_USE_OPENAI=1
+#     OPENAI_BASE_URL=http://localhost:11434/v1
+#     OPENAI_API_KEY=ollama
+#     OPENAI_MODEL=llama3.2
+#
+#   Option 6 — LM Studio (local):
+#     CLAUDE_CODE_USE_OPENAI=1
+#     OPENAI_BASE_URL=http://localhost:1234/v1
+#     OPENAI_MODEL=your-model-id-here
+#     OPENAI_API_KEY=lmstudio                        (optional)
+#
+#   Option 7 — AWS Bedrock (may also need: aws configure):
+#     CLAUDE_CODE_USE_BEDROCK=1
+#     AWS_REGION=us-east-1
+#     AWS_DEFAULT_REGION=us-east-1
+#     AWS_BEARER_TOKEN_BEDROCK=your-bearer-token-here
+#     ANTHROPIC_BEDROCK_BASE_URL=https://bedrock-runtime.us-east-1.amazonaws.com
+#
+#   Option 8 — Google Vertex AI:
+#     CLAUDE_CODE_USE_VERTEX=1
+#     ANTHROPIC_VERTEX_PROJECT_ID=your-gcp-project-id
+#     CLOUD_ML_REGION=us-east5
+#     GOOGLE_CLOUD_PROJECT=your-gcp-project-id
+#
+# ── How to set variables on each OS ──────────────────────────────────
+#
+# macOS (zsh):
+#   1. Open:  nano ~/.zshrc
+#   2. Add each variable as:  export VAR_NAME=value
+#   3. Save and reload:  source ~/.zshrc
+#
+# Linux (bash):
+#   1. Open:  nano ~/.bashrc
+#   2. Add each variable as:  export VAR_NAME=value
+#   3. Save and reload:  source ~/.bashrc
+#
+# Windows (PowerShell):
+#   Run for each variable:
+#     [System.Environment]::SetEnvironmentVariable('VAR_NAME', 'value', 'User')
+#   Then restart your terminal.
+#
+# Windows (Command Prompt):
+#   Run for each variable:
+#     setx VAR_NAME value
+#   Then restart your terminal.
+#
+# Windows (GUI):
+#   Settings > System > About > Advanced System Settings >
+#   Environment Variables > under "User variables" click New,
+#   then add each variable.
+#
+# ── Important notes ──────────────────────────────────────────────────
+#
+# LOCAL SERVERS: If using LM Studio or Ollama, the server MUST be
+# running with a model loaded before you launch OpenClaude —
+# otherwise you'll get connection errors.
+#
+# SWITCHING PROVIDERS: To temporarily switch, unset the relevant
+# variables in your current terminal session:
+#
+#   macOS / Linux:
+#     unset VAR_NAME
+#     # e.g.: unset CLAUDE_CODE_USE_OPENAI OPENAI_BASE_URL OPENAI_MODEL
+#
+#   Windows (PowerShell — current session only):
+#     Remove-Item Env:VAR_NAME
+#
+#   To permanently remove a variable on Windows:
+#     [System.Environment]::SetEnvironmentVariable('VAR_NAME', $null, 'User')
+#
+# LOAD ORDER:
+#   Shell and system environment variables are inherited by the process.
+#   Project .env files are only used if your launcher or shell loads them
+#   before starting OpenClaude.
+# COMPATIBILITY:
+#   System-wide variables work regardless of how you run OpenClaude:
+#   npx, global npm install, bun run, or node directly. Any process
+#   launched from your terminal inherits your shell's environment.
+#
+# REMINDER: Make sure .env is in your .gitignore to avoid committing secrets.
+# =============================================================================
+
+# =============================================================================
+# PROVIDER SELECTION — uncomment ONE block below
+# =============================================================================
+
+# -----------------------------------------------------------------------------
+# Option 1: Anthropic (default — no provider flag needed)
+# -----------------------------------------------------------------------------
+ANTHROPIC_API_KEY=sk-ant-your-key-here
+
+# Override the default model (optional)
+# ANTHROPIC_MODEL=claude-sonnet-4-5
+
+# Use a custom Anthropic-compatible endpoint (optional)
+# ANTHROPIC_BASE_URL=https://api.anthropic.com
+
+
+# -----------------------------------------------------------------------------
+# Option 2: OpenAI
+# -----------------------------------------------------------------------------
+# CLAUDE_CODE_USE_OPENAI=1
+# OPENAI_API_KEY=sk-your-key-here
+# OPENAI_MODEL=gpt-4o
+
+# Use a custom OpenAI-compatible endpoint (optional — defaults to api.openai.com)
+# OPENAI_BASE_URL=https://api.openai.com/v1
+
+
+# -----------------------------------------------------------------------------
+# Option 3: Google Gemini
+# -----------------------------------------------------------------------------
+# CLAUDE_CODE_USE_GEMINI=1
+# GEMINI_API_KEY=your-gemini-key-here
+# GEMINI_MODEL=gemini-2.0-flash
+
+# Use a custom Gemini endpoint (optional)
+# GEMINI_BASE_URL=https://generativelanguage.googleapis.com/v1beta/openai
+
+
+# -----------------------------------------------------------------------------
+# Option 4: GitHub Models
+# -----------------------------------------------------------------------------
+# CLAUDE_CODE_USE_GITHUB=1
+# GITHUB_TOKEN=ghp_your-token-here
+
+
+# -----------------------------------------------------------------------------
+# Option 5: Ollama (local models)
+# -----------------------------------------------------------------------------
+# CLAUDE_CODE_USE_OPENAI=1
+# OPENAI_BASE_URL=http://localhost:11434/v1
+# OPENAI_API_KEY=ollama
+# OPENAI_MODEL=llama3.2
+
+# -----------------------------------------------------------------------------
+# Option 6: LM Studio (local models)
+# -----------------------------------------------------------------------------
+# LM Studio exposes an OpenAI-compatible API, so we use the OpenAI provider.
+# Make sure LM Studio is running with the Developer server enabled
+# (Developer tab > toggle server ON).
+#
+# Steps:
+#   1. Download and install LM Studio from https://lmstudio.ai
+#   2. Search for and download a model (e.g. any coding or instruct model)
+#   3. Load the model and start the Developer server
+#   4. Set OPENAI_MODEL to the model ID shown in LM Studio's Developer tab
+#
+# The default server URL is http://localhost:1234 — change the port below
+# if you've configured a different one in LM Studio.
+#
+# OPENAI_API_KEY is optional — LM Studio runs locally and ignores it.
+# Some clients require a non-empty value; if you get auth errors, set it
+# to any dummy value (e.g. "lmstudio").
+#
+# CLAUDE_CODE_USE_OPENAI=1
+# OPENAI_BASE_URL=http://localhost:1234/v1
+# OPENAI_MODEL=your-model-id-here
+
+
+# -----------------------------------------------------------------------------
+# Option 7: AWS Bedrock
+# -----------------------------------------------------------------------------
+
+# You may also need AWS CLI credentials configured (run: aws configure)
+# or have AWS_ACCESS_KEY_ID and AWS_SECRET_ACCESS_KEY set in your
+# environment in addition to the variables below.
+#
+# CLAUDE_CODE_USE_BEDROCK=1
+# AWS_REGION=us-east-1
+# AWS_DEFAULT_REGION=us-east-1
+# AWS_BEARER_TOKEN_BEDROCK=your-bearer-token-here
+# ANTHROPIC_BEDROCK_BASE_URL=https://bedrock-runtime.us-east-1.amazonaws.com
+
+
+# -----------------------------------------------------------------------------
+# Option 8: Google Vertex AI
+# -----------------------------------------------------------------------------
+# CLAUDE_CODE_USE_VERTEX=1
+# ANTHROPIC_VERTEX_PROJECT_ID=your-gcp-project-id
+# CLOUD_ML_REGION=us-east5
+# GOOGLE_CLOUD_PROJECT=your-gcp-project-id
+
+
+# -----------------------------------------------------------------------------
+# Option 9: NVIDIA NIM
+# -----------------------------------------------------------------------------
+# NVIDIA NIM provides hosted inference endpoints for NVIDIA models.
+# Get your API key from https://build.nvidia.com/
+#
+# CLAUDE_CODE_USE_OPENAI=1
+# NVIDIA_API_KEY=nvapi-your-key-here
+# OPENAI_BASE_URL=https://integrate.api.nvidia.com/v1
+# OPENAI_MODEL=nvidia/llama-3.1-nemotron-70b-instruct
+
+
+# -----------------------------------------------------------------------------
+# Option 10: MiniMax
+# -----------------------------------------------------------------------------
+# MiniMax API provides text generation models.
+# Get your API key from https://platform.minimax.io/
+#
+# CLAUDE_CODE_USE_OPENAI=1
+# MINIMAX_API_KEY=your-minimax-key-here
+# OPENAI_BASE_URL=https://api.minimax.io/v1
+# OPENAI_MODEL=MiniMax-M2.5
+
+
+# =============================================================================
+# OPTIONAL TUNING
+# =============================================================================
+
+# Max number of API retries on failure (default: 10)
+# CLAUDE_CODE_MAX_RETRIES=10
+
+# Enable persistent retry mode for unattended/CI sessions
+# Retries 429/529 indefinitely with smart backoff
+# CLAUDE_CODE_UNATTENDED_RETRY=1
+
+# Enable extended key reporting (Kitty keyboard protocol)
+# Useful for iTerm2, WezTerm, Ghostty if modifier keys feel off
+# OPENCLAUDE_ENABLE_EXTENDED_KEYS=1
+
+# Disable "Co-authored-by" line in git commits made by OpenClaude
+# OPENCLAUDE_DISABLE_CO_AUTHORED_BY=1
+
+# Disable strict tool schema normalization for non-Gemini providers
+# Useful when MCP tools with complex optional params (e.g. list[dict])
+# trigger "Extra required key ... supplied" errors from OpenAI-compatible endpoints
+# OPENCLAUDE_DISABLE_STRICT_TOOLS=1
+
+# Disable hidden <system-reminder> messages injected into tool output
+# Suppresses the file-read cyber-risk reminder and the todo/task tool nudges
+# Useful for users who want full transparency over what the model sees
+# OPENCLAUDE_DISABLE_TOOL_REMINDERS=1
+
+# Custom timeout for API requests in milliseconds (default: varies)
+# API_TIMEOUT_MS=60000
+
+# Enable debug logging
+# CLAUDE_DEBUG=1
+
+
+# =============================================================================
+# WEB SEARCH (OPTIONAL)
+# =============================================================================
+# OpenClaude includes a web search tool. By default it uses DuckDuckGo (free)
+# or the provider's native search (Anthropic firstParty / vertex).
+#
+# Set one API key below to enable a provider. That's it.
+
+# ── Provider API keys — set ONE of these ────────────────────────────
+
+# Tavily (AI-optimized search, recommended)
+# TAVILY_API_KEY=tvly-your-key-here
+
+# Exa (neural/semantic search)
+# EXA_API_KEY=your-exa-key-here
+
+# You.com (RAG-ready snippets)
+# YOU_API_KEY=your-you-key-here
+
+# Jina (s.jina.ai endpoint)
+# JINA_API_KEY=your-jina-key-here
+
+# Bing Web Search
+# BING_API_KEY=your-bing-key-here
+
+# Mojeek (privacy-focused)
+# MOJEEK_API_KEY=your-mojeek-key-here
+
+# Linkup
+# LINKUP_API_KEY=your-linkup-key-here
+
+# Firecrawl (premium, uses @mendable/firecrawl-js)
+# FIRECRAWL_API_KEY=fc-your-key-here
+
+# ── Provider selection mode ─────────────────────────────────────────
+#
+# WEB_SEARCH_PROVIDER controls fallback behavior:
+#
+#   "auto"      (default) — try all configured providers, fall through on failure
+#   "custom"    — custom API only, throw on failure (NOT in auto chain)
+#   "firecrawl" — firecrawl only
+#   "tavily"    — tavily only
+#   "exa"       — exa only
+#   "you"       — you.com only
+#   "jina"      — jina only
+#   "bing"      — bing only
+#   "mojeek"    — mojeek only
+#   "linkup"    — linkup only
+#   "ddg"       — duckduckgo only
+#   "native"    — anthropic native / codex only
+#
+# Auto mode priority: firecrawl → tavily → exa → you → jina → bing → mojeek →
+#                     linkup → ddg
+# Note: "custom" is NOT in the auto chain. To use the custom API provider,
+# you must explicitly set WEB_SEARCH_PROVIDER=custom.
+#
+# WEB_SEARCH_PROVIDER=auto
+
+# ── Built-in custom API presets ─────────────────────────────────────
+#
+# Use with WEB_KEY for the API key:
+# WEB_PROVIDER=searxng|google|brave|serpapi
+# WEB_KEY=your-api-key-here
+
+# ── Custom API endpoint (advanced) ──────────────────────────────────
+#
+# WEB_SEARCH_API    — base URL of your search endpoint
+# WEB_QUERY_PARAM         — query parameter name (default: "q")
+# WEB_METHOD        — GET or POST (default: GET)
+# WEB_PARAMS        — extra static query params as JSON: {"lang":"en","count":"10"}
+# WEB_URL_TEMPLATE  — URL template with {query} for path embedding
+# WEB_BODY_TEMPLATE — custom POST body with {query} placeholder
+# WEB_AUTH_HEADER   — header name for API key (default: "Authorization")
+# WEB_AUTH_SCHEME   — prefix before key (default: "Bearer")
+# WEB_HEADERS       — extra headers as "Name: value; Name2: value2"
+# WEB_JSON_PATH     — dot-path to results array in response
+
+# ── Custom API security guardrails ──────────────────────────────────
+#
+# The custom provider enforces security guardrails by default.
+# Override these only if you understand the risks.
+#
+# WEB_CUSTOM_TIMEOUT_SEC=15                  — request timeout in seconds (default 15)
+# WEB_CUSTOM_MAX_BODY_KB=300                 — max POST body size in KB (default 300)
+# WEB_CUSTOM_ALLOW_ARBITRARY_HEADERS=false   — set "true" to use non-standard headers
+# WEB_CUSTOM_ALLOW_HTTP=false                — set "true" to allow http:// URLs
+# WEB_CUSTOM_ALLOW_PRIVATE=false             — set "true" to target localhost/private IPs
+#                                               (needed for self-hosted SearXNG)

vendor/openclaude/.gitignore

@@ -0,0 +1,14 @@
+node_modules/
+dist/
+*.tsbuildinfo
+.env
+.env.*
+!.env.example
+.openclaude-profile.json
+reports/
+GEMINI.md
+CLAUDE.md
+package-lock.json
+/.claude
+coverage/
+agent.log

vendor/openclaude/.release-please-manifest.json

@@ -0,0 +1,3 @@
+{
+  ".": "0.6.0"
+}

vendor/openclaude/ANDROID_INSTALL.md

@@ -0,0 +1,162 @@
+# OpenClaude on Android (Termux)
+
+A complete guide to running OpenClaude on Android using Termux + proot Ubuntu.
+
+---
+
+## Prerequisites
+
+- Android phone with ~700MB free storage
+- [Termux](https://f-droid.org/en/packages/com.termux/) installed from **F-Droid** (not Play Store)
+- An [OpenRouter](https://openrouter.ai) API key (free, no credit card required)
+
+---
+
+## Why This Setup?
+
+OpenClaude requires [Bun](https://bun.sh) to build, and Bun does not support Android natively. The workaround is running a real Ubuntu environment inside Termux via `proot-distro`, where Bun's Linux binary works correctly.
+
+---
+
+## Installation
+
+### Step 1 — Update Termux
+
+```bash
+pkg update && pkg upgrade
+```
+
+Press `N` or Enter for any config file conflict prompts.
+
+### Step 2 — Install dependencies
+
+```bash
+pkg install nodejs-lts git proot-distro
+```
+
+Verify Node.js:
+```bash
+node --version  # should be v20+
+```
+
+### Step 3 — Clone OpenClaude
+
+```bash
+git clone https://github.com/Gitlawb/openclaude.git
+cd openclaude
+npm install
+npm link
+```
+
+### Step 4 — Install Ubuntu via proot
+
+```bash
+proot-distro install ubuntu
+```
+
+This downloads ~200–400MB. Wait for it to complete.
+
+### Step 5 — Install Bun inside Ubuntu
+
+```bash
+proot-distro login ubuntu
+curl -fsSL https://bun.sh/install | bash
+source ~/.bashrc
+bun --version  # should show 1.3.11+
+```
+
+### Step 6 — Build OpenClaude
+
+```bash
+cd /data/data/com.termux/files/home/openclaude
+bun run build
+```
+
+You should see:
+```
+✓ Built openclaude v0.1.6 → dist/cli.mjs
+```
+
+### Step 7 — Save env vars permanently
+
+Still inside Ubuntu, add your OpenRouter config to `.bashrc`:
+
+```bash
+echo 'export CLAUDE_CODE_USE_OPENAI=1' >> ~/.bashrc
+echo 'export OPENAI_API_KEY=your_openrouter_key_here' >> ~/.bashrc
+echo 'export OPENAI_BASE_URL=https://openrouter.ai/api/v1' >> ~/.bashrc
+echo 'export OPENAI_MODEL=qwen/qwen3.6-plus-preview:free' >> ~/.bashrc
+source ~/.bashrc
+```
+
+Replace `your_openrouter_key_here` with your actual key from [openrouter.ai/keys](https://openrouter.ai/keys).
+
+### Step 8 — Run OpenClaude
+
+```bash
+node dist/cli.mjs
+```
+
+Select **3** (3rd-party platform) at the login screen. Your env vars will be detected automatically.
+
+---
+
+## Restarting After Closing Termux
+
+Every time you reopen Termux after killing it, run:
+
+```bash
+proot-distro login ubuntu
+cd /data/data/com.termux/files/home/openclaude
+node dist/cli.mjs
+```
+
+---
+
+## Recommended Free Model
+
+**`qwen/qwen3.6-plus-preview:free`** — Best free model on OpenRouter as of April 2026.
+
+- 1M token context window
+- Beats Claude 4.5 Opus on Terminal-Bench 2.0 agentic coding (61.6 vs 59.3)
+- Built-in chain-of-thought reasoning
+- Native tool use and function calling
+- $0/M tokens (preview period)
+
+> ⚠️ Free status may change when the preview period ends. Check [openrouter.ai](https://openrouter.ai/qwen/qwen3.6-plus-preview:free) for current pricing.
+
+---
+
+## Alternative Free Models (OpenRouter)
+
+| Model ID | Context | Notes |
+|---|---|---|
+| `qwen/qwen3-coder:free` | 262K | Best for pure coding tasks |
+| `openai/gpt-oss-120b:free` | 131K | OpenAI open model, strong tool calling |
+| `nvidia/nemotron-3-super-120b-a12b:free` | 262K | Hybrid MoE, good general use |
+| `meta-llama/llama-3.3-70b-instruct:free` | 66K | Reliable, widely tested |
+
+Switch models anytime:
+```bash
+export OPENAI_MODEL=qwen/qwen3-coder:free
+node dist/cli.mjs
+```
+
+---
+
+## Why Not Groq or Cerebras?
+
+Both were tested and fail due to OpenClaude's large system prompt (~50K tokens):
+
+- **Groq free tier**: TPM limits too low (6K–12K tokens/min)
+- **Cerebras free tier**: TPM limits exceeded, even on `llama3.1-8b`
+
+OpenRouter free models have no TPM restrictions — only 20 req/min and 200 req/day.
+
+---
+
+## Tips
+
+- **Don't swipe Termux away** from recent apps mid-session — use the home button to minimize instead.
+- The Ubuntu environment persists between Termux sessions; your build and config are saved.
+- Run `bun run build` again only if you pull updates to the OpenClaude repo.

vendor/openclaude/CHANGELOG.md

@@ -0,0 +1,176 @@
+# Changelog
+
+## [0.6.0](https://github.com/Gitlawb/openclaude/compare/v0.5.2...v0.6.0) (2026-04-22)
+
+
+### Features
+
+* add model caching and benchmarking utilities ([#671](https://github.com/Gitlawb/openclaude/issues/671)) ([2b15e16](https://github.com/Gitlawb/openclaude/commit/2b15e16421f793f954a92c53933a07094544b29d))
+* add thinking token extraction ([#798](https://github.com/Gitlawb/openclaude/issues/798)) ([268c039](https://github.com/Gitlawb/openclaude/commit/268c0398e4bf1ab898069c61500a2b3c226a0322))
+* **api:** compress old tool_result content for small-context providers ([#801](https://github.com/Gitlawb/openclaude/issues/801)) ([a6a3de5](https://github.com/Gitlawb/openclaude/commit/a6a3de5ac155fe9d00befbfcab98d439314effd8))
+* **api:** improve local provider reliability with readiness and self-healing ([#738](https://github.com/Gitlawb/openclaude/issues/738)) ([4cb963e](https://github.com/Gitlawb/openclaude/commit/4cb963e660dbd6ee438c04042700db05a9d32c59))
+* **api:** smart model routing primitive (cheap-for-simple, strong-for-hard) ([#785](https://github.com/Gitlawb/openclaude/issues/785)) ([e908864](https://github.com/Gitlawb/openclaude/commit/e908864da7e7c987a98053ac5d18d702e192db2b))
+* enable 15 additional feature flags in open build ([#667](https://github.com/Gitlawb/openclaude/issues/667)) ([6a62e3f](https://github.com/Gitlawb/openclaude/commit/6a62e3ff76ba9ba446b8e20cf2bb139ee76a9387))
+* native Anthropic API mode for Claude models on GitHub Copilot ([#579](https://github.com/Gitlawb/openclaude/issues/579)) ([fdef4a1](https://github.com/Gitlawb/openclaude/commit/fdef4a1b4ce218ded4937ca83b30acce7c726472))
+* **provider:** expose Atomic Chat in /provider picker with autodetect ([#810](https://github.com/Gitlawb/openclaude/issues/810)) ([ee19159](https://github.com/Gitlawb/openclaude/commit/ee19159c17b3de3b4a8b4a4541a6569f4261d54e))
+* **provider:** zero-config autodetection primitive ([#784](https://github.com/Gitlawb/openclaude/issues/784)) ([a5bfcbb](https://github.com/Gitlawb/openclaude/commit/a5bfcbbadf8e9a1fd42f3e103d295524b8da64b0))
+
+
+### Bug Fixes
+
+* **api:** ensure strict role sequence and filter empty assistant messages after interruption ([#745](https://github.com/Gitlawb/openclaude/issues/745) regression) ([#794](https://github.com/Gitlawb/openclaude/issues/794)) ([06e7684](https://github.com/Gitlawb/openclaude/commit/06e7684eb56df8e694ac784575e163641931c44c))
+* Collapse all-text arrays to string for DeepSeek compatibility ([#806](https://github.com/Gitlawb/openclaude/issues/806)) ([761924d](https://github.com/Gitlawb/openclaude/commit/761924daa7e225fe8acf41651408c7cae639a511))
+* **model:** codex/nvidia-nim/minimax now read OPENAI_MODEL env ([#815](https://github.com/Gitlawb/openclaude/issues/815)) ([4581208](https://github.com/Gitlawb/openclaude/commit/458120889f6ce54cc9f0b287461d5e38eae48a20))
+* **provider:** saved profile ignored when stale CLAUDE_CODE_USE_* in shell ([#807](https://github.com/Gitlawb/openclaude/issues/807)) ([13de4e8](https://github.com/Gitlawb/openclaude/commit/13de4e85df7f5fadc8cd15a76076374dc112360b))
+* rename .claude.json to .openclaude.json with legacy fallback ([#582](https://github.com/Gitlawb/openclaude/issues/582)) ([4d4fb28](https://github.com/Gitlawb/openclaude/commit/4d4fb2880e4d0e3a62d8715e1ec13d932e736279))
+* replace discontinued gemini-2.5-pro-preview-03-25 with stable gemini-2.5-pro ([#802](https://github.com/Gitlawb/openclaude/issues/802)) ([64582c1](https://github.com/Gitlawb/openclaude/commit/64582c119d5d0278195271379da4a68d59a89c1f)), closes [#398](https://github.com/Gitlawb/openclaude/issues/398)
+* **security:** harden project settings trust boundary + MCP sanitization ([#789](https://github.com/Gitlawb/openclaude/issues/789)) ([ae3b723](https://github.com/Gitlawb/openclaude/commit/ae3b723f3b297b49925cada4728f3174aee8bf12))
+* **test:** autoCompact floor assertion is flag-sensitive ([#816](https://github.com/Gitlawb/openclaude/issues/816)) ([c13842e](https://github.com/Gitlawb/openclaude/commit/c13842e91c7227246520955de6ae0636b30def9a))
+* **ui:** prevent provider manager lag by deferring sync I/O ([#803](https://github.com/Gitlawb/openclaude/issues/803)) ([85eab27](https://github.com/Gitlawb/openclaude/commit/85eab2751e7d351bb0ed6a3fe0e15461d241c9cb))
+
+## [0.5.2](https://github.com/Gitlawb/openclaude/compare/v0.5.1...v0.5.2) (2026-04-20)
+
+
+### Bug Fixes
+
+* **api:** replace phrase-based reasoning sanitizer with tag-based filter ([#779](https://github.com/Gitlawb/openclaude/issues/779)) ([336ddcc](https://github.com/Gitlawb/openclaude/commit/336ddcc50d59d79ebff50993f2673652aecb0d7d))
+
+## [0.5.1](https://github.com/Gitlawb/openclaude/compare/v0.5.0...v0.5.1) (2026-04-20)
+
+
+### Bug Fixes
+
+* enforce Bash path constraints after sandbox allow ([#777](https://github.com/Gitlawb/openclaude/issues/777)) ([7002cb3](https://github.com/Gitlawb/openclaude/commit/7002cb302b78ea2a19da3f26226de24e2903fa1d))
+* enforce MCP OAuth callback state before errors ([#775](https://github.com/Gitlawb/openclaude/issues/775)) ([739b8d1](https://github.com/Gitlawb/openclaude/commit/739b8d1f40fde0e401a5cbd2b9a55d88bd5124ad))
+* require trusted approval for sandbox override ([#778](https://github.com/Gitlawb/openclaude/issues/778)) ([aab4890](https://github.com/Gitlawb/openclaude/commit/aab489055c53dd64369414116fe93226d2656273))
+
+## [0.5.0](https://github.com/Gitlawb/openclaude/compare/v0.4.0...v0.5.0) (2026-04-20)
+
+
+### Features
+
+* add OPENCLAUDE_DISABLE_STRICT_TOOLS env var to opt out of strict MCP tool schema normalization ([#770](https://github.com/Gitlawb/openclaude/issues/770)) ([e6e8d9a](https://github.com/Gitlawb/openclaude/commit/e6e8d9a24897e4c9ef08b72df20fabbf8ef27f38))
+* mask provider api key input ([#772](https://github.com/Gitlawb/openclaude/issues/772)) ([13e9f22](https://github.com/Gitlawb/openclaude/commit/13e9f22a83a2b0f85f557b1e12c9442ba61241e4))
+
+
+### Bug Fixes
+
+* allow provider recovery during startup ([#765](https://github.com/Gitlawb/openclaude/issues/765)) ([f828171](https://github.com/Gitlawb/openclaude/commit/f828171ef1ab94e2acf73a28a292799e4e26cc0d))
+* **api:** drop orphan tool results to satisfy strict role sequence ([#745](https://github.com/Gitlawb/openclaude/issues/745)) ([b786b76](https://github.com/Gitlawb/openclaude/commit/b786b765f01f392652eaf28ed3579a96b7260a53))
+* **help:** prevent /help tab crash from undefined descriptions ([#732](https://github.com/Gitlawb/openclaude/issues/732)) ([3d1979f](https://github.com/Gitlawb/openclaude/commit/3d1979ff066db32415e0c8321af916d81f5f2621))
+* **mcp:** sync required array with properties in tool schemas ([#754](https://github.com/Gitlawb/openclaude/issues/754)) ([002a8f1](https://github.com/Gitlawb/openclaude/commit/002a8f1f6de2fcfc917165d828501d3047bad61f))
+* remove cached mcpClient in diagnostic tracking to prevent stale references ([#727](https://github.com/Gitlawb/openclaude/issues/727)) ([2c98be7](https://github.com/Gitlawb/openclaude/commit/2c98be700274a4241963b5f43530bf3bd8f8963f))
+* use raw context window for auto-compact percentage display ([#748](https://github.com/Gitlawb/openclaude/issues/748)) ([55c5f26](https://github.com/Gitlawb/openclaude/commit/55c5f262a9a5a8be0aa9ae8dc6c7dafc465eb2c6))
+
+## [0.4.0](https://github.com/Gitlawb/openclaude/compare/v0.3.0...v0.4.0) (2026-04-17)
+
+
+### Features
+
+* add Alibaba Coding Plan (DashScope) provider support ([#509](https://github.com/Gitlawb/openclaude/issues/509)) ([43ac6db](https://github.com/Gitlawb/openclaude/commit/43ac6dba75537282da1e2ad8f855082bc4e25f1e))
+* add NVIDIA NIM and MiniMax provider support ([#552](https://github.com/Gitlawb/openclaude/issues/552)) ([51191d6](https://github.com/Gitlawb/openclaude/commit/51191d61326e1f8319d70b3a3c0d9229e185a564))
+* add ripgrep to Dockerfile for faster file searching ([#688](https://github.com/Gitlawb/openclaude/issues/688)) ([12dd375](https://github.com/Gitlawb/openclaude/commit/12dd3755c619cc27af3b151ae8fdb9d425a7b9a2))
+* **api:** classify openai-compatible provider failures ([#708](https://github.com/Gitlawb/openclaude/issues/708)) ([80a00ac](https://github.com/Gitlawb/openclaude/commit/80a00acc2c6dc4657a78de7366f7a9ebc920bfbb))
+* **vscode:** add full chat interface to OpenClaude extension ([#608](https://github.com/Gitlawb/openclaude/issues/608)) ([fbcd928](https://github.com/Gitlawb/openclaude/commit/fbcd928f7f8511da795aea3ad318bddf0ab9a1a7))
+
+
+### Bug Fixes
+
+* focus "Done" option after completing provider manager actions ([#718](https://github.com/Gitlawb/openclaude/issues/718)) ([d6f5130](https://github.com/Gitlawb/openclaude/commit/d6f5130c204d8ffe582212466768706cd7fd6774))
+* **models:** prevent /models crash from non-string saved model values ([#691](https://github.com/Gitlawb/openclaude/issues/691)) ([6b2121d](https://github.com/Gitlawb/openclaude/commit/6b2121da12189fa7ce1f33394d18abd24cf8a01b))
+* prevent crash in commands tab when description is undefined ([#730](https://github.com/Gitlawb/openclaude/issues/730)) ([eed77e6](https://github.com/Gitlawb/openclaude/commit/eed77e6579866a98384dcc948a0ad6406614ede3))
+* strip comments before scanning for missing imports ([#676](https://github.com/Gitlawb/openclaude/issues/676)) ([a00b792](https://github.com/Gitlawb/openclaude/commit/a00b7928de9662ffb7ef6abd8cd040afe6f4f122))
+* **ui:** show correct endpoint URL in intro screen for custom Anthropic endpoints ([#735](https://github.com/Gitlawb/openclaude/issues/735)) ([3424663](https://github.com/Gitlawb/openclaude/commit/34246635fb9a09499047a52e7f96ca9b36c8a85a))
+
+## [0.3.0](https://github.com/Gitlawb/openclaude/compare/v0.2.3...v0.3.0) (2026-04-14)
+
+
+### Features
+
+* activate coordinator mode in open build ([#647](https://github.com/Gitlawb/openclaude/issues/647)) ([99a1714](https://github.com/Gitlawb/openclaude/commit/99a17144ee285b892a0801acb6abcc9af68879af))
+* activate local-only team memory in open build ([#648](https://github.com/Gitlawb/openclaude/issues/648)) ([24d485f](https://github.com/Gitlawb/openclaude/commit/24d485f42f5b1405d2fab13f2f497d5edd3b5300))
+* activate message actions in open build ([#632](https://github.com/Gitlawb/openclaude/issues/632)) ([252808b](https://github.com/Gitlawb/openclaude/commit/252808bbd0a12a6ccf97e2cb09752a0212ea3acd))
+* add allowBypassPermissionsMode setting ([#658](https://github.com/Gitlawb/openclaude/issues/658)) ([31be66d](https://github.com/Gitlawb/openclaude/commit/31be66d7645ea3473334c9ce89ea1a5095b8df6e))
+* add Docker image build and push to GHCR on release ([#656](https://github.com/Gitlawb/openclaude/issues/656)) ([658d076](https://github.com/Gitlawb/openclaude/commit/658d076909e14eb0459bcb98aee9aa0472118265))
+* implement /loop command with fixed and dynamic scheduling ([#621](https://github.com/Gitlawb/openclaude/issues/621)) ([64298a6](https://github.com/Gitlawb/openclaude/commit/64298a663f1391b16aa1f5a49e8a877e1d3742f2))
+* implement Monitor tool for streaming shell output ([#649](https://github.com/Gitlawb/openclaude/issues/649)) ([b818dd5](https://github.com/Gitlawb/openclaude/commit/b818dd5958f4e8428566ce25a1a6be5fd4fe66f8))
+* local feature flag overrides via ~/.claude/feature-flags.json ([#639](https://github.com/Gitlawb/openclaude/issues/639)) ([0e48884](https://github.com/Gitlawb/openclaude/commit/0e48884f56c6c008f047a7926d3b2cb924170625))
+* open useful USER_TYPE-gated features to all users ([#644](https://github.com/Gitlawb/openclaude/issues/644)) ([c1beea9](https://github.com/Gitlawb/openclaude/commit/c1beea98676a413c54152a45a6b9fbe7fb9ed028))
+
+
+### Bug Fixes
+
+* bump axios 1.14.0 → 1.15.0 (Dependabot [#4](https://github.com/Gitlawb/openclaude/issues/4), [#5](https://github.com/Gitlawb/openclaude/issues/5)) ([#670](https://github.com/Gitlawb/openclaude/issues/670)) ([a07e5ef](https://github.com/Gitlawb/openclaude/commit/a07e5ef990a5ed01a72e83fdbd1fcab36f515a08))
+* extend provider guard to protect anthropic profiles from cross-terminal override ([#641](https://github.com/Gitlawb/openclaude/issues/641)) ([03e0b06](https://github.com/Gitlawb/openclaude/commit/03e0b06e0784e4ea46945b3950840b10b6e3ca49))
+* improve fetch diagnostics for bootstrap and session requests ([#646](https://github.com/Gitlawb/openclaude/issues/646)) ([df2b9f2](https://github.com/Gitlawb/openclaude/commit/df2b9f2b7b4c661ee3d9ed5dc58b3064de0599d1))
+* **openai-shim:** preserve tool result images and local token caps ([#659](https://github.com/Gitlawb/openclaude/issues/659)) ([30c866d](https://github.com/Gitlawb/openclaude/commit/30c866d31ad8538496460667d86ed5efbd4a8547))
+* replace broken bun:bundle shim with source pre-processing ([#657](https://github.com/Gitlawb/openclaude/issues/657)) ([adbe391](https://github.com/Gitlawb/openclaude/commit/adbe391e63721918b5d147f4f845111c1a3143db))
+* resolve 12 bugs across API, MCP, agent tools, web search, and context overflow ([#674](https://github.com/Gitlawb/openclaude/issues/674)) ([25ce2ca](https://github.com/Gitlawb/openclaude/commit/25ce2ca7bff8937b0b79ad7f85c6dc1c68432069))
+* route OpenAI Codex shortcuts to correct endpoint ([#566](https://github.com/Gitlawb/openclaude/issues/566)) ([7c8bdcc](https://github.com/Gitlawb/openclaude/commit/7c8bdcc3e2ac1ecb98286c705c85671044be3d6b))
+
+## [0.2.3](https://github.com/Gitlawb/openclaude/compare/v0.2.2...v0.2.3) (2026-04-12)
+
+
+### Bug Fixes
+
+* prevent infinite auto-compact loop for unknown 3P models ([#635](https://github.com/Gitlawb/openclaude/issues/635)) ([#636](https://github.com/Gitlawb/openclaude/issues/636)) ([aeaa658](https://github.com/Gitlawb/openclaude/commit/aeaa658f776fb8df95721e8b8962385f8b00f66a))
+
+## [0.2.2](https://github.com/Gitlawb/openclaude/compare/v0.2.1...v0.2.2) (2026-04-12)
+
+
+### Bug Fixes
+
+* **read/edit:** make compact line prefix unambiguous for tab-indented files ([#613](https://github.com/Gitlawb/openclaude/issues/613)) ([08cc6f3](https://github.com/Gitlawb/openclaude/commit/08cc6f328711cd93ce9fa53351266c29a0b0a341))
+
+## [0.2.1](https://github.com/Gitlawb/openclaude/compare/v0.2.0...v0.2.1) (2026-04-12)
+
+
+### Bug Fixes
+
+* **provider:** add recovery guidance for missing OpenAI API key ([#616](https://github.com/Gitlawb/openclaude/issues/616)) ([9419e8a](https://github.com/Gitlawb/openclaude/commit/9419e8a4a21b3771d9ddb10f7072e0a8c5b5b631))
+
+## [0.2.0](https://github.com/Gitlawb/openclaude/compare/v0.1.8...v0.2.0) (2026-04-12)
+
+
+### Features
+
+* add /cache-probe diagnostic command ([#580](https://github.com/Gitlawb/openclaude/issues/580)) ([9ccaa7a](https://github.com/Gitlawb/openclaude/commit/9ccaa7a6759b6991f4a566b4118c06e68a2398fe)), closes [#515](https://github.com/Gitlawb/openclaude/issues/515)
+* add auto-fix service — auto-lint and test after AI file edits ([#508](https://github.com/Gitlawb/openclaude/issues/508)) ([c385047](https://github.com/Gitlawb/openclaude/commit/c385047abba4366866f4c87bfb5e0b0bd4dcbb9d))
+* Add Gemini support with thought_signature fix  ([#404](https://github.com/Gitlawb/openclaude/issues/404)) ([5012c16](https://github.com/Gitlawb/openclaude/commit/5012c160c9a2dff9418e7ee19dc9a4d29ef2b024))
+* add headless gRPC server for external agent integration ([#278](https://github.com/Gitlawb/openclaude/issues/278)) ([26eef92](https://github.com/Gitlawb/openclaude/commit/26eef92fe72e9c3958d61435b8d3571e12bf2b74))
+* add wiki mvp commands ([#532](https://github.com/Gitlawb/openclaude/issues/532)) ([c328fdf](https://github.com/Gitlawb/openclaude/commit/c328fdf9e2fe59ad101b049301298ce9ff24caca))
+* GitHub provider lifecycle and onboarding hardening ([#351](https://github.com/Gitlawb/openclaude/issues/351)) ([ff7d499](https://github.com/Gitlawb/openclaude/commit/ff7d49990de515825ddbe4099f3a39b944b61370))
+
+
+### Bug Fixes
+
+* add File polyfill for Node < 20 to prevent startup deadlock with proxy ([#442](https://github.com/Gitlawb/openclaude/issues/442)) ([85aa8b0](https://github.com/Gitlawb/openclaude/commit/85aa8b0985c8f3cb8801efa5141114a0ab0f6a83))
+* add GitHub Copilot model context windows and output limits ([#576](https://github.com/Gitlawb/openclaude/issues/576)) ([a7f5982](https://github.com/Gitlawb/openclaude/commit/a7f5982f6438ab0ddc3f0daae31ea68ac7ac206c)), closes [#515](https://github.com/Gitlawb/openclaude/issues/515)
+* add LiteLLM-style aliases for GitHub Copilot context windows ([#606](https://github.com/Gitlawb/openclaude/issues/606)) ([2e0e14d](https://github.com/Gitlawb/openclaude/commit/2e0e14d71313e0e501efaa9e55c6c56f2742fb10))
+* add store:false to Chat Completions and /responses fallback ([#578](https://github.com/Gitlawb/openclaude/issues/578)) ([8aaa4f2](https://github.com/Gitlawb/openclaude/commit/8aaa4f22ac5b942d82aa9cad54af30d56034515a))
+* address code scanning alerts ([#434](https://github.com/Gitlawb/openclaude/issues/434)) ([e365cb4](https://github.com/Gitlawb/openclaude/commit/e365cb4010becabacd7cbccb4c3e59ea23a41e90))
+* avoid sync github credential reads in provider manager ([#428](https://github.com/Gitlawb/openclaude/issues/428)) ([aff2bd8](https://github.com/Gitlawb/openclaude/commit/aff2bd87e4f2821992f74fb95481c505d0ba5d5d))
+* convert dragged file paths to [@mentions](https://github.com/mentions) for attachment ([#382](https://github.com/Gitlawb/openclaude/issues/382)) ([112df59](https://github.com/Gitlawb/openclaude/commit/112df5911791ea71ee9efbb98ea59c5ded1ea161))
+* custom web search — WEB_URL_TEMPLATE not recognized, timeout too short, silent native fallback ([#537](https://github.com/Gitlawb/openclaude/issues/537)) ([32fbd0c](https://github.com/Gitlawb/openclaude/commit/32fbd0c7b4168b32dcb13a5b69342e2727269201))
+* defer startup checks and suppress recommendation dialogs during startup window (issue [#363](https://github.com/Gitlawb/openclaude/issues/363)) ([#504](https://github.com/Gitlawb/openclaude/issues/504)) ([2caf2fd](https://github.com/Gitlawb/openclaude/commit/2caf2fd982af1ec845c50152ad9d28d1a597f82f))
+* display selected model in startup screen instead of hardcoded sonnet 4.6 ([#587](https://github.com/Gitlawb/openclaude/issues/587)) ([b126e38](https://github.com/Gitlawb/openclaude/commit/b126e38b1affddd2de83fcc3ba26f2e44b42a509))
+* handle missing skill parameter in SkillTool ([#485](https://github.com/Gitlawb/openclaude/issues/485)) ([f9ce81b](https://github.com/Gitlawb/openclaude/commit/f9ce81bfb384e909353813fb6f6760cadd508ae7))
+* include MCP tool results in microcompact to reduce token waste ([#348](https://github.com/Gitlawb/openclaude/issues/348)) ([52d33a8](https://github.com/Gitlawb/openclaude/commit/52d33a87a047b943aedaaaf772cd48636c263509))
+* **ink:** restore host prop updates in React 19 reconciler ([#589](https://github.com/Gitlawb/openclaude/issues/589)) ([6e94dd9](https://github.com/Gitlawb/openclaude/commit/6e94dd913688b2d6433a9abe62a245c5f031b776))
+* let saved provider profiles win on restart ([#513](https://github.com/Gitlawb/openclaude/issues/513)) ([cb8f8b7](https://github.com/Gitlawb/openclaude/commit/cb8f8b7ac2e3e74516ee219a3a48156db7c6ed78))
+* normalize malformed Bash tool arguments from OpenAI-compatible providers ([#385](https://github.com/Gitlawb/openclaude/issues/385)) ([b4bd95b](https://github.com/Gitlawb/openclaude/commit/b4bd95b47715c9896240d708c106777507fd26ec))
+* preserve only originally-required properties in strict tool schemas ([#471](https://github.com/Gitlawb/openclaude/issues/471)) ([ccaa193](https://github.com/Gitlawb/openclaude/commit/ccaa193eec5761f0972ffb58eb3189a81a9244b0))
+* preserve unicode in Windows clipboard fallback ([#388](https://github.com/Gitlawb/openclaude/issues/388)) ([c193497](https://github.com/Gitlawb/openclaude/commit/c1934974aaf64db460cc850a044bd13cc744cce7))
+* rebrand prompt identity to openclaude ([#496](https://github.com/Gitlawb/openclaude/issues/496)) ([598651f](https://github.com/Gitlawb/openclaude/commit/598651f42389ce76311ec00e8a9c701c939ead27))
+* replace isDeepStrictEqual with navigation-aware options comparison ([#507](https://github.com/Gitlawb/openclaude/issues/507)) ([537c469](https://github.com/Gitlawb/openclaude/commit/537c469c3a2f7cb0eed05fa2f54dca57b6bc273f)), closes [#472](https://github.com/Gitlawb/openclaude/issues/472)
+* report cache reads in streaming and correct cost calculation ([#577](https://github.com/Gitlawb/openclaude/issues/577)) ([f4ac709](https://github.com/Gitlawb/openclaude/commit/f4ac709fa6eda732bf45204fcab625ba6c5674b9))
+* restore default context window for unknown 3p models ([#494](https://github.com/Gitlawb/openclaude/issues/494)) ([69ea1f1](https://github.com/Gitlawb/openclaude/commit/69ea1f1e4a99e9436215d8cb391a116a64442b94))
+* restore Grep and Glob reliability on OpenAI paths ([#461](https://github.com/Gitlawb/openclaude/issues/461)) ([600c01f](https://github.com/Gitlawb/openclaude/commit/600c01faf761a080a2c7dede872ddbe05a132f23))
+* restore Ollama auto-detect in first-run setup ([#561](https://github.com/Gitlawb/openclaude/issues/561)) ([68c2968](https://github.com/Gitlawb/openclaude/commit/68c296833dcef54ce44cb18b24357230b5204dbc))
+* scrub canonical Anthropic headers from 3P shim requests ([#499](https://github.com/Gitlawb/openclaude/issues/499)) ([07621a6](https://github.com/Gitlawb/openclaude/commit/07621a6f8d0918170281869a47b5dbff90e71594))
+* strip Anthropic params from 3P resume paths ([#479](https://github.com/Gitlawb/openclaude/issues/479)) ([4975cfc](https://github.com/Gitlawb/openclaude/commit/4975cfc2e0ddbe34aa4e8e3f52ee5eba07fbe465))
+* suppress startup dialogs when input is buffered ([#423](https://github.com/Gitlawb/openclaude/issues/423)) ([8ece290](https://github.com/Gitlawb/openclaude/commit/8ece2900872dadd157e798ef501ddf126dac66c4))
+* **tui:** restore prompt rendering on startup ([#498](https://github.com/Gitlawb/openclaude/issues/498)) ([e30ad17](https://github.com/Gitlawb/openclaude/commit/e30ad17ae0056787273be2caafd6cf5340b6ab57))
+* update theme preview on focus change ([#562](https://github.com/Gitlawb/openclaude/issues/562)) ([6924718](https://github.com/Gitlawb/openclaude/commit/692471850fc789ee0797190089272407f9a4d953))
+* **web-search:** close SSRF bypasses in custom provider hostname guard ([#610](https://github.com/Gitlawb/openclaude/issues/610)) ([a02c441](https://github.com/Gitlawb/openclaude/commit/a02c44143b257fbee7f38f1b93873cc0ea68a1f9))
+* WebSearch providers + MCPTool bugs ([#593](https://github.com/Gitlawb/openclaude/issues/593)) ([91e4cfb](https://github.com/Gitlawb/openclaude/commit/91e4cfb15b62c04615834fd3c417fe38b4feb914))

vendor/openclaude/CODE_OF_CONDUCT.md

@@ -0,0 +1,126 @@
+# Contributor Covenant Code of Conduct
+
+## Our Pledge
+
+We as members, contributors, and maintainers pledge to make participation in
+our community a harassment-free experience for everyone, regardless of age,
+body size, visible or invisible disability, ethnicity, sex characteristics,
+gender identity and expression, level of experience, education, socio-economic
+status, nationality, personal appearance, race, religion, or sexual identity
+and orientation.
+
+We pledge to act and interact in ways that contribute to an open, welcoming,
+diverse, inclusive, and healthy community.
+
+## Our Standards
+
+Examples of behavior that contributes to a positive environment for our
+community include:
+
+- Demonstrating empathy and kindness toward other people
+- Being respectful of differing opinions, viewpoints, and experiences
+- Giving and gracefully accepting constructive feedback
+- Accepting responsibility and apologizing to those affected by our mistakes,
+  and learning from the experience
+- Focusing on what is best not just for us as individuals, but for the
+  overall community
+
+Examples of unacceptable behavior include:
+
+- The use of sexualized language or imagery, and sexual attention or
+  advances of any kind
+- Trolling, insulting or derogatory comments, and personal or political attacks
+- Public or private harassment
+- Publishing others' private information, such as a physical or email
+  address, without their explicit permission
+- Other conduct which could reasonably be considered inappropriate in a
+  professional setting
+
+## Enforcement Responsibilities
+
+Community leaders are responsible for clarifying and enforcing our standards of
+acceptable behavior and will take appropriate and fair corrective action in
+response to any behavior that they deem inappropriate, threatening, offensive,
+or harmful.
+
+Community leaders have the right and responsibility to remove, edit, or reject
+comments, commits, code, wiki edits, issues, and other contributions that are
+not aligned to this Code of Conduct, and will communicate reasons for
+moderation decisions when appropriate.
+
+## Scope
+
+This Code of Conduct applies within all community spaces, and also applies when
+an individual is officially representing the community in public spaces.
+Examples of representing our community include using an official email address,
+posting via an official social media account, or acting as an appointed
+representative at an online or offline event.
+
+## Enforcement
+
+Instances of abusive, harassing, or otherwise unacceptable behavior may be
+reported to the project maintainers through the repository maintainers or
+security/community contact paths available in the repository.
+
+All complaints will be reviewed and investigated promptly and fairly.
+
+All community leaders are obligated to respect the privacy and security of the
+reporter of any incident.
+
+## Enforcement Guidelines
+
+Community leaders will follow these Community Impact Guidelines in determining
+the consequences for any action they deem in violation of this Code of Conduct:
+
+### 1. Correction
+
+**Community Impact**: Use of inappropriate language or other behavior deemed
+unprofessional or unwelcome in the community.
+
+**Consequence**: A private, written warning from community leaders, providing
+clarity around the nature of the violation and an explanation of why the
+behavior was inappropriate. A public apology may be requested.
+
+### 2. Warning
+
+**Community Impact**: A violation through a single incident or series
+of actions.
+
+**Consequence**: A warning with consequences for continued behavior. No
+interaction with the people involved, including unsolicited interaction with
+those enforcing the Code of Conduct, for a specified period of time. This
+includes avoiding interactions in community spaces as well as external channels
+like social media. Violating these terms may lead to a temporary or permanent
+ban.
+
+### 3. Temporary Ban
+
+**Community Impact**: A serious violation of community standards, including
+sustained inappropriate behavior.
+
+**Consequence**: A temporary ban from any sort of interaction or public
+communication with the community for a specified period of time. No public or
+private interaction with the people involved, including unsolicited interaction
+with those enforcing the Code of Conduct, is allowed during this period.
+Violating these terms may lead to a permanent ban.
+
+### 4. Permanent Ban
+
+**Community Impact**: Demonstrating a pattern of violation of community
+standards, including sustained inappropriate behavior, harassment of an
+individual, or aggression toward or disparagement of classes of individuals.
+
+**Consequence**: A permanent ban from any sort of public interaction within
+the community.
+
+## Attribution
+
+This Code of Conduct is adapted from the [Contributor Covenant](https://www.contributor-covenant.org/),
+version 2.1, available at
+[https://www.contributor-covenant.org/version/2/1/code_of_conduct.html](https://www.contributor-covenant.org/version/2/1/code_of_conduct.html).
+
+Community Impact Guidelines were inspired by
+[Mozilla's code of conduct enforcement ladder](https://github.com/mozilla/diversity).
+
+For answers to common questions about this code of conduct, see the FAQ at
+[https://www.contributor-covenant.org/faq](https://www.contributor-covenant.org/faq).

vendor/openclaude/CONTRIBUTING.md

@@ -0,0 +1,119 @@
+# Contributing to OpenClaude
+
+Thanks for contributing.
+
+OpenClaude is a fast-moving open-source coding-agent CLI with support for multiple providers, local backends, MCP, and a terminal-first workflow. The best contributions here are focused, well-tested, and easy to review.
+
+## Before You Start
+
+- Search existing [issues](https://github.com/Gitlawb/openclaude/issues) and [discussions](https://github.com/Gitlawb/openclaude/discussions) before opening a new thread.
+- Use issues for confirmed bugs and actionable feature work.
+- Use discussions for setup help, ideas, and general community conversation.
+- For larger changes, open an issue first so the scope is clear before implementation.
+- For security reports, follow [SECURITY.md](SECURITY.md).
+
+## Local Setup
+
+Install dependencies:
+
+```bash
+bun install
+```
+
+Build the CLI:
+
+```bash
+bun run build
+```
+
+Smoke test:
+
+```bash
+bun run smoke
+```
+
+Run the app locally:
+
+```bash
+bun run dev
+```
+
+If you are working on provider setup or saved profiles, useful commands include:
+
+```bash
+bun run profile:init
+bun run dev:profile
+```
+
+## Development Workflow
+
+- Keep PRs focused on one problem or feature.
+- Avoid mixing unrelated cleanup into the same change.
+- Preserve existing repo patterns unless the change is intentionally refactoring them.
+- Add or update tests when the change affects behavior.
+- Update docs when setup, commands, or user-facing behavior changes.
+
+## Validation
+
+At minimum, run the most relevant checks for your change.
+
+Common checks:
+
+```bash
+bun run build
+bun run smoke
+```
+
+Focused tests:
+
+```bash
+bun test ./path/to/test-file.test.ts
+```
+
+When working on provider/runtime setup, this can also help:
+
+```bash
+bun run doctor:runtime
+```
+
+## Pull Requests
+
+Good PRs usually include:
+
+- a short explanation of what changed
+- why it changed
+- the user or developer impact
+- the exact checks you ran
+
+If the PR touches UI, terminal presentation, or the VS Code extension, include screenshots when useful.
+
+If the PR changes provider behavior, mention which provider path was tested.
+
+## Code Style
+
+- Follow the existing code style in the touched files.
+- Prefer small, readable changes over broad rewrites.
+- Do not reformat unrelated files just because they are nearby.
+- Keep comments useful and concise.
+
+## Provider Changes
+
+OpenClaude supports multiple provider paths. If you change provider logic:
+
+- be explicit about which providers are affected
+- avoid breaking third-party providers while fixing first-party behavior
+- test the exact provider/model path you changed when possible
+- call out any limitations or follow-up work in the PR description
+
+## Community
+
+Please be respectful and constructive with other contributors.
+
+Maintainers may ask for:
+
+- narrower scope
+- focused follow-up PRs
+- stronger validation
+- docs updates for behavior changes
+
+That is normal and helps keep the project reviewable as it grows.

vendor/openclaude/Dockerfile

@@ -0,0 +1,46 @@
+# ---- build stage ----
+FROM node:22-slim AS build
+
+# Install Bun
+RUN npm install -g bun@1.3.11
+
+WORKDIR /app
+
+# Copy dependency manifests first for better layer caching
+COPY package.json bun.lock ./
+
+# Install all dependencies (including devDependencies for build)
+RUN bun install --frozen-lockfile
+
+# Copy source code
+COPY src/ src/
+COPY scripts/ scripts/
+COPY bin/ bin/
+COPY tsconfig.json ./
+
+# Build the CLI bundle
+RUN bun run build
+
+# Prune devDependencies
+RUN rm -rf node_modules && bun install --frozen-lockfile --production
+
+# ---- runtime stage ----
+FROM node:22-slim
+
+WORKDIR /app
+
+# Copy only what's needed to run
+COPY --from=build /app/dist/cli.mjs dist/cli.mjs
+COPY --from=build /app/bin/ bin/
+COPY --from=build /app/node_modules/ node_modules/
+COPY --from=build /app/package.json package.json
+COPY README.md ./
+
+# Install git and ripgrep — many CLI tool operations depend on them
+RUN apt-get update && apt-get install -y --no-install-recommends git ripgrep \
+    && rm -rf /var/lib/apt/lists/*
+
+# Run as non-root user
+USER node
+
+ENTRYPOINT ["node", "/app/dist/cli.mjs"]

vendor/openclaude/LICENSE

@@ -0,0 +1,29 @@
+NOTICE
+
+This repository contains code derived from Anthropic's Claude Code CLI.
+
+The original Claude Code source is proprietary software:
+  Copyright (c) Anthropic PBC. All rights reserved.
+  Subject to Anthropic's Commercial Terms of Service.
+
+Modifications and additions by OpenClaude contributors are offered under
+the MIT License where legally permissible:
+
+  MIT License
+  Copyright (c) 2026 OpenClaude contributors (modifications only)
+
+  Permission is hereby granted, free of charge, to any person obtaining
+  a copy of the modifications made by OpenClaude contributors, to deal
+  in those modifications without restriction, including without limitation
+  the rights to use, copy, modify, merge, publish, distribute, sublicense,
+  and/or sell copies, subject to the following conditions:
+
+  The above copyright notice and this permission notice shall be included
+  in all copies or substantial portions of the modifications.
+
+  THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND.
+
+The underlying derived code remains subject to Anthropic's copyright.
+This project does not have Anthropic's authorization to distribute
+their proprietary source. Users and contributors should evaluate their
+own legal position.

vendor/openclaude/PLAYBOOK.md

@@ -0,0 +1,322 @@
+# OpenClaude Local Agent Playbook
+
+This playbook is a practical guide to run OpenClaude with a local model (Ollama), work safely, and get strong day-to-day results.
+
+## 1. What You Have
+
+- A CLI agent loop that can read/write files, run terminal commands, and help with coding workflows.
+- A local provider profile system (`profile:init` and `dev:profile`).
+- Runtime checks (`doctor:runtime`) and reporting (`doctor:report`).
+- A local model profile currently set to `llama3.1:8b`.
+
+## 2. Daily Start (Fast Path)
+
+Run this in your project root:
+
+```powershell
+bun run dev:profile
+```
+
+For quick switches:
+
+```powershell
+# low latency preset
+bun run dev:fast
+
+# better coding quality preset
+bun run dev:code
+```
+
+If everything is healthy, OpenClaude starts directly.
+
+## 3. One-Time Setup (If Needed)
+
+### 3.1 Initialize a local profile
+
+```powershell
+bun run profile:init -- --provider ollama --model llama3.1:8b
+```
+
+Or let OpenClaude recommend the best local model for your goal:
+
+```powershell
+bun run profile:init -- --provider ollama --goal coding
+```
+
+Preview recommendations before saving:
+
+```powershell
+bun run profile:recommend -- --goal coding --benchmark
+```
+
+### 3.2 Confirm profile file
+
+```powershell
+Get-Content .\.openclaude-profile.json
+```
+
+### 3.3 Validate environment
+
+```powershell
+bun run doctor:runtime
+```
+
+## 4. Health and Diagnostics
+
+### 4.1 Human-readable checks
+
+```powershell
+bun run doctor:runtime
+```
+
+### 4.2 JSON diagnostics (automation/logging)
+
+```powershell
+bun run doctor:runtime:json
+```
+
+### 4.3 Persist runtime report
+
+```powershell
+bun run doctor:report
+```
+
+Report output:
+
+- `reports/doctor-runtime.json`
+
+### 4.4 Hardening checks
+
+```powershell
+# practical checks (smoke + runtime doctor)
+bun run hardening:check
+
+# strict checks (includes typecheck)
+bun run hardening:strict
+```
+
+## 5. Provider Modes
+
+## 5.1 Local mode (Ollama)
+
+```powershell
+bun run profile:init -- --provider ollama --model llama3.1:8b
+bun run dev:profile
+```
+
+Expected behavior:
+
+- No API key required.
+- `OPENAI_BASE_URL` should be `http://localhost:11434/v1`.
+
+## 5.2 OpenAI mode
+
+```powershell
+bun run profile:init -- --provider openai --api-key sk-... --model gpt-4o
+bun run dev:profile
+```
+
+Expected behavior:
+
+- Real API key required.
+- Placeholder values fail fast.
+
+## 6. Troubleshooting Matrix
+
+## 6.1 `Script not found "dev"`
+
+Cause:
+
+- You ran command in the wrong folder.
+
+Fix:
+
+```powershell
+cd C:\Users\Lucas Pedry\Documents\openclaude\openclaude
+bun run dev:profile
+```
+
+## 6.2 `ollama: term not recognized`
+
+Cause:
+
+- Ollama not installed or PATH not loaded in this terminal.
+
+Fix:
+
+- Install Ollama from https://ollama.com/download/windows or `winget install Ollama.Ollama`.
+- Open a new terminal and run:
+
+```powershell
+ollama --version
+```
+
+## 6.3 `Provider reachability failed` for localhost
+
+Cause:
+
+- Ollama service not running.
+
+Fix:
+
+```powershell
+ollama serve
+```
+
+Then, in another terminal:
+
+```powershell
+bun run doctor:runtime
+```
+
+## 6.4 `Missing key for non-local provider URL`
+
+Cause:
+
+- `OPENAI_BASE_URL` points to remote endpoint without key.
+
+Fix:
+
+- Re-initialize profile for ollama:
+
+```powershell
+bun run profile:init -- --provider ollama --model llama3.1:8b
+```
+
+Or pick a local Ollama profile automatically by goal:
+
+```powershell
+bun run profile:init -- --provider ollama --goal balanced
+```
+
+## 6.5 Placeholder key (`SUA_CHAVE`) error
+
+Cause:
+
+- Placeholder was used instead of real key.
+
+Fix:
+
+- For OpenAI: use a real key.
+- For Ollama: no key needed; keep localhost base URL.
+
+## 7. Recommended Local Models
+
+- Fast/general: `llama3.1:8b`
+- Better coding quality (if hardware supports): `qwen2.5-coder:14b`
+- Low-resource fallback: smaller instruct model
+
+Switch model quickly:
+
+```powershell
+bun run profile:init -- --provider ollama --model qwen2.5-coder:14b
+bun run dev:profile
+```
+
+Preset shortcuts already configured:
+
+```powershell
+bun run profile:fast   # llama3.2:3b
+bun run profile:code   # qwen2.5-coder:7b
+```
+
+Goal-based local auto-selection:
+
+```powershell
+bun run profile:init -- --provider ollama --goal latency
+bun run profile:init -- --provider ollama --goal balanced
+bun run profile:init -- --provider ollama --goal coding
+```
+
+`profile:auto` is a best-available provider picker, not a local-only command. Use `--provider ollama` when you want to stay on a local model.
+
+## 8. Practical Prompt Playbook (Copy/Paste)
+
+## 8.1 Code understanding
+
+- "Map this repository architecture and explain the execution flow from entrypoint to tool invocation."
+- "Find the top 5 risky modules and explain why."
+
+## 8.2 Refactoring
+
+- "Refactor this module for clarity without behavior change, then run checks and summarize diff impact."
+- "Extract shared logic from duplicated functions and add minimal tests."
+
+## 8.3 Debugging
+
+- "Reproduce the failure, identify root cause, implement fix, and validate with commands."
+- "Trace this error path and list likely failure points with confidence levels."
+
+## 8.4 Reliability
+
+- "Add runtime guardrails and fail-fast messages for invalid provider env vars."
+- "Create a diagnostic command that outputs JSON report for CI artifacts."
+
+## 8.5 Review mode
+
+- "Do a code review of unstaged changes, prioritize bugs/regressions, and suggest concrete patches."
+
+## 9. Safe Working Rules
+
+- Run `doctor:runtime` before debugging provider issues.
+- Prefer `dev:profile` over manual env edits.
+- Keep `.openclaude-profile.json` local (already gitignored).
+- Use `doctor:report` before asking for help so you have a reproducible snapshot.
+
+## 10. Quick Recovery Checklist
+
+When something breaks, run in order:
+
+```powershell
+bun run doctor:runtime
+bun run doctor:report
+bun run smoke
+```
+
+If answers are very slow, check processor mode:
+
+```powershell
+ollama ps
+```
+
+If `PROCESSOR` shows `CPU`, your setup is valid but latency will be higher for large models.
+
+If local model mode is failing:
+
+```powershell
+ollama --version
+ollama serve
+bun run doctor:runtime
+bun run dev:profile
+```
+
+## 11. Command Reference
+
+```powershell
+# profile
+bun run profile:init -- --provider ollama --model llama3.1:8b
+bun run profile:init -- --provider openai --api-key sk-... --model gpt-4o
+
+# launch
+bun run dev:profile
+bun run dev:ollama
+bun run dev:openai
+
+# diagnostics
+bun run doctor:runtime
+bun run doctor:runtime:json
+bun run doctor:report
+
+# quality
+bun run smoke
+bun run hardening:check
+bun run hardening:strict
+```
+
+## 12. Success Criteria
+
+Your setup is healthy when:
+
+- `bun run doctor:runtime` passes provider and reachability checks.
+- `bun run dev:profile` opens the CLI normally.
+- Model shown in the UI matches your selected profile model.

vendor/openclaude/README.md

@@ -0,0 +1,345 @@
+# OpenClaude
+
+OpenClaude is an open-source coding-agent CLI for cloud and local model providers.
+
+Use OpenAI-compatible APIs, Gemini, GitHub Models, Codex OAuth, Codex, Ollama, Atomic Chat, and other supported backends while keeping one terminal-first workflow: prompts, tools, agents, MCP, slash commands, and streaming output.
+
+[![PR Checks](https://github.com/Gitlawb/openclaude/actions/workflows/pr-checks.yml/badge.svg?branch=main)](https://github.com/Gitlawb/openclaude/actions/workflows/pr-checks.yml)
+[![Release](https://img.shields.io/github/v/tag/Gitlawb/openclaude?label=release&color=0ea5e9)](https://github.com/Gitlawb/openclaude/tags)
+[![Discussions](https://img.shields.io/badge/discussions-open-7c3aed)](https://github.com/Gitlawb/openclaude/discussions)
+[![Security Policy](https://img.shields.io/badge/security-policy-0f766e)](SECURITY.md)
+[![License](https://img.shields.io/badge/license-MIT-2563eb)](LICENSE)
+
+OpenClaude is also mirrored to GitLawb:
+[gitlawb.com/node/repos/z6MkqDnb/openclaude](https://gitlawb.com/node/repos/z6MkqDnb/openclaude)
+
+[Quick Start](#quick-start) | [Setup Guides](#setup-guides) | [Providers](#supported-providers) | [Source Build](#source-build-and-local-development) | [VS Code Extension](#vs-code-extension) | [Community](#community)
+
+## Star History
+
+[![Star History Chart](https://api.star-history.com/chart?repos=gitlawb/openclaude&type=date&legend=top-left)](https://www.star-history.com/?repos=gitlawb%2Fopenclaude&type=date&legend=top-left)
+
+## Why OpenClaude
+
+- Use one CLI across cloud APIs and local model backends
+- Save provider profiles inside the app with `/provider`
+- Run with OpenAI-compatible services, Gemini, GitHub Models, Codex OAuth, Codex, Ollama, Atomic Chat, and other supported providers
+- Keep coding-agent workflows in one place: bash, file tools, grep, glob, agents, tasks, MCP, and web tools
+- Use the bundled VS Code extension for launch integration and theme support
+
+## Quick Start
+
+### Install
+
+```bash
+npm install -g @gitlawb/openclaude
+```
+
+If the install later reports `ripgrep not found`, install ripgrep system-wide and confirm `rg --version` works in the same terminal before starting OpenClaude.
+
+### Start
+
+```bash
+openclaude
+```
+
+Inside OpenClaude:
+
+- run `/provider` for guided provider setup and saved profiles
+- run `/onboard-github` for GitHub Models onboarding
+
+### Fastest OpenAI setup
+
+macOS / Linux:
+
+```bash
+export CLAUDE_CODE_USE_OPENAI=1
+export OPENAI_API_KEY=sk-your-key-here
+export OPENAI_MODEL=gpt-4o
+
+openclaude
+```
+
+Windows PowerShell:
+
+```powershell
+$env:CLAUDE_CODE_USE_OPENAI="1"
+$env:OPENAI_API_KEY="sk-your-key-here"
+$env:OPENAI_MODEL="gpt-4o"
+
+openclaude
+```
+
+### Fastest local Ollama setup
+
+macOS / Linux:
+
+```bash
+export CLAUDE_CODE_USE_OPENAI=1
+export OPENAI_BASE_URL=http://localhost:11434/v1
+export OPENAI_MODEL=qwen2.5-coder:7b
+
+openclaude
+```
+
+Windows PowerShell:
+
+```powershell
+$env:CLAUDE_CODE_USE_OPENAI="1"
+$env:OPENAI_BASE_URL="http://localhost:11434/v1"
+$env:OPENAI_MODEL="qwen2.5-coder:7b"
+
+openclaude
+```
+
+### Using Ollama's launch command
+
+If you have [Ollama](https://ollama.com) installed, you can skip the env var setup entirely:
+
+```bash
+ollama launch openclaude --model qwen2.5-coder:7b
+```
+
+This automatically sets `ANTHROPIC_BASE_URL`, model routing, and auth so all API traffic goes through your local Ollama instance. Works with any model you have pulled — local or cloud.
+
+## Setup Guides
+
+Beginner-friendly guides:
+
+- [Non-Technical Setup](docs/non-technical-setup.md)
+- [Windows Quick Start](docs/quick-start-windows.md)
+- [macOS / Linux Quick Start](docs/quick-start-mac-linux.md)
+
+Advanced and source-build guides:
+
+- [Advanced Setup](docs/advanced-setup.md)
+- [Android Install](ANDROID_INSTALL.md)
+
+## Supported Providers
+
+| Provider | Setup Path | Notes |
+| --- | --- | --- |
+| OpenAI-compatible | `/provider` or env vars | Works with OpenAI, OpenRouter, DeepSeek, Groq, Mistral, LM Studio, and other compatible `/v1` servers |
+| Gemini | `/provider` or env vars | Supports API key, access token, or local ADC workflow on current `main` |
+| GitHub Models | `/onboard-github` | Interactive onboarding with saved credentials |
+| Codex OAuth | `/provider` | Opens ChatGPT sign-in in your browser and stores Codex credentials securely |
+| Codex | `/provider` | Uses existing Codex CLI auth, OpenClaude secure storage, or env credentials |
+| Ollama | `/provider`, env vars, or `ollama launch` | Local inference with no API key |
+| Atomic Chat | `/provider`, env vars, or `bun run dev:atomic-chat` | Local Model Provider; auto-detects loaded models |
+| Bedrock / Vertex / Foundry | env vars | Additional provider integrations for supported environments |
+
+## What Works
+
+- **Tool-driven coding workflows**: Bash, file read/write/edit, grep, glob, agents, tasks, MCP, and slash commands
+- **Streaming responses**: Real-time token output and tool progress
+- **Tool calling**: Multi-step tool loops with model calls, tool execution, and follow-up responses
+- **Images**: URL and base64 image inputs for providers that support vision
+- **Provider profiles**: Guided setup plus saved `.openclaude-profile.json` support
+- **Local and remote model backends**: Cloud APIs, local servers, and Apple Silicon local inference
+
+## Provider Notes
+
+OpenClaude supports multiple providers, but behavior is not identical across all of them.
+
+- Anthropic-specific features may not exist on other providers
+- Tool quality depends heavily on the selected model
+- Smaller local models can struggle with long multi-step tool flows
+- Some providers impose lower output caps than the CLI defaults, and OpenClaude adapts where possible
+
+For best results, use models with strong tool/function calling support.
+
+## Agent Routing
+
+OpenClaude can route different agents to different models through settings-based routing. This is useful for cost optimization or splitting work by model strength.
+
+Add to `~/.claude/settings.json`:
+
+```json
+{
+  "agentModels": {
+    "deepseek-chat": {
+      "base_url": "https://api.deepseek.com/v1",
+      "api_key": "sk-your-key"
+    },
+    "gpt-4o": {
+      "base_url": "https://api.openai.com/v1",
+      "api_key": "sk-your-key"
+    }
+  },
+  "agentRouting": {
+    "Explore": "deepseek-chat",
+    "Plan": "gpt-4o",
+    "general-purpose": "gpt-4o",
+    "frontend-dev": "deepseek-chat",
+    "default": "gpt-4o"
+  }
+}
+```
+
+When no routing match is found, the global provider remains the fallback.
+
+> **Note:** `api_key` values in `settings.json` are stored in plaintext. Keep this file private and do not commit it to version control.
+
+## Web Search and Fetch
+
+By default, `WebSearch` works on non-Anthropic models using DuckDuckGo. This gives GPT-4o, DeepSeek, Gemini, Ollama, and other OpenAI-compatible providers a free web search path out of the box.
+
+> **Note:** DuckDuckGo fallback works by scraping search results and may be rate-limited, blocked, or subject to DuckDuckGo's Terms of Service. If you want a more reliable supported option, configure Firecrawl.
+
+For Anthropic-native backends and Codex responses, OpenClaude keeps the native provider web search behavior.
+
+`WebFetch` works, but its basic HTTP plus HTML-to-markdown path can still fail on JavaScript-rendered sites or sites that block plain HTTP requests.
+
+Set a [Firecrawl](https://firecrawl.dev) API key if you want Firecrawl-powered search/fetch behavior:
+
+```bash
+export FIRECRAWL_API_KEY=your-key-here
+```
+
+With Firecrawl enabled:
+
+- `WebSearch` can use Firecrawl's search API while DuckDuckGo remains the default free path for non-Claude models
+- `WebFetch` uses Firecrawl's scrape endpoint instead of raw HTTP, handling JS-rendered pages correctly
+
+Free tier at [firecrawl.dev](https://firecrawl.dev) includes 500 credits. The key is optional.
+
+---
+
+## Headless gRPC Server
+
+OpenClaude can be run as a headless gRPC service, allowing you to integrate its agentic capabilities (tools, bash, file editing) into other applications, CI/CD pipelines, or custom user interfaces. The server uses bidirectional streaming to send real-time text chunks, tool calls, and request permissions for sensitive commands.
+
+### 1. Start the gRPC Server
+
+Start the core engine as a gRPC service on `localhost:50051`:
+
+```bash
+npm run dev:grpc
+```
+
+#### Configuration
+
+| Variable | Default | Description |
+|-----------|-------------|------------------------------------------------|
+| `GRPC_PORT` | `50051` | Port the gRPC server listens on |
+| `GRPC_HOST` | `localhost` | Bind address. Use `0.0.0.0` to expose on all interfaces (not recommended without authentication) |
+
+### 2. Run the Test CLI Client
+
+We provide a lightweight CLI client that communicates exclusively over gRPC. It acts just like the main interactive CLI, rendering colors, streaming tokens, and prompting you for tool permissions (y/n) via the gRPC `action_required` event.
+
+In a separate terminal, run:
+
+```bash
+npm run dev:grpc:cli
+```
+
+*Note: The gRPC definitions are located in `src/proto/openclaude.proto`. You can use this file to generate clients in Python, Go, Rust, or any other language.*
+
+---
+
+## Source Build And Local Development
+
+```bash
+bun install
+bun run build
+node dist/cli.mjs
+```
+
+Helpful commands:
+
+- `bun run dev`
+- `bun test`
+- `bun run test:coverage`
+- `bun run security:pr-scan -- --base origin/main`
+- `bun run smoke`
+- `bun run doctor:runtime`
+- `bun run verify:privacy`
+- focused `bun test ...` runs for the areas you touch
+
+## Testing And Coverage
+
+OpenClaude uses Bun's built-in test runner for unit tests.
+
+Run the full unit suite:
+
+```bash
+bun test
+```
+
+Generate unit test coverage:
+
+```bash
+bun run test:coverage
+```
+
+Open the visual coverage report:
+
+```bash
+open coverage/index.html
+```
+
+If you already have `coverage/lcov.info` and only want to rebuild the UI:
+
+```bash
+bun run test:coverage:ui
+```
+
+Use focused test runs when you only touch one area:
+
+- `bun run test:provider`
+- `bun run test:provider-recommendation`
+- `bun test path/to/file.test.ts`
+
+Recommended contributor validation before opening a PR:
+
+- `bun run build`
+- `bun run smoke`
+- `bun run test:coverage` for broader unit coverage when your change affects shared runtime or provider logic
+- focused `bun test ...` runs for the files and flows you changed
+
+Coverage output is written to `coverage/lcov.info`, and OpenClaude also generates a git-activity-style heatmap at `coverage/index.html`.
+## Repository Structure
+
+- `src/` - core CLI/runtime
+- `scripts/` - build, verification, and maintenance scripts
+- `docs/` - setup, contributor, and project documentation
+- `python/` - standalone Python helpers and their tests
+- `vscode-extension/openclaude-vscode/` - VS Code extension
+- `.github/` - repo automation, templates, and CI configuration
+- `bin/` - CLI launcher entrypoints
+
+## VS Code Extension
+
+The repo includes a VS Code extension in [`vscode-extension/openclaude-vscode`](vscode-extension/openclaude-vscode) for OpenClaude launch integration, provider-aware control-center UI, and theme support.
+
+## Security
+
+If you believe you found a security issue, see [SECURITY.md](SECURITY.md).
+
+## Community
+
+- Use [GitHub Discussions](https://github.com/Gitlawb/openclaude/discussions) for Q&A, ideas, and community conversation
+- Use [GitHub Issues](https://github.com/Gitlawb/openclaude/issues) for confirmed bugs and actionable feature work
+
+## Contributing
+
+Contributions are welcome.
+
+For larger changes, open an issue first so the scope is clear before implementation. Helpful validation commands include:
+
+- `bun run build`
+- `bun run test:coverage`
+- `bun run smoke`
+- focused `bun test ...` runs for files and flows you changed
+
+
+## Disclaimer
+
+OpenClaude is an independent community project and is not affiliated with, endorsed by, or sponsored by Anthropic.
+
+OpenClaude originated from the Claude Code codebase and has since been substantially modified to support multiple providers and open use. "Claude" and "Claude Code" are trademarks of Anthropic PBC. See [LICENSE](LICENSE) for details.
+
+## License
+
+See [LICENSE](LICENSE).

vendor/openclaude/SECURITY.md

@@ -0,0 +1,69 @@
+# Security Policy
+
+## Supported Versions
+
+Open Claude is currently maintained on the latest `main` branch and the latest
+npm release only.
+
+| Version | Supported |
+| ------- | --------- |
+| Latest release | :white_check_mark: |
+| Older releases | :x: |
+| Unreleased forks / modified builds | :x: |
+
+Security fixes are generally released in the next patch version and may also be
+landed directly on `main` before a package release is published.
+
+## Reporting a Vulnerability
+
+If you believe you have found a security vulnerability in Open Claude, please
+report it privately.
+
+Preferred reporting channel:
+
+- GitHub Security Advisories / private vulnerability reporting for this
+  repository
+
+Please include:
+
+- a clear description of the issue
+- affected version, commit, or environment
+- reproduction steps or a proof of concept
+- impact assessment
+- any suggested remediation, if available
+
+Please do **not** open a public issue for an unpatched vulnerability.
+
+## Response Process
+
+Our general goals are:
+
+- initial triage acknowledgment within 7 days
+- follow-up after validation when we can reproduce the issue
+- coordinated disclosure after a fix is available
+
+Severity, exploitability, and maintenance bandwidth may affect timelines.
+
+## Disclosure and CVEs
+
+Valid reports may be fixed privately first and disclosed after a patch is
+available.
+
+If a report is accepted and the issue is significant enough to warrant formal
+tracking, we may publish a GitHub Security Advisory and request or assign a CVE
+through the appropriate channel. CVE issuance is not guaranteed for every
+report.
+
+## Scope
+
+This policy applies to:
+
+- the Open Claude source code in this repository
+- official release artifacts published from this repository
+- the `@gitlawb/openclaude` npm package
+
+This policy does not cover:
+
+- third-party model providers, endpoints, or hosted services
+- local misconfiguration on the reporter's machine
+- vulnerabilities in unofficial forks, mirrors, or downstream repackages

vendor/openclaude/bin/import-specifier.mjs

@@ -0,0 +1,13 @@
+import { join, win32 } from 'path'
+import { pathToFileURL } from 'url'
+
+export function getDistImportSpecifier(baseDir) {
+  if (/^[A-Za-z]:\\/.test(baseDir)) {
+    const distPath = win32.join(baseDir, '..', 'dist', 'cli.mjs')
+    return `file:///${distPath.replace(/\\/g, '/')}`
+  }
+
+  const joinImpl = join
+  const distPath = joinImpl(baseDir, '..', 'dist', 'cli.mjs')
+  return pathToFileURL(distPath).href
+}

vendor/openclaude/bin/import-specifier.test.mjs

@@ -0,0 +1,13 @@
+import assert from 'node:assert/strict'
+import test from 'node:test'
+
+import { getDistImportSpecifier } from './import-specifier.mjs'
+
+test('builds a file URL import specifier for dist/cli.mjs', () => {
+  const specifier = getDistImportSpecifier('C:\\repo\\bin')
+
+  assert.equal(
+    specifier,
+    'file:///C:/repo/dist/cli.mjs',
+  )
+})

vendor/openclaude/bin/openclaude

@@ -0,0 +1,32 @@
+#!/usr/bin/env node
+
+/**
+ * OpenClaude — Claude Code with any LLM
+ *
+ * If dist/cli.mjs exists (built), run that.
+ * Otherwise, tell the user to build first or use `bun run dev`.
+ */
+
+import { existsSync } from 'fs'
+import { join, dirname } from 'path'
+import { fileURLToPath, pathToFileURL } from 'url'
+
+const __dirname = dirname(fileURLToPath(import.meta.url))
+const distPath = join(__dirname, '..', 'dist', 'cli.mjs')
+
+if (existsSync(distPath)) {
+  await import(pathToFileURL(distPath).href)
+} else {
+  console.error(`
+  openclaude: dist/cli.mjs not found.
+
+  Build first:
+    bun run build
+
+  Or run directly with Bun:
+    bun run dev
+
+  See README.md for setup instructions.
+`)
+  process.exit(1)
+}

vendor/openclaude/docs/advanced-setup.md

@@ -0,0 +1,273 @@
+# OpenClaude Advanced Setup
+
+This guide is for users who want source builds, Bun workflows, provider profiles, diagnostics, or more control over runtime behavior.
+
+## Install Options
+
+### Option A: npm
+
+```bash
+npm install -g @gitlawb/openclaude
+```
+
+### Option B: From source with Bun
+
+Use Bun `1.3.11` or newer for source builds on Windows. Older Bun versions can fail during `bun run build`.
+
+```bash
+git clone https://node.gitlawb.com/z6MkqDnb7Siv3Cwj7pGJq4T5EsUisECqR8KpnDLwcaZq5TPr/openclaude.git
+cd openclaude
+
+bun install
+bun run build
+npm link
+```
+
+### Option C: Run directly with Bun
+
+```bash
+git clone https://node.gitlawb.com/z6MkqDnb7Siv3Cwj7pGJq4T5EsUisECqR8KpnDLwcaZq5TPr/openclaude.git
+cd openclaude
+
+bun install
+bun run dev
+```
+
+## Provider Examples
+
+### OpenAI
+
+```bash
+export CLAUDE_CODE_USE_OPENAI=1
+export OPENAI_API_KEY=sk-...
+export OPENAI_MODEL=gpt-4o
+```
+
+### Codex via ChatGPT auth
+
+`codexplan` maps to GPT-5.4 on the Codex backend with high reasoning.
+`codexspark` maps to GPT-5.3 Codex Spark for faster loops.
+
+If you use the in-app provider wizard, choose `Codex OAuth` to open ChatGPT sign-in in your browser and let OpenClaude store Codex credentials securely.
+
+If you already use the Codex CLI, OpenClaude reads `~/.codex/auth.json` automatically. You can also point it elsewhere with `CODEX_AUTH_JSON_PATH` or override the token directly with `CODEX_API_KEY`.
+
+```bash
+export CLAUDE_CODE_USE_OPENAI=1
+export OPENAI_MODEL=codexplan
+
+# optional if you do not already have ~/.codex/auth.json
+export CODEX_API_KEY=...
+
+openclaude
+```
+
+### DeepSeek
+
+```bash
+export CLAUDE_CODE_USE_OPENAI=1
+export OPENAI_API_KEY=sk-...
+export OPENAI_BASE_URL=https://api.deepseek.com/v1
+export OPENAI_MODEL=deepseek-chat
+```
+
+### Google Gemini via OpenRouter
+
+```bash
+export CLAUDE_CODE_USE_OPENAI=1
+export OPENAI_API_KEY=sk-or-...
+export OPENAI_BASE_URL=https://openrouter.ai/api/v1
+export OPENAI_MODEL=google/gemini-2.0-flash-001
+```
+
+OpenRouter model availability changes over time. If a model stops working, try another current OpenRouter model before assuming the integration is broken.
+
+### Ollama
+
+Using `ollama launch` (recommended if you have Ollama installed):
+
+```bash
+ollama launch openclaude --model llama3.3:70b
+```
+
+This handles all environment setup automatically — no env vars needed. Works with any local or cloud model available in your Ollama instance.
+
+Using environment variables manually:
+
+```bash
+ollama pull llama3.3:70b
+
+export CLAUDE_CODE_USE_OPENAI=1
+export OPENAI_BASE_URL=http://localhost:11434/v1
+export OPENAI_MODEL=llama3.3:70b
+```
+
+### Atomic Chat (local, Apple Silicon)
+
+```bash
+export CLAUDE_CODE_USE_OPENAI=1
+export OPENAI_BASE_URL=http://127.0.0.1:1337/v1
+export OPENAI_MODEL=your-model-name
+```
+
+No API key is needed for Atomic Chat local models.
+
+Or use the profile launcher:
+
+```bash
+bun run dev:atomic-chat
+```
+
+Download Atomic Chat from [atomic.chat](https://atomic.chat/). The app must be running with a model loaded before launching.
+
+### LM Studio
+
+```bash
+export CLAUDE_CODE_USE_OPENAI=1
+export OPENAI_BASE_URL=http://localhost:1234/v1
+export OPENAI_MODEL=your-model-name
+```
+
+### Together AI
+
+```bash
+export CLAUDE_CODE_USE_OPENAI=1
+export OPENAI_API_KEY=...
+export OPENAI_BASE_URL=https://api.together.xyz/v1
+export OPENAI_MODEL=meta-llama/Llama-3.3-70B-Instruct-Turbo
+```
+
+### Groq
+
+```bash
+export CLAUDE_CODE_USE_OPENAI=1
+export OPENAI_API_KEY=gsk_...
+export OPENAI_BASE_URL=https://api.groq.com/openai/v1
+export OPENAI_MODEL=llama-3.3-70b-versatile
+```
+
+### Mistral
+
+```bash
+export CLAUDE_CODE_USE_MISTRAL=1
+export MISTRAL_API_KEY=...
+export MISTRAL_MODEL=mistral-large-latest
+```
+
+### Azure OpenAI
+
+```bash
+export CLAUDE_CODE_USE_OPENAI=1
+export OPENAI_API_KEY=your-azure-key
+export OPENAI_BASE_URL=https://your-resource.openai.azure.com/openai/deployments/your-deployment/v1
+export OPENAI_MODEL=gpt-4o
+```
+
+## Environment Variables
+
+| Variable | Required | Description |
+|----------|----------|-------------|
+| `CLAUDE_CODE_USE_OPENAI` | Yes | Set to `1` to enable the OpenAI provider |
+| `OPENAI_API_KEY` | Yes* | Your API key (`*` not needed for local models like Ollama or Atomic Chat) |
+| `OPENAI_MODEL` | Yes | Model name such as `gpt-4o`, `deepseek-chat`, or `llama3.3:70b` |
+| `OPENAI_BASE_URL` | No | API endpoint, defaulting to `https://api.openai.com/v1` |
+| `CODEX_API_KEY` | Codex only | Codex or ChatGPT access token override |
+| `CODEX_AUTH_JSON_PATH` | Codex only | Path to a Codex CLI `auth.json` file |
+| `CODEX_HOME` | Codex only | Alternative Codex home directory |
+| `OPENCLAUDE_DISABLE_CO_AUTHORED_BY` | No | Suppress the default `Co-Authored-By` trailer in generated git commits |
+
+You can also use `ANTHROPIC_MODEL` to override the model name. `OPENAI_MODEL` takes priority.
+
+## Runtime Hardening
+
+Use these commands to validate your setup and catch mistakes early:
+
+```bash
+# quick startup sanity check
+bun run smoke
+
+# validate provider env + reachability
+bun run doctor:runtime
+
+# print machine-readable runtime diagnostics
+bun run doctor:runtime:json
+
+# persist a diagnostics report to reports/doctor-runtime.json
+bun run doctor:report
+
+# full local hardening check (smoke + runtime doctor)
+bun run hardening:check
+
+# strict hardening (includes project-wide typecheck)
+bun run hardening:strict
+```
+
+Notes:
+
+- `doctor:runtime` fails fast if `CLAUDE_CODE_USE_OPENAI=1` with a placeholder key or a missing key for non-local providers.
+- Local providers such as `http://localhost:11434/v1`, `http://10.0.0.1:11434/v1`, and `http://127.0.0.1:1337/v1` can run without `OPENAI_API_KEY`.
+- Codex profiles validate `CODEX_API_KEY` or the Codex CLI auth file and probe `POST /responses` instead of `GET /models`.
+
+## Provider Launch Profiles
+
+Use profile launchers to avoid repeated environment setup:
+
+```bash
+# one-time profile bootstrap (prefer viable local Ollama, otherwise OpenAI)
+bun run profile:init
+
+# preview the best provider/model for your goal
+bun run profile:recommend -- --goal coding --benchmark
+
+# auto-apply the best available local/openai provider/model for your goal
+bun run profile:auto -- --goal latency
+
+# codex bootstrap (defaults to codexplan and ~/.codex/auth.json)
+bun run profile:codex
+
+# openai bootstrap with explicit key
+bun run profile:init -- --provider openai --api-key sk-...
+
+# ollama bootstrap with custom model
+bun run profile:init -- --provider ollama --model llama3.1:8b
+
+# ollama bootstrap with intelligent model auto-selection
+bun run profile:init -- --provider ollama --goal coding
+
+# atomic-chat bootstrap (auto-detects running model)
+bun run profile:init -- --provider atomic-chat
+
+# codex bootstrap with a fast model alias
+bun run profile:init -- --provider codex --model codexspark
+
+# launch using persisted profile (.openclaude-profile.json)
+bun run dev:profile
+
+# codex profile (uses CODEX_API_KEY or ~/.codex/auth.json)
+bun run dev:codex
+
+# OpenAI profile (requires OPENAI_API_KEY in your shell)
+bun run dev:openai
+
+# Ollama profile (defaults: localhost:11434, llama3.1:8b)
+bun run dev:ollama
+
+# Atomic Chat profile (Apple Silicon local LLMs at 127.0.0.1:1337)
+bun run dev:atomic-chat
+```
+
+`profile:recommend` ranks installed Ollama models for `latency`, `balanced`, or `coding`, and `profile:auto` can persist the recommendation directly.
+
+If no profile exists yet, `dev:profile` uses the same goal-aware defaults when picking the initial model.
+
+Use `--provider ollama` when you want a local-only path. Auto mode falls back to OpenAI when no viable local chat model is installed.
+
+Use `--provider atomic-chat` when you want Atomic Chat as the local Apple Silicon provider.
+
+Use `profile:codex` or `--provider codex` when you want the ChatGPT Codex backend.
+
+`dev:openai`, `dev:ollama`, `dev:atomic-chat`, and `dev:codex` run `doctor:runtime` first and only launch the app if checks pass.
+
+For `dev:ollama`, make sure Ollama is running locally before launch.
+
+For `dev:atomic-chat`, make sure Atomic Chat is running with a model loaded before launch.

vendor/openclaude/docs/hook-chains.md

@@ -0,0 +1,333 @@
+# Hook Chains (Self-Healing Agent Mesh MVP)
+
+Hook Chains provide an event-driven recovery layer for important workflow failures.
+When a matching hook event occurs, OpenClaude evaluates declarative rules and can dispatch remediation actions such as:
+
+- `spawn_fallback_agent`
+- `notify_team`
+- `warm_remote_capacity`
+
+## Disabled-By-Default Rollout
+
+> **Rollout recommendation:** keep Hook Chains disabled until you validate rules in your environment.
+>
+> - Set top-level config to `"enabled": false` initially.
+> - Enable per environment when ready.
+> - Dispatch is gated by `feature('HOOK_CHAINS')`.
+> - Env gate defaults to off unless `CLAUDE_CODE_ENABLE_HOOK_CHAINS=1` is set.
+
+This keeps existing workflows unchanged while you tune guard windows and action behavior.
+
+## Feature Overview
+
+Hook Chains are loaded from a deterministic config file and evaluated on dispatched hook events.
+
+MVP runtime trigger wiring:
+
+- `PostToolUseFailure` hooks dispatch Hook Chains with outcome `failed`.
+- `TaskCompleted` hooks dispatch Hook Chains with outcome:
+  - `success` when completion hooks did not block.
+  - `failed` when completion hooks returned blocking errors or prevented continuation.
+
+Default config path:
+
+- `.openclaude/hook-chains.json`
+
+Override path:
+
+- `CLAUDE_CODE_HOOK_CHAINS_CONFIG_PATH=/abs/or/relative/path/to/hook-chains.json`
+
+Global gate:
+
+- `feature('HOOK_CHAINS')` must be enabled in the build
+- `CLAUDE_CODE_ENABLE_HOOK_CHAINS=0|1` (defaults to disabled when unset)
+
+## Safety Guarantees
+
+The runtime is intentionally conservative:
+
+- **Depth guard:** chain dispatch is blocked when `chainDepth >= maxChainDepth`.
+- **Rule cooldown:** each rule can only re-fire after cooldown expires.
+- **Dedup window:** identical event/action combinations are suppressed for a window.
+- **Abort-safe behavior:** if the current signal is aborted, actions skip safely.
+- **Policy-aware remote warm:** `warm_remote_capacity` skips when remote sessions are policy denied.
+- **Bridge inactive no-op:** `warm_remote_capacity` safely skips when no active bridge handle exists.
+- **Missing team context safety:** `notify_team` skips with structured reason if no team context/team file is available.
+- **Fallback launcher safety:** `spawn_fallback_agent` fails with a structured reason when launch permissions/context are unavailable.
+
+## Configuration Schema Reference
+
+Top-level object:
+
+```json
+{
+  "version": 1,
+  "enabled": true,
+  "maxChainDepth": 2,
+  "defaultCooldownMs": 30000,
+  "defaultDedupWindowMs": 30000,
+  "rules": []
+}
+```
+
+### Top-Level Fields
+
+| Field | Type | Required | Notes |
+|---|---|---:|---|
+| `version` | `1` | No | Defaults to `1`. |
+| `enabled` | `boolean` | No | Global feature switch for this config file. |
+| `maxChainDepth` | `integer` | No | Global depth guard (default `2`, max `10`). |
+| `defaultCooldownMs` | `integer` | No | Default rule cooldown in ms (default `30000`). |
+| `defaultDedupWindowMs` | `integer` | No | Default action dedup window in ms (default `30000`). |
+| `rules` | `HookChainRule[]` | No | Defaults to `[]`. May be omitted or empty; when no rules are present, dispatch is a no-op and returns `enabled: false`. |
+
+> **Note:** An empty ruleset is valid and can be used to keep Hook Chains configured but effectively disabled until rules are added.
+### Rule Object (`HookChainRule`)
+
+```json
+{
+  "id": "task-failure-recovery",
+  "enabled": true,
+  "trigger": {
+    "event": "TaskCompleted",
+    "outcome": "failed"
+  },
+  "condition": {
+    "toolNames": ["Edit"],
+    "taskStatuses": ["failed"],
+    "errorIncludes": ["timeout", "permission denied"],
+    "eventFieldEquals": {
+      "meta.source": "scheduler"
+    }
+  },
+  "cooldownMs": 60000,
+  "dedupWindowMs": 30000,
+  "maxDepth": 2,
+  "actions": []
+}
+```
+
+| Field | Type | Required | Notes |
+|---|---|---:|---|
+| `id` | `string` | Yes | Stable identifier used in telemetry/guards. |
+| `enabled` | `boolean` | No | Per-rule switch. |
+| `trigger.event` | `HookEvent` | Yes | Event name to match. |
+| `trigger.outcome` | `"success"|"failed"|"timeout"|"unknown"` | No | Single outcome matcher. |
+| `trigger.outcomes` | `Outcome[]` | No | Multi-outcome matcher. Use either `outcome` or `outcomes`. |
+| `condition` | `object` | No | Optional extra matching constraints. |
+| `cooldownMs` | `integer` | No | Overrides global cooldown for this rule. |
+| `dedupWindowMs` | `integer` | No | Overrides global dedup for this rule. |
+| `maxDepth` | `integer` | No | Per-rule depth cap. |
+| `actions` | `HookChainAction[]` | Yes | One or more actions to execute in order. |
+
+### Condition Fields
+
+| Field | Type | Notes |
+|---|---|---|
+| `toolNames` | `string[]` | Matches `tool_name` / `toolName` in event payload. |
+| `taskStatuses` | `string[]` | Matches `task_status` / `taskStatus` / `status`. |
+| `errorIncludes` | `string[]` | Case-insensitive substring match against `error` / `reason` / `message`. |
+| `eventFieldEquals` | `Record<string, string\|number\|boolean>` | Dot-path equality against payload (example: `"meta.source": "scheduler"`). |
+
+### Actions
+
+#### `spawn_fallback_agent`
+
+```json
+{
+  "type": "spawn_fallback_agent",
+  "id": "fallback-1",
+  "enabled": true,
+  "dedupWindowMs": 30000,
+  "description": "Fallback recovery for failed task",
+  "promptTemplate": "Recover task ${TASK_SUBJECT}. Event=${EVENT_NAME}, outcome=${OUTCOME}, error=${ERROR}. Payload=${PAYLOAD_JSON}",
+  "agentType": "general-purpose",
+  "model": "sonnet"
+}
+```
+
+#### `notify_team`
+
+```json
+{
+  "type": "notify_team",
+  "id": "notify-ops",
+  "enabled": true,
+  "dedupWindowMs": 30000,
+  "teamName": "mesh-team",
+  "recipients": ["*"],
+  "summary": "Hook chain ${RULE_ID} fired",
+  "messageTemplate": "Event=${EVENT_NAME} outcome=${OUTCOME}\nTask=${TASK_ID}\nError=${ERROR}\nPayload=${PAYLOAD_JSON}"
+}
+```
+
+#### `warm_remote_capacity`
+
+```json
+{
+  "type": "warm_remote_capacity",
+  "id": "warm-bridge",
+  "enabled": true,
+  "dedupWindowMs": 60000,
+  "createDefaultEnvironmentIfMissing": false
+}
+```
+
+## Complete Example Configs
+
+### 1) Retry via Fallback Agent
+
+```json
+{
+  "version": 1,
+  "enabled": true,
+  "maxChainDepth": 2,
+  "defaultCooldownMs": 30000,
+  "defaultDedupWindowMs": 30000,
+  "rules": [
+    {
+      "id": "retry-task-via-fallback",
+      "trigger": {
+        "event": "TaskCompleted",
+        "outcome": "failed"
+      },
+      "cooldownMs": 60000,
+      "actions": [
+        {
+          "type": "spawn_fallback_agent",
+          "id": "spawn-retry-agent",
+          "description": "Retry failed task with fallback agent",
+          "promptTemplate": "A task failed. Recover it safely.\nTask=${TASK_SUBJECT}\nDescription=${TASK_DESCRIPTION}\nError=${ERROR}\nPayload=${PAYLOAD_JSON}",
+          "agentType": "general-purpose",
+          "model": "sonnet"
+        }
+      ]
+    }
+  ]
+}
+```
+
+### 2) Notify Only
+
+```json
+{
+  "version": 1,
+  "enabled": true,
+  "maxChainDepth": 2,
+  "defaultCooldownMs": 30000,
+  "defaultDedupWindowMs": 30000,
+  "rules": [
+    {
+      "id": "notify-on-tool-failure",
+      "trigger": {
+        "event": "PostToolUseFailure",
+        "outcome": "failed"
+      },
+      "condition": {
+        "toolNames": ["Edit", "Write", "Bash"]
+      },
+      "actions": [
+        {
+          "type": "notify_team",
+          "id": "notify-team-failure",
+          "recipients": ["*"],
+          "summary": "Tool failure detected",
+          "messageTemplate": "Tool failure detected.\nEvent=${EVENT_NAME} outcome=${OUTCOME}\nError=${ERROR}\nPayload=${PAYLOAD_JSON}"
+        }
+      ]
+    }
+  ]
+}
+```
+
+### 3) Combined Fallback + Notify + Bridge Warm
+
+```json
+{
+  "version": 1,
+  "enabled": true,
+  "maxChainDepth": 2,
+  "defaultCooldownMs": 45000,
+  "defaultDedupWindowMs": 30000,
+  "rules": [
+    {
+      "id": "full-recovery-chain",
+      "trigger": {
+        "event": "TaskCompleted",
+        "outcomes": ["failed", "timeout"]
+      },
+      "condition": {
+        "errorIncludes": ["timeout", "capacity", "connection"]
+      },
+      "cooldownMs": 90000,
+      "actions": [
+        {
+          "type": "spawn_fallback_agent",
+          "id": "fallback-agent",
+          "description": "Recover failed task execution",
+          "promptTemplate": "Recover failed task and produce a concise fix summary.\nTask=${TASK_SUBJECT}\nError=${ERROR}\nPayload=${PAYLOAD_JSON}"
+        },
+        {
+          "type": "notify_team",
+          "id": "notify-team",
+          "recipients": ["*"],
+          "summary": "Recovery chain triggered",
+          "messageTemplate": "Recovery chain ${RULE_ID} fired.\nOutcome=${OUTCOME}\nTask=${TASK_SUBJECT}\nError=${ERROR}"
+        },
+        {
+          "type": "warm_remote_capacity",
+          "id": "warm-capacity",
+          "createDefaultEnvironmentIfMissing": false
+        }
+      ]
+    }
+  ]
+}
+```
+
+## Template Variables
+
+The following placeholders are supported by `promptTemplate`, `summary`, and `messageTemplate`:
+
+- `${EVENT_NAME}`
+- `${OUTCOME}`
+- `${RULE_ID}`
+- `${TASK_SUBJECT}`
+- `${TASK_DESCRIPTION}`
+- `${TASK_ID}`
+- `${ERROR}`
+- `${PAYLOAD_JSON}`
+
+## Troubleshooting
+
+### Rule never triggers
+
+- Verify `trigger.event` and `trigger.outcome`/`trigger.outcomes` exactly match dispatched event data.
+- Check `condition` filters (especially `toolNames` and `eventFieldEquals` dot-path keys).
+- Confirm the config file is valid JSON and schema-valid.
+
+### Actions show as skipped
+
+Common skip reasons:
+
+- `action disabled`
+- `rule cooldown active ...`
+- `dedup window active ...`
+- `max chain depth reached ...`
+- `No team context is available ...`
+- `Team file not found ...`
+- `Remote sessions are blocked by policy`
+- `Bridge is not active; warm_remote_capacity is a safe no-op`
+- `No fallback agent launcher is registered in runtime context`
+
+### Config changes not reflected
+
+- Loader uses memoization by file mtime/size.
+- Ensure your editor writes the file fully and updates mtime.
+- If needed, force reload from the caller side with `forceReloadConfig: true`.
+
+### Existing workflows changed unexpectedly
+
+- Set `"enabled": false` at top-level.
+- Or globally disable with `CLAUDE_CODE_ENABLE_HOOK_CHAINS=0`.
+- Re-enable gradually after validating one rule at a time.

vendor/openclaude/docs/litellm-setup.md

@@ -0,0 +1,144 @@
+# LiteLLM Setup
+
+OpenClaude can connect to LiteLLM through LiteLLM's OpenAI-compatible proxy.
+
+## Overview
+
+LiteLLM is an open-source LLM gateway that provides a unified API to 100+ model providers. By running the LiteLLM Proxy, you can route OpenClaude requests through LiteLLM to access any of its supported providers — all while using OpenClaude's existing OpenAI-compatible provider path.
+
+## Prerequisites
+
+- LiteLLM installed (`pip install litellm[proxy]`)
+- A `litellm_config.yaml` or equivalent LiteLLM configuration
+- LiteLLM Proxy running on a local or remote port
+
+## 1. Start the LiteLLM Proxy
+
+### Basic installation
+
+```bash
+pip install litellm[proxy]
+```
+
+### Configure LiteLLM
+
+Create a `litellm_config.yaml` with your desired model aliases:
+
+```yaml
+model_list:
+  - model_name: gpt-4o
+    litellm_params:
+      model: openai/gpt-4o
+      api_key: os.environ/OPENAI_API_KEY
+
+  - model_name: claude-sonnet-4
+    litellm_params:
+      model: anthropic/claude-sonnet-4-5-20250929
+      api_key: os.environ/ANTHROPIC_API_KEY
+
+  - model_name: gemini-2.5-flash
+    litellm_params:
+      model: gemini/gemini-2.5-flash
+      api_key: os.environ/GEMINI_API_KEY
+
+  - model_name: llama-3.3-70b
+    litellm_params:
+      model: together_ai/meta-llama/Llama-3.3-70B-Instruct-Turbo
+      api_key: os.environ/TOGETHER_API_KEY
+```
+
+### Run the proxy
+
+```bash
+litellm --config litellm_config.yaml --port 4000
+```
+
+The proxy will start at `http://localhost:4000` by default.
+
+## 2. Point OpenClaude to LiteLLM
+
+### Option A: Environment Variables
+
+```bash
+export CLAUDE_CODE_USE_OPENAI=1
+export OPENAI_BASE_URL=http://localhost:4000
+export OPENAI_API_KEY=<your-master-key-or-placeholder>
+export OPENAI_MODEL=<your-litellm-model-alias>
+openclaude
+```
+
+Replace `<your-litellm-model-alias>` with a model name from your `litellm_config.yaml` (e.g., `gpt-4o`, `claude-sonnet-4`, `gemini-2.5-flash`).
+
+### Option B: Using /provider
+
+1. Run `openclaude`
+2. Type `/provider` to open the provider setup flow
+3. Choose the **OpenAI-compatible** option
+4. When prompted for the API key, enter the key required by your LiteLLM proxy
+   If your local LiteLLM setup does not enforce auth, you may still need to enter a placeholder value
+      - 5. When prompted for the base URL, enter `http://localhost:4000`
+        6. 6. When prompted for the model, enter the LiteLLM model name or alias you configured
+           7. 7. Save the provider configuration
+
+## 3. Example LiteLLM Configs
+
+### Multi-provider routing with spend tracking
+
+```yaml
+model_list:
+  - model_name: gpt-4o
+    litellm_params:
+      model: openai/gpt-4o
+      api_key: os.environ/OPENAI_API_KEY
+
+  - model_name: claude-sonnet-4
+    litellm_params:
+      model: anthropic/claude-sonnet-4-5-20250929
+      api_key: os.environ/ANTHROPIC_API_KEY
+
+  - model_name: deepseek-chat
+    litellm_params:
+      model: deepseek/deepseek-chat
+      api_key: os.environ/DEEPSEEK_API_KEY
+
+litellm_settings:
+  set_verbose: false
+  num_retries: 3
+```
+
+### With a master key for auth
+
+```bash
+# Start proxy with a master key
+litellm --config litellm_config.yaml --port 4000 --master_key sk-my-master-key
+
+# Connect OpenClaude
+export CLAUDE_CODE_USE_OPENAI=1
+export OPENAI_BASE_URL=http://localhost:4000
+export OPENAI_API_KEY=sk-my-master-key
+export OPENAI_MODEL=gpt-4o
+openclaude
+```
+
+## 4. Notes
+
+- `OPENAI_MODEL` must match the **LiteLLM model alias** defined in your config, not the upstream raw provider model name.
+- If your proxy requires authentication, use the proxy key (or `master_key`) in `OPENAI_API_KEY`.
+- LiteLLM's OpenAI-compatible endpoint accepts the same request format as OpenAI, so OpenClaude works without any code changes.
+- You can switch between any provider configured in LiteLLM by simply changing the `OPENAI_MODEL` value — no need to reconfigure OpenClaude.
+
+## 5. Troubleshooting
+
+| Issue | Likely Cause | Fix |
+|-------|--------------|-----|
+| 404 or Model Not Found | Model alias doesn't exist in LiteLLM config | Verify the `model_name` in `litellm_config.yaml` matches `OPENAI_MODEL` |
+| Connection Refused | LiteLLM proxy isn't running | Start the proxy with `litellm --config litellm_config.yaml --port 4000` |
+| Auth Failed | Missing or wrong `master_key` | Set the correct key in `OPENAI_API_KEY` |
+| Upstream provider error | The backend provider key is missing or invalid | Ensure the upstream API key (e.g., `OPENAI_API_KEY`) is set in your LiteLLM proxy process environment |
+| Tools fail but chat works | The selected model has weak function/tool calling support | Switch to a model with strong tool support (e.g., GPT-4o, Claude Sonnet) |
+
+## 6. Resources
+
+- [LiteLLM Proxy Docs](https://docs.litellm.ai/docs/proxy/quick_start)
+- [LiteLLM Provider List](https://docs.litellm.ai/docs/providers)
+- [LiteLLM OpenAI-Compatible Endpoints](https://docs.litellm.ai/docs/proxy/openai_compatible_proxy)

vendor/openclaude/docs/non-technical-setup.md

@@ -0,0 +1,116 @@
+# OpenClaude for Non-Technical Users
+
+This guide is for people who want the easiest setup path.
+
+You do not need to build from source. You do not need Bun. You do not need to understand the full codebase.
+
+If you can copy and paste commands into a terminal, you can set this up.
+
+## What OpenClaude Does
+
+OpenClaude lets you use an AI coding assistant with different model providers such as:
+
+- OpenAI
+- DeepSeek
+- Gemini
+- Ollama
+- Codex
+
+For most first-time users, OpenAI is the easiest option.
+
+## Before You Start
+
+You need:
+
+1. Node.js 20 or newer installed
+2. A terminal window
+3. An API key from your provider, unless you are using a local model like Ollama
+
+## Fastest Path
+
+1. Install OpenClaude with npm
+2. Set 3 environment variables
+3. Run `openclaude`
+
+## Choose Your Operating System
+
+- Windows: [Windows Quick Start](quick-start-windows.md)
+- macOS / Linux: [macOS / Linux Quick Start](quick-start-mac-linux.md)
+
+## Which Provider Should You Choose?
+
+### OpenAI
+
+Choose this if:
+
+- you want the easiest setup
+- you already have an OpenAI API key
+
+### Ollama
+
+Choose this if:
+
+- you want to run models locally
+- you do not want to depend on a cloud API for testing
+
+### Codex
+
+Choose this if:
+
+- you already use the Codex CLI
+- you already have Codex or ChatGPT auth configured
+
+## What Success Looks Like
+
+After you run `openclaude`, the CLI should start and wait for your prompt.
+
+At that point, you can ask it to:
+
+- explain code
+- edit files
+- run commands
+- review changes
+
+## Common Problems
+
+### `openclaude` command not found
+
+Cause:
+
+- npm installed the package, but your terminal has not refreshed yet
+
+Fix:
+
+1. Close the terminal
+2. Open a new terminal
+3. Run `openclaude` again
+
+### Invalid API key
+
+Cause:
+
+- the key is wrong, expired, or copied incorrectly
+
+Fix:
+
+1. Get a fresh key from your provider
+2. Paste it again carefully
+3. Re-run `openclaude`
+
+### Ollama not working
+
+Cause:
+
+- Ollama is not installed or not running
+
+Fix:
+
+1. Install Ollama from `https://ollama.com/download`
+2. Start Ollama
+3. Try again
+
+## Want More Control?
+
+If you want source builds, advanced provider profiles, diagnostics, or Bun-based workflows, use:
+
+- [Advanced Setup](advanced-setup.md)

vendor/openclaude/docs/quick-start-mac-linux.md

@@ -0,0 +1,143 @@
+# OpenClaude Quick Start for macOS and Linux
+
+This guide uses a standard shell such as Terminal, iTerm, bash, or zsh.
+
+## 1. Install Node.js
+
+Install Node.js 20 or newer from:
+
+- `https://nodejs.org/`
+
+Then check it:
+
+```bash
+node --version
+npm --version
+```
+
+## 2. Install OpenClaude
+
+```bash
+npm install -g @gitlawb/openclaude
+```
+
+## 3. Pick One Provider
+
+### Option A: OpenAI
+
+Replace `sk-your-key-here` with your real key.
+
+```bash
+export CLAUDE_CODE_USE_OPENAI=1
+export OPENAI_API_KEY=sk-your-key-here
+export OPENAI_MODEL=gpt-4o
+
+openclaude
+```
+
+### Option B: DeepSeek
+
+```bash
+export CLAUDE_CODE_USE_OPENAI=1
+export OPENAI_API_KEY=sk-your-key-here
+export OPENAI_BASE_URL=https://api.deepseek.com/v1
+export OPENAI_MODEL=deepseek-chat
+
+openclaude
+```
+
+### Option C: Ollama
+
+Install Ollama first from:
+
+- `https://ollama.com/download`
+
+Then run:
+
+```bash
+ollama pull llama3.1:8b
+
+export CLAUDE_CODE_USE_OPENAI=1
+export OPENAI_BASE_URL=http://localhost:11434/v1
+export OPENAI_MODEL=llama3.1:8b
+
+openclaude
+```
+
+No API key is needed for Ollama local models.
+
+### Option D: LM Studio
+
+Install LM Studio first from:
+
+- `https://lmstudio.ai/`
+
+Then in LM Studio:
+
+1. Download a model (e.g., Llama 3.1 8B, Mistral 7B)
+2. Go to the "Developer" tab
+3. Select your model and enable the server via the toggle
+
+Then run:
+
+```bash
+export CLAUDE_CODE_USE_OPENAI=1
+export OPENAI_BASE_URL=http://localhost:1234/v1
+export OPENAI_MODEL=your-model-name
+# export OPENAI_API_KEY=lmstudio  # optional: some users need a dummy key
+
+openclaude
+```
+
+Replace `your-model-name` with the model name shown in LM Studio.
+
+No API key is needed for LM Studio local models (but uncomment the `OPENAI_API_KEY` line if you hit auth errors).
+
+## 4. If `openclaude` Is Not Found
+
+Close the terminal, open a new one, and try again:
+
+```bash
+openclaude
+```
+
+## 5. If Your Provider Fails
+
+Check the basics:
+
+### For OpenAI or DeepSeek
+
+- make sure the key is real
+- make sure you copied it fully
+
+### For Ollama
+
+- make sure Ollama is installed
+- make sure Ollama is running
+- make sure the model was pulled successfully
+
+### For LM Studio
+
+- make sure LM Studio is installed
+- make sure LM Studio is running
+- make sure the server is enabled (toggle on in the "Developer" tab)
+- make sure a model is loaded in LM Studio
+- make sure the model name matches what you set in `OPENAI_MODEL`
+
+## 6. Updating OpenClaude
+
+```bash
+npm install -g @gitlawb/openclaude@latest
+```
+
+## 7. Uninstalling OpenClaude
+
+```bash
+npm uninstall -g @gitlawb/openclaude
+```
+
+## Need Advanced Setup?
+
+Use:
+
+- [Advanced Setup](advanced-setup.md)

vendor/openclaude/docs/quick-start-windows.md

@@ -0,0 +1,143 @@
+# OpenClaude Quick Start for Windows
+
+This guide uses Windows PowerShell.
+
+## 1. Install Node.js
+
+Install Node.js 20 or newer from:
+
+- `https://nodejs.org/`
+
+Then open PowerShell and check it:
+
+```powershell
+node --version
+npm --version
+```
+
+## 2. Install OpenClaude
+
+```powershell
+npm install -g @gitlawb/openclaude
+```
+
+## 3. Pick One Provider
+
+### Option A: OpenAI
+
+Replace `sk-your-key-here` with your real key.
+
+```powershell
+$env:CLAUDE_CODE_USE_OPENAI="1"
+$env:OPENAI_API_KEY="sk-your-key-here"
+$env:OPENAI_MODEL="gpt-4o"
+
+openclaude
+```
+
+### Option B: DeepSeek
+
+```powershell
+$env:CLAUDE_CODE_USE_OPENAI="1"
+$env:OPENAI_API_KEY="sk-your-key-here"
+$env:OPENAI_BASE_URL="https://api.deepseek.com/v1"
+$env:OPENAI_MODEL="deepseek-chat"
+
+openclaude
+```
+
+### Option C: Ollama
+
+Install Ollama first from:
+
+- `https://ollama.com/download/windows`
+
+Then run:
+
+```powershell
+ollama pull llama3.1:8b
+
+$env:CLAUDE_CODE_USE_OPENAI="1"
+$env:OPENAI_BASE_URL="http://localhost:11434/v1"
+$env:OPENAI_MODEL="llama3.1:8b"
+
+openclaude
+```
+
+No API key is needed for Ollama local models.
+
+### Option D: LM Studio
+
+Install LM Studio first from:
+
+- `https://lmstudio.ai/`
+
+Then in LM Studio:
+
+1. Download a model (e.g., Llama 3.1 8B, Mistral 7B)
+2. Go to the "Developer" tab
+3. Select your model and enable the server via the toggle
+
+Then run:
+
+```powershell
+$env:CLAUDE_CODE_USE_OPENAI="1"
+$env:OPENAI_BASE_URL="http://localhost:1234/v1"
+$env:OPENAI_MODEL="your-model-name"
+# $env:OPENAI_API_KEY="lmstudio"  # optional: some users need a dummy key
+
+openclaude
+```
+
+Replace `your-model-name` with the model name shown in LM Studio.
+
+No API key is needed for LM Studio local models (but uncomment the `OPENAI_API_KEY` line if you hit auth errors).
+
+## 4. If `openclaude` Is Not Found
+
+Close PowerShell, open a new one, and try again:
+
+```powershell
+openclaude
+```
+
+## 5. If Your Provider Fails
+
+Check the basics:
+
+### For OpenAI or DeepSeek
+
+- make sure the key is real
+- make sure you copied it fully
+
+### For Ollama
+
+- make sure Ollama is installed
+- make sure Ollama is running
+- make sure the model was pulled successfully
+
+### For LM Studio
+
+- make sure LM Studio is installed
+- make sure LM Studio is running
+- make sure the server is enabled (toggle on in the "Developer" tab)
+- make sure a model is loaded in LM Studio
+- make sure the model name matches what you set in `OPENAI_MODEL`
+
+## 6. Updating OpenClaude
+
+```powershell
+npm install -g @gitlawb/openclaude@latest
+```
+
+## 7. Uninstalling OpenClaude
+
+```powershell
+npm uninstall -g @gitlawb/openclaude
+```
+
+## Need Advanced Setup?
+
+Use:
+
+- [Advanced Setup](advanced-setup.md)

vendor/openclaude/package.json

@@ -0,0 +1,162 @@
+{
+  "name": "@gitlawb/openclaude",
+  "version": "0.6.0",
+  "description": "Claude Code opened to any LLM — OpenAI, Gemini, DeepSeek, Ollama, and 200+ models",
+  "type": "module",
+  "bin": {
+    "openclaude": "./bin/openclaude"
+  },
+  "files": [
+    "bin/",
+    "dist/cli.mjs",
+    "README.md"
+  ],
+  "scripts": {
+    "build": "bun run scripts/build.ts",
+    "dev": "bun run build && node dist/cli.mjs",
+    "dev:profile": "bun run scripts/provider-launch.ts",
+    "dev:profile:fast": "bun run scripts/provider-launch.ts auto --fast --bare",
+    "dev:codex": "bun run scripts/provider-launch.ts codex",
+    "dev:openai": "bun run scripts/provider-launch.ts openai",
+    "dev:gemini": "bun run scripts/provider-launch.ts gemini",
+    "dev:ollama": "bun run scripts/provider-launch.ts ollama",
+    "dev:ollama:fast": "bun run scripts/provider-launch.ts ollama --fast --bare",
+    "dev:atomic-chat": "bun run scripts/provider-launch.ts atomic-chat",
+    "profile:init": "bun run scripts/provider-bootstrap.ts",
+    "profile:recommend": "bun run scripts/provider-recommend.ts",
+    "profile:auto": "bun run scripts/provider-recommend.ts --apply",
+    "profile:codex": "bun run profile:init -- --provider codex --model codexplan",
+    "profile:fast": "bun run profile:init -- --provider ollama --model llama3.2:3b",
+    "profile:code": "bun run profile:init -- --provider ollama --model qwen2.5-coder:7b",
+    "dev:fast": "bun run profile:fast && bun run dev:ollama:fast",
+    "dev:code": "bun run profile:code && bun run dev:profile",
+    "dev:grpc": "bun run scripts/start-grpc.ts",
+    "dev:grpc:cli": "bun run scripts/grpc-cli.ts",
+    "start": "node dist/cli.mjs",
+    "test": "bun test",
+    "test:coverage": "bun test --coverage --coverage-reporter=lcov --coverage-dir=coverage --max-concurrency=1 && bun run scripts/render-coverage-heatmap.ts",
+    "test:coverage:ui": "bun run scripts/render-coverage-heatmap.ts",
+    "security:pr-scan": "bun run scripts/pr-intent-scan.ts",
+    "test:provider-recommendation": "bun test src/utils/providerRecommendation.test.ts src/utils/providerProfile.test.ts",
+    "typecheck": "tsc --noEmit",
+    "smoke": "bun run build && node dist/cli.mjs --version",
+    "verify:privacy": "bun run scripts/verify-no-phone-home.ts",
+    "build:verified": "bun run build && bun run verify:privacy",
+    "test:provider": "bun test src/services/api/*.test.ts src/utils/context.test.ts",
+    "doctor:runtime": "bun run scripts/system-check.ts",
+    "doctor:runtime:json": "bun run scripts/system-check.ts --json",
+    "doctor:report": "bun run scripts/system-check.ts --out reports/doctor-runtime.json",
+    "hardening:check": "bun run smoke && bun run doctor:runtime",
+    "hardening:strict": "bun run typecheck && bun run hardening:check",
+    "prepack": "npm run build"
+  },
+  "dependencies": {
+    "@alcalzone/ansi-tokenize": "0.3.0",
+    "@anthropic-ai/bedrock-sdk": "0.26.4",
+    "@anthropic-ai/foundry-sdk": "0.2.3",
+    "@anthropic-ai/sandbox-runtime": "0.0.46",
+    "@anthropic-ai/sdk": "0.81.0",
+    "@anthropic-ai/vertex-sdk": "0.14.4",
+    "@commander-js/extra-typings": "12.1.0",
+    "@growthbook/growthbook": "1.6.5",
+    "@grpc/grpc-js": "^1.14.3",
+    "@grpc/proto-loader": "^0.8.0",
+    "@mendable/firecrawl-js": "4.18.1",
+    "@modelcontextprotocol/sdk": "1.29.0",
+    "@opentelemetry/api": "1.9.1",
+    "@opentelemetry/api-logs": "0.214.0",
+    "@opentelemetry/core": "2.6.1",
+    "@opentelemetry/exporter-logs-otlp-http": "0.214.0",
+    "@opentelemetry/exporter-trace-otlp-grpc": "0.57.2",
+    "@opentelemetry/resources": "2.6.1",
+    "@opentelemetry/sdk-logs": "0.214.0",
+    "@opentelemetry/sdk-metrics": "2.6.1",
+    "@opentelemetry/sdk-trace-base": "2.6.1",
+    "@opentelemetry/sdk-trace-node": "2.6.1",
+    "@opentelemetry/semantic-conventions": "1.40.0",
+    "ajv": "8.18.0",
+    "auto-bind": "5.0.1",
+    "axios": "1.15.0",
+    "bidi-js": "1.0.3",
+    "chalk": "5.6.2",
+    "chokidar": "4.0.3",
+    "cli-boxes": "3.0.0",
+    "cli-highlight": "2.1.11",
+    "code-excerpt": "4.0.0",
+    "commander": "12.1.0",
+    "cross-spawn": "7.0.6",
+    "diff": "8.0.3",
+    "duck-duck-scrape": "^2.2.7",
+    "emoji-regex": "10.6.0",
+    "env-paths": "3.0.0",
+    "execa": "9.6.1",
+    "fflate": "0.8.2",
+    "figures": "6.1.0",
+    "fuse.js": "7.1.0",
+    "get-east-asian-width": "1.5.0",
+    "google-auth-library": "9.15.1",
+    "https-proxy-agent": "7.0.6",
+    "ignore": "7.0.5",
+    "indent-string": "5.0.0",
+    "jsonc-parser": "3.3.1",
+    "lodash-es": "4.18.1",
+    "lru-cache": "11.2.7",
+    "marked": "15.0.12",
+    "p-map": "7.0.4",
+    "picomatch": "4.0.4",
+    "proper-lockfile": "4.1.2",
+    "qrcode": "1.5.4",
+    "react": "19.2.4",
+    "react-compiler-runtime": "1.0.0",
+    "react-reconciler": "0.33.0",
+    "semver": "7.7.4",
+    "sharp": "^0.34.5",
+    "shell-quote": "1.8.3",
+    "signal-exit": "4.1.0",
+    "stack-utils": "2.0.6",
+    "strip-ansi": "7.2.0",
+    "supports-hyperlinks": "3.2.0",
+    "tree-kill": "1.2.2",
+    "turndown": "7.2.2",
+    "type-fest": "4.41.0",
+    "undici": "7.24.6",
+    "usehooks-ts": "3.1.1",
+    "vscode-languageserver-protocol": "3.17.5",
+    "wrap-ansi": "9.0.2",
+    "ws": "8.20.0",
+    "xss": "1.0.15",
+    "yaml": "2.8.3",
+    "zod": "3.25.76"
+  },
+  "devDependencies": {
+    "@types/bun": "1.3.11",
+    "@types/node": "25.5.0",
+    "@types/react": "19.2.14",
+    "tsx": "^4.21.0",
+    "typescript": "5.9.3"
+  },
+  "engines": {
+    "node": ">=20.0.0"
+  },
+  "repository": {
+    "type": "git",
+    "url": "https://github.com/Gitlawb/openclaude.git"
+  },
+  "keywords": [
+    "claude-code",
+    "openai",
+    "llm",
+    "cli",
+    "agent",
+    "deepseek",
+    "ollama",
+    "gemini"
+  ],
+  "license": "SEE LICENSE FILE",
+  "publishConfig": {
+    "access": "public"
+  },
+  "overrides": {
+    "lodash-es": "4.18.1"
+  }
+}

vendor/openclaude/python/__init__.py

@@ -0,0 +1 @@
+# Python helper package for standalone provider-side utilities.

vendor/openclaude/python/atomic_chat_provider.py

@@ -0,0 +1,146 @@
+"""
+atomic_chat_provider.py
+-----------------------
+Adds native Atomic Chat support to openclaude.
+Lets Claude Code route requests to any locally-running model via
+Atomic Chat (Apple Silicon only) at 127.0.0.1:1337.
+
+Atomic Chat exposes an OpenAI-compatible API, so messages are forwarded
+directly without translation.
+
+Usage (.env):
+    PREFERRED_PROVIDER=atomic-chat
+    ATOMIC_CHAT_BASE_URL=http://127.0.0.1:1337
+"""
+
+import httpx
+import json
+import logging
+import os
+from typing import AsyncIterator
+
+logger = logging.getLogger(__name__)
+ATOMIC_CHAT_BASE_URL = os.getenv("ATOMIC_CHAT_BASE_URL", "http://127.0.0.1:1337")
+
+
+def _api_url(path: str) -> str:
+    return f"{ATOMIC_CHAT_BASE_URL}/v1{path}"
+
+
+async def check_atomic_chat_running() -> bool:
+    try:
+        async with httpx.AsyncClient(timeout=3.0) as client:
+            resp = await client.get(_api_url("/models"))
+            return resp.status_code == 200
+    except Exception:
+        return False
+
+
+async def list_atomic_chat_models() -> list[str]:
+    try:
+        async with httpx.AsyncClient(timeout=5.0) as client:
+            resp = await client.get(_api_url("/models"))
+            resp.raise_for_status()
+            data = resp.json()
+            return [m["id"] for m in data.get("data", [])]
+    except Exception as e:
+        logger.warning(f"Could not list Atomic Chat models: {e}")
+        return []
+
+
+async def atomic_chat(
+    model: str,
+    messages: list[dict],
+    system: str | None = None,
+    max_tokens: int = 4096,
+    temperature: float = 1.0,
+) -> dict:
+    chat_messages = list(messages)
+    if system:
+        chat_messages.insert(0, {"role": "system", "content": system})
+
+    payload = {
+        "model": model,
+        "messages": chat_messages,
+        "max_tokens": max_tokens,
+        "temperature": temperature,
+        "stream": False,
+    }
+
+    async with httpx.AsyncClient(timeout=120.0) as client:
+        resp = await client.post(_api_url("/chat/completions"), json=payload)
+        resp.raise_for_status()
+        data = resp.json()
+
+    choice = data.get("choices", [{}])[0]
+    assistant_text = choice.get("message", {}).get("content", "")
+    usage = data.get("usage", {})
+
+    return {
+        "id": data.get("id", "msg_atomic_chat"),
+        "type": "message",
+        "role": "assistant",
+        "content": [{"type": "text", "text": assistant_text}],
+        "model": model,
+        "stop_reason": "end_turn",
+        "stop_sequence": None,
+        "usage": {
+            "input_tokens": usage.get("prompt_tokens", 0),
+            "output_tokens": usage.get("completion_tokens", 0),
+        },
+    }
+
+
+async def atomic_chat_stream(
+    model: str,
+    messages: list[dict],
+    system: str | None = None,
+    max_tokens: int = 4096,
+    temperature: float = 1.0,
+) -> AsyncIterator[str]:
+    chat_messages = list(messages)
+    if system:
+        chat_messages.insert(0, {"role": "system", "content": system})
+
+    payload = {
+        "model": model,
+        "messages": chat_messages,
+        "max_tokens": max_tokens,
+        "temperature": temperature,
+        "stream": True,
+    }
+
+    yield "event: message_start\n"
+    yield f'data: {json.dumps({"type": "message_start", "message": {"id": "msg_atomic_chat_stream", "type": "message", "role": "assistant", "content": [], "model": model, "stop_reason": None, "usage": {"input_tokens": 0, "output_tokens": 0}}})}\n\n'
+    yield "event: content_block_start\n"
+    yield f'data: {json.dumps({"type": "content_block_start", "index": 0, "content_block": {"type": "text", "text": ""}})}\n\n'
+
+    async with httpx.AsyncClient(timeout=120.0) as client:
+        async with client.stream("POST", _api_url("/chat/completions"), json=payload) as resp:
+            resp.raise_for_status()
+            async for line in resp.aiter_lines():
+                if not line or not line.startswith("data: "):
+                    continue
+                raw = line[len("data: "):]
+                if raw.strip() == "[DONE]":
+                    break
+                try:
+                    chunk = json.loads(raw)
+                    delta = chunk.get("choices", [{}])[0].get("delta", {})
+                    delta_text = delta.get("content", "")
+                    if delta_text:
+                        yield "event: content_block_delta\n"
+                        yield f'data: {json.dumps({"type": "content_block_delta", "index": 0, "delta": {"type": "text_delta", "text": delta_text}})}\n\n'
+
+                    finish_reason = chunk.get("choices", [{}])[0].get("finish_reason")
+                    if finish_reason:
+                        usage = chunk.get("usage", {})
+                        yield "event: content_block_stop\n"
+                        yield f'data: {json.dumps({"type": "content_block_stop", "index": 0})}\n\n'
+                        yield "event: message_delta\n"
+                        yield f'data: {json.dumps({"type": "message_delta", "delta": {"stop_reason": "end_turn", "stop_sequence": None}, "usage": {"output_tokens": usage.get("completion_tokens", 0)}})}\n\n'
+                        yield "event: message_stop\n"
+                        yield f'data: {json.dumps({"type": "message_stop"})}\n\n'
+                        break
+                except json.JSONDecodeError:
+                    continue

vendor/openclaude/python/ollama_provider.py

@@ -0,0 +1,173 @@
+"""
+ollama_provider.py
+------------------
+Adds native Ollama support to openclaude.
+Lets Claude Code route requests to any locally-running Ollama model
+(llama3, mistral, codellama, phi3, qwen2, deepseek-coder, etc.)
+without needing an API key.
+
+Usage (.env):
+    PREFERRED_PROVIDER=ollama
+    OLLAMA_BASE_URL=http://localhost:11434
+    BIG_MODEL=codellama:34b
+    SMALL_MODEL=llama3:8b
+"""
+
+import httpx
+import logging
+import os
+from typing import AsyncIterator
+
+logger = logging.getLogger(__name__)
+OLLAMA_BASE_URL = os.getenv("OLLAMA_BASE_URL", "http://localhost:11434")
+
+
+async def check_ollama_running() -> bool:
+    try:
+        async with httpx.AsyncClient(timeout=3.0) as client:
+            resp = await client.get(f"{OLLAMA_BASE_URL}/api/tags")
+            return resp.status_code == 200
+    except Exception:
+        return False
+
+
+async def list_ollama_models() -> list[str]:
+    try:
+        async with httpx.AsyncClient(timeout=5.0) as client:
+            resp = await client.get(f"{OLLAMA_BASE_URL}/api/tags")
+            resp.raise_for_status()
+            data = resp.json()
+            return [m["name"] for m in data.get("models", [])]
+    except Exception as e:
+        logger.warning(f"Could not list Ollama models: {e}")
+        return []
+
+
+def normalize_ollama_model(model_name: str) -> str:
+    if model_name.startswith("ollama/"):
+        return model_name[len("ollama/"):]
+    return model_name
+
+
+def _extract_ollama_image_data(block: dict) -> str | None:
+    source = block.get("source")
+    if not isinstance(source, dict):
+        return None
+    if source.get("type") != "base64":
+        return None
+    data = source.get("data")
+    if isinstance(data, str) and data:
+        return data
+    return None
+
+
+def anthropic_to_ollama_messages(messages: list[dict]) -> list[dict]:
+    ollama_messages = []
+    for msg in messages:
+        role = msg.get("role", "user")
+        content = msg.get("content", "")
+        if isinstance(content, str):
+            ollama_messages.append({"role": role, "content": content})
+        elif isinstance(content, list):
+            text_parts = []
+            image_parts = []
+            for block in content:
+                if isinstance(block, dict):
+                    if block.get("type") == "text":
+                        text_parts.append(block.get("text", ""))
+                    elif block.get("type") == "image":
+                        image_data = _extract_ollama_image_data(block)
+                        if image_data:
+                            image_parts.append(image_data)
+                        else:
+                            text_parts.append("[image]")
+                elif isinstance(block, str):
+                    text_parts.append(block)
+            ollama_message = {"role": role, "content": "\n".join(text_parts)}
+            if image_parts:
+                ollama_message["images"] = image_parts
+            ollama_messages.append(ollama_message)
+    return ollama_messages
+
+
+async def ollama_chat(
+    model: str,
+    messages: list[dict],
+    system: str | None = None,
+    max_tokens: int = 4096,
+    temperature: float = 1.0,
+) -> dict:
+    model = normalize_ollama_model(model)
+    ollama_messages = anthropic_to_ollama_messages(messages)
+    if system:
+        ollama_messages.insert(0, {"role": "system", "content": system})
+    payload = {
+        "model": model,
+        "messages": ollama_messages,
+        "stream": False,
+        "options": {"num_predict": max_tokens, "temperature": temperature},
+    }
+    async with httpx.AsyncClient(timeout=120.0) as client:
+        resp = await client.post(f"{OLLAMA_BASE_URL}/api/chat", json=payload)
+        resp.raise_for_status()
+        data = resp.json()
+    assistant_text = data.get("message", {}).get("content", "")
+    return {
+        "id": f"msg_ollama_{data.get('created_at', 'unknown')}",
+        "type": "message",
+        "role": "assistant",
+        "content": [{"type": "text", "text": assistant_text}],
+        "model": model,
+        "stop_reason": "end_turn",
+        "stop_sequence": None,
+        "usage": {
+            "input_tokens": data.get("prompt_eval_count", 0),
+            "output_tokens": data.get("eval_count", 0),
+        },
+    }
+
+
+async def ollama_chat_stream(
+    model: str,
+    messages: list[dict],
+    system: str | None = None,
+    max_tokens: int = 4096,
+    temperature: float = 1.0,
+) -> AsyncIterator[str]:
+    import json
+    model = normalize_ollama_model(model)
+    ollama_messages = anthropic_to_ollama_messages(messages)
+    if system:
+        ollama_messages.insert(0, {"role": "system", "content": system})
+    payload = {
+        "model": model,
+        "messages": ollama_messages,
+        "stream": True,
+        "options": {"num_predict": max_tokens, "temperature": temperature},
+    }
+    yield "event: message_start\n"
+    yield f'data: {json.dumps({"type": "message_start", "message": {"id": "msg_ollama_stream", "type": "message", "role": "assistant", "content": [], "model": model, "stop_reason": None, "usage": {"input_tokens": 0, "output_tokens": 0}}})}\n\n'
+    yield "event: content_block_start\n"
+    yield f'data: {json.dumps({"type": "content_block_start", "index": 0, "content_block": {"type": "text", "text": ""}})}\n\n'
+    async with httpx.AsyncClient(timeout=120.0) as client:
+        async with client.stream("POST", f"{OLLAMA_BASE_URL}/api/chat", json=payload) as resp:
+            resp.raise_for_status()
+            async for line in resp.aiter_lines():
+                if not line:
+                    continue
+                try:
+                    chunk = json.loads(line)
+                    delta_text = chunk.get("message", {}).get("content", "")
+                    if delta_text:
+                        yield "event: content_block_delta\n"
+                        yield f'data: {json.dumps({"type": "content_block_delta", "index": 0, "delta": {"type": "text_delta", "text": delta_text}})}\n\n'
+                    if chunk.get("done"):
+                        yield "event: content_block_stop\n"
+                        yield f'data: {json.dumps({"type": "content_block_stop", "index": 0})}\n\n'
+                        yield "event: message_delta\n"
+                        yield f'data: {json.dumps({"type": "message_delta", "delta": {"stop_reason": "end_turn", "stop_sequence": None}, "usage": {"output_tokens": chunk.get("eval_count", 0)}})}\n\n'
+                        yield "event: message_stop\n"
+                        yield f'data: {json.dumps({"type": "message_stop"})}\n\n'
+                        break
+                except json.JSONDecodeError:
+                    continue

vendor/openclaude/python/requirements.txt

@@ -0,0 +1,3 @@
+pytest==7.4.4
+pytest-asyncio==0.23.3
+httpx==0.25.2

vendor/openclaude/python/smart_router.py

@@ -0,0 +1,387 @@
+"""
+smart_router.py
+---------------
+Intelligent auto-router for openclaude.
+
+Instead of always using one fixed provider, the smart router:
+- Pings all configured providers on startup
+- Scores them by latency, cost, and health
+- Routes each request to the optimal provider
+- Falls back automatically if a provider fails
+- Learns from real request timings over time
+
+Usage in server.py:
+    from smart_router import SmartRouter
+    router = SmartRouter()
+    await router.initialize()
+    result = await router.route(messages, model, stream)
+
+.env config:
+    ROUTER_MODE=smart          # or: fixed (default behaviour)
+    ROUTER_STRATEGY=latency    # or: cost, balanced
+    ROUTER_FALLBACK=true       # auto-retry on failure
+
+Contribution to: https://github.com/Gitlawb/openclaude
+"""
+
+import asyncio
+import logging
+import os
+import time
+from dataclasses import dataclass, field
+from typing import Optional
+import httpx
+
+logger = logging.getLogger(__name__)
+
+# ── Provider definitions ──────────────────────────────────────────────────────
+
+@dataclass
+class Provider:
+    name: str                        # e.g. "openai", "gemini", "ollama"
+    ping_url: str                    # URL used to check health
+    api_key_env: str                 # env var name for API key
+    cost_per_1k_tokens: float        # estimated cost USD per 1k tokens
+    big_model: str                   # model for sonnet/large requests
+    small_model: str                 # model for haiku/small requests
+    latency_ms: float = 9999.0       # updated by benchmark
+    healthy: bool = True             # updated by health checks
+    request_count: int = 0           # total requests routed here
+    error_count: int = 0             # total errors from this provider
+    avg_latency_ms: float = 9999.0   # rolling average from real requests
+
+    @property
+    def api_key(self) -> Optional[str]:
+        return os.getenv(self.api_key_env)
+
+    @property
+    def is_configured(self) -> bool:
+        """True if the provider has an API key set."""
+        if self.name in ("ollama", "atomic-chat"):
+            return True  # Local providers need no API key
+        return bool(self.api_key)
+
+    @property
+    def error_rate(self) -> float:
+        if self.request_count == 0:
+            return 0.0
+        return self.error_count / self.request_count
+
+    def score(self, strategy: str = "balanced") -> float:
+        """
+        Lower score = better provider.
+        strategy: 'latency' | 'cost' | 'balanced'
+        """
+        if not self.healthy or not self.is_configured:
+            return float("inf")
+
+        latency_score = self.avg_latency_ms / 1000.0   # normalize to seconds
+        cost_score = self.cost_per_1k_tokens * 100      # normalize to similar scale
+        error_penalty = self.error_rate * 500           # heavy penalty for errors
+
+        if strategy == "latency":
+            return latency_score + error_penalty
+        elif strategy == "cost":
+            return cost_score + error_penalty
+        else:  # balanced
+            return (latency_score * 0.5) + (cost_score * 0.5) + error_penalty
+
+
+# ── Default provider catalogue ────────────────────────────────────────────────
+
+def build_default_providers() -> list[Provider]:
+    big = os.getenv("BIG_MODEL", "gpt-4.1")
+    small = os.getenv("SMALL_MODEL", "gpt-4.1-mini")
+    ollama_url = os.getenv("OLLAMA_BASE_URL", "http://localhost:11434")
+    atomic_chat_url = os.getenv("ATOMIC_CHAT_BASE_URL", "http://127.0.0.1:1337")
+
+    return [
+        Provider(
+            name="openai",
+            ping_url="https://api.openai.com/v1/models",
+            api_key_env="OPENAI_API_KEY",
+            cost_per_1k_tokens=0.002,
+            big_model=big if "gpt" in big else "gpt-4.1",
+            small_model=small if "gpt" in small else "gpt-4.1-mini",
+        ),
+        Provider(
+            name="gemini",
+            ping_url="https://generativelanguage.googleapis.com/v1/models",
+            api_key_env="GEMINI_API_KEY",
+            cost_per_1k_tokens=0.0005,
+            big_model=big if "gemini" in big else "gemini-2.5-pro",
+            small_model=small if "gemini" in small else "gemini-2.0-flash",
+        ),
+        Provider(
+          name="mistral",
+          ping_url="",
+          api_key_env="MISTRAL_API_KEY",
+          cost_per_1k_tokens=0.0001,
+          big_model=big if "mistral" in big else "devstral-latest",
+          small_model=small if "small" in small else "ministral-3b-latest",  
+        ),
+        Provider(
+            name="ollama",
+            ping_url=f"{ollama_url}/api/tags",
+            api_key_env="",
+            cost_per_1k_tokens=0.0,   # free — local
+            big_model=big if "gemini" not in big and "gpt" not in big else "llama3:8b",
+            small_model=small if "gemini" not in small and "gpt" not in small else "llama3:8b",
+        ),
+        Provider(
+            name="atomic-chat",
+            ping_url=f"{atomic_chat_url}/v1/models",
+            api_key_env="",
+            cost_per_1k_tokens=0.0,   # free — local (Apple Silicon)
+            big_model=big if "gemini" not in big and "gpt" not in big else "llama3:8b",
+            small_model=small if "gemini" not in small and "gpt" not in small else "llama3:8b",
+        ),
+    ]
+
+
+# ── Smart Router ──────────────────────────────────────────────────────────────
+
+class SmartRouter:
+    """
+    Intelligently routes Claude Code API requests to the best
+    available LLM provider based on latency, cost, and health.
+    """
+
+    def __init__(
+        self,
+        providers: Optional[list[Provider]] = None,
+        strategy: Optional[str] = None,
+        fallback_enabled: Optional[bool] = None,
+    ):
+        self.providers = providers or build_default_providers()
+        self.strategy = strategy or os.getenv("ROUTER_STRATEGY", "balanced")
+        self.fallback_enabled = (
+            fallback_enabled
+            if fallback_enabled is not None
+            else os.getenv("ROUTER_FALLBACK", "true").lower() == "true"
+        )
+        self._initialized = False
+
+    # ── Initialization ────────────────────────────────────────────────────────
+
+    async def initialize(self) -> None:
+        """Ping all providers and build initial latency scores."""
+        logger.info("SmartRouter: benchmarking providers...")
+        await asyncio.gather(
+            *[self._ping_provider(p) for p in self.providers],
+            return_exceptions=True,
+        )
+        available = [p for p in self.providers if p.healthy and p.is_configured]
+        logger.info(
+            f"SmartRouter ready. Available providers: "
+            f"{[p.name for p in available]}"
+        )
+        if not available:
+            logger.warning(
+                "SmartRouter: no providers available! "
+                "Check your API keys in .env"
+            )
+        self._initialized = True
+
+    async def _ping_provider(self, provider: Provider) -> None:
+        """Measure latency to a provider's health endpoint."""
+        if not provider.is_configured:
+            provider.healthy = False
+            logger.debug(f"SmartRouter: {provider.name} skipped — no API key")
+            return
+
+        headers = {}
+        if provider.api_key:
+            headers["Authorization"] = f"Bearer {provider.api_key}"
+
+        start = time.monotonic()
+        try:
+            async with httpx.AsyncClient(timeout=5.0) as client:
+                resp = await client.get(provider.ping_url, headers=headers)
+                elapsed_ms = (time.monotonic() - start) * 1000
+                if resp.status_code in (200, 400, 401, 403):
+                    # 400/401/403 means reachable, just possibly bad key
+                    # We still mark healthy for routing purposes
+                    provider.healthy = True
+                    provider.latency_ms = elapsed_ms
+                    provider.avg_latency_ms = elapsed_ms
+                    logger.info(
+                        f"SmartRouter: {provider.name} OK "
+                        f"({elapsed_ms:.0f}ms, status={resp.status_code})"
+                    )
+                else:
+                    provider.healthy = False
+                    logger.warning(
+                        f"SmartRouter: {provider.name} unhealthy "
+                        f"(status={resp.status_code})"
+                    )
+        except Exception as e:
+            provider.healthy = False
+            logger.warning(f"SmartRouter: {provider.name} unreachable — {e}")
+
+    # ── Routing logic ─────────────────────────────────────────────────────────
+
+    def select_provider(self, is_large_request: bool = False) -> Optional[Provider]:
+        """
+        Pick the best available provider for this request.
+        Returns None if no providers are available.
+        """
+        available = [
+            p for p in self.providers
+            if p.healthy and p.is_configured
+        ]
+        if not available:
+            return None
+
+        return min(available, key=lambda p: p.score(self.strategy))
+
+    def get_model_for_provider(
+        self,
+        provider: Provider,
+        claude_model: str,
+        is_large_request: bool = False,
+    ) -> str:
+        """Map a Claude model name to the provider's actual model."""
+        if is_large_request:
+            return provider.big_model
+        is_large = any(
+            keyword in claude_model.lower()
+            for keyword in ["opus", "sonnet", "large", "big"]
+        )
+        return provider.big_model if is_large else provider.small_model
+
+    def is_large_request(self, messages: list[dict]) -> bool:
+        """Estimate if this is a large request based on message length."""
+        total_chars = sum(
+            len(str(m.get("content", ""))) for m in messages
+        )
+        return total_chars > 2000  # >2000 chars = treat as large
+
+    def _update_latency(self, provider: Provider, duration_ms: float) -> None:
+        """Exponential moving average update for latency tracking."""
+        alpha = 0.3  # weight for new observation
+        provider.avg_latency_ms = (
+            alpha * duration_ms + (1 - alpha) * provider.avg_latency_ms
+        )
+
+    # ── Main routing entry point ──────────────────────────────────────────────
+
+    async def route(
+        self,
+        messages: list[dict],
+        claude_model: str = "claude-sonnet",
+        attempt: int = 0,
+        exclude_providers: Optional[list[str]] = None,
+    ) -> dict:
+        """
+        Route a request to the best provider.
+        Returns a dict with routing decision info:
+          {
+            "provider": provider name,
+            "model": actual model to use,
+            "api_key": API key for the provider,
+            "base_url": base URL for the provider,
+          }
+        Raises RuntimeError if no providers available.
+        """
+        if not self._initialized:
+            await self.initialize()
+
+        exclude = set(exclude_providers or [])
+        large = self.is_large_request(messages)
+
+        available = [
+            p for p in self.providers
+            if p.healthy and p.is_configured and p.name not in exclude
+        ]
+
+        if not available:
+            raise RuntimeError(
+                "SmartRouter: no providers available. "
+                "Check your API keys and provider health."
+            )
+
+        provider = min(available, key=lambda p: p.score(self.strategy))
+        model = self.get_model_for_provider(
+            provider,
+            claude_model,
+            is_large_request=large,
+        )
+
+        logger.debug(
+            f"SmartRouter: routing to {provider.name}/{model} "
+            f"(strategy={self.strategy}, large={large}, attempt={attempt})"
+        )
+
+        return {
+            "provider": provider.name,
+            "model": model,
+            "api_key": provider.api_key or "none",
+            "provider_object": provider,
+        }
+
+    async def record_result(
+        self,
+        provider_name: str,
+        success: bool,
+        duration_ms: float,
+    ) -> None:
+        """
+        Record the outcome of a request.
+        Called after each proxied request to update provider scores.
+        """
+        provider = next(
+            (p for p in self.providers if p.name == provider_name), None
+        )
+        if not provider:
+            return
+
+        provider.request_count += 1
+        if success:
+            self._update_latency(provider, duration_ms)
+        else:
+            provider.error_count += 1
+            # After 3 consecutive failures, mark unhealthy temporarily
+            recent_errors = provider.error_count
+            recent_total = provider.request_count
+            if recent_total >= 3 and (recent_errors / recent_total) > 0.7:
+                logger.warning(
+                    f"SmartRouter: {provider_name} error rate high "
+                    f"({provider.error_rate:.0%}), marking unhealthy"
+                )
+                provider.healthy = False
+                # Schedule re-check after 60s
+                asyncio.create_task(self._recheck_provider(provider, delay=60))
+
+    async def _recheck_provider(
+        self, provider: Provider, delay: float = 60
+    ) -> None:
+        """Re-ping a provider after a delay and restore if healthy."""
+        await asyncio.sleep(delay)
+        await self._ping_provider(provider)
+        if provider.healthy:
+            logger.info(
+                f"SmartRouter: {provider.name} recovered, "
+                f"re-adding to pool"
+            )
+
+    # ── Status report ─────────────────────────────────────────────────────────
+
+    def status(self) -> list[dict]:
+        """Return current provider status for monitoring."""
+        return [
+            {
+                "provider": p.name,
+                "healthy": p.healthy,
+                "configured": p.is_configured,
+                "latency_ms": round(p.avg_latency_ms, 1),
+                "cost_per_1k": p.cost_per_1k_tokens,
+                "requests": p.request_count,
+                "errors": p.error_count,
+                "error_rate": f"{p.error_rate:.1%}",
+                "score": round(p.score(self.strategy), 3)
+                if p.healthy and p.is_configured
+                else "N/A",
+            }
+            for p in self.providers
+        ]

vendor/openclaude/python/tests/__init__.py

@@ -0,0 +1 @@
+# Pytest package marker for the Python helper test suite.

vendor/openclaude/python/tests/conftest.py

@@ -0,0 +1,5 @@
+from pathlib import Path
+import sys
+
+# Make the sibling `python/` helper modules importable from this test package.
+sys.path.insert(0, str(Path(__file__).resolve().parents[1]))

vendor/openclaude/python/tests/test_atomic_chat_provider.py

@@ -0,0 +1,130 @@
+"""
+test_atomic_chat_provider.py
+Run: pytest python/tests/test_atomic_chat_provider.py -v
+"""
+
+import pytest
+from unittest.mock import AsyncMock, MagicMock, patch
+from atomic_chat_provider import (
+    atomic_chat,
+    list_atomic_chat_models,
+    check_atomic_chat_running,
+)
+
+
+@pytest.mark.asyncio
+async def test_atomic_chat_running_true():
+    mock_response = MagicMock()
+    mock_response.status_code = 200
+    with patch("atomic_chat_provider.httpx.AsyncClient") as MockClient:
+        MockClient.return_value.__aenter__.return_value.get = AsyncMock(return_value=mock_response)
+        result = await check_atomic_chat_running()
+    assert result is True
+
+
+@pytest.mark.asyncio
+async def test_atomic_chat_running_false_on_exception():
+    with patch("atomic_chat_provider.httpx.AsyncClient") as MockClient:
+        MockClient.return_value.__aenter__.return_value.get = AsyncMock(side_effect=Exception("refused"))
+        result = await check_atomic_chat_running()
+    assert result is False
+
+
+@pytest.mark.asyncio
+async def test_list_models_returns_ids():
+    mock_response = MagicMock()
+    mock_response.status_code = 200
+    mock_response.json.return_value = {
+        "data": [{"id": "llama-3.1-8b"}, {"id": "mistral-7b"}],
+    }
+    mock_response.raise_for_status = MagicMock()
+    with patch("atomic_chat_provider.httpx.AsyncClient") as MockClient:
+        MockClient.return_value.__aenter__.return_value.get = AsyncMock(return_value=mock_response)
+        models = await list_atomic_chat_models()
+    assert "llama-3.1-8b" in models
+    assert "mistral-7b" in models
+
+
+@pytest.mark.asyncio
+async def test_list_models_empty_on_failure():
+    with patch("atomic_chat_provider.httpx.AsyncClient") as MockClient:
+        MockClient.return_value.__aenter__.return_value.get = AsyncMock(side_effect=Exception("down"))
+        models = await list_atomic_chat_models()
+    assert models == []
+
+
+@pytest.mark.asyncio
+async def test_atomic_chat_returns_anthropic_format():
+    mock_response = MagicMock()
+    mock_response.raise_for_status = MagicMock()
+    mock_response.json.return_value = {
+        "id": "chatcmpl-abc123",
+        "choices": [{"message": {"content": "42 is the answer."}}],
+        "usage": {"prompt_tokens": 10, "completion_tokens": 8},
+    }
+    with patch("atomic_chat_provider.httpx.AsyncClient") as MockClient:
+        MockClient.return_value.__aenter__.return_value.post = AsyncMock(return_value=mock_response)
+        result = await atomic_chat(
+            model="llama-3.1-8b",
+            messages=[{"role": "user", "content": "What is 6*7?"}],
+        )
+    assert result["type"] == "message"
+    assert result["role"] == "assistant"
+    assert "42" in result["content"][0]["text"]
+    assert result["usage"]["input_tokens"] == 10
+    assert result["usage"]["output_tokens"] == 8
+
+
+@pytest.mark.asyncio
+async def test_atomic_chat_prepends_system():
+    captured = {}
+
+    async def mock_post(url, json=None, **kwargs):
+        captured.update(json or {})
+        m = MagicMock()
+        m.raise_for_status = MagicMock()
+        m.json.return_value = {
+            "id": "chatcmpl-xyz",
+            "choices": [{"message": {"content": "ok"}}],
+            "usage": {"prompt_tokens": 1, "completion_tokens": 1},
+        }
+        return m
+
+    with patch("atomic_chat_provider.httpx.AsyncClient") as MockClient:
+        MockClient.return_value.__aenter__.return_value.post = mock_post
+        await atomic_chat(
+            model="llama-3.1-8b",
+            messages=[{"role": "user", "content": "Hi"}],
+            system="Be helpful.",
+        )
+    assert captured["messages"][0]["role"] == "system"
+    assert "helpful" in captured["messages"][0]["content"]
+
+
+@pytest.mark.asyncio
+async def test_atomic_chat_sends_correct_payload():
+    captured = {}
+
+    async def mock_post(url, json=None, **kwargs):
+        captured.update(json or {})
+        m = MagicMock()
+        m.raise_for_status = MagicMock()
+        m.json.return_value = {
+            "id": "chatcmpl-xyz",
+            "choices": [{"message": {"content": "ok"}}],
+            "usage": {"prompt_tokens": 1, "completion_tokens": 1},
+        }
+        return m
+
+    with patch("atomic_chat_provider.httpx.AsyncClient") as MockClient:
+        MockClient.return_value.__aenter__.return_value.post = mock_post
+        await atomic_chat(
+            model="test-model",
+            messages=[{"role": "user", "content": "Test"}],
+            max_tokens=2048,
+            temperature=0.5,
+        )
+    assert captured["model"] == "test-model"
+    assert captured["max_tokens"] == 2048
+    assert captured["temperature"] == 0.5
+    assert captured["stream"] is False

vendor/openclaude/python/tests/test_ollama_provider.py

@@ -0,0 +1,192 @@
+"""
+test_ollama_provider.py
+Run: pytest python/tests/test_ollama_provider.py -v
+"""
+
+import pytest
+from unittest.mock import AsyncMock, MagicMock, patch
+from ollama_provider import (
+    normalize_ollama_model,
+    anthropic_to_ollama_messages,
+    ollama_chat,
+    list_ollama_models,
+    check_ollama_running,
+)
+
+
+def test_normalize_strips_prefix():
+    assert normalize_ollama_model("ollama/llama3:8b") == "llama3:8b"
+
+
+def test_normalize_no_prefix():
+    assert normalize_ollama_model("codellama:34b") == "codellama:34b"
+
+
+def test_normalize_empty():
+    assert normalize_ollama_model("") == ""
+
+
+def test_converts_string_content():
+    messages = [{"role": "user", "content": "Hello!"}]
+    result = anthropic_to_ollama_messages(messages)
+    assert result == [{"role": "user", "content": "Hello!"}]
+
+
+def test_converts_text_block_list():
+    messages = [{"role": "user", "content": [{"type": "text", "text": "What is Python?"}]}]
+    result = anthropic_to_ollama_messages(messages)
+    assert result[0]["content"] == "What is Python?"
+
+
+def test_converts_image_block_to_placeholder():
+    messages = [{"role": "user", "content": [{"type": "image", "source": {}}, {"type": "text", "text": "Describe this"}]}]
+    result = anthropic_to_ollama_messages(messages)
+    assert "[image]" in result[0]["content"]
+    assert "Describe this" in result[0]["content"]
+
+
+def test_converts_base64_image_block_to_ollama_images():
+    messages = [{
+        "role": "user",
+        "content": [
+            {
+                "type": "image",
+                "source": {
+                    "type": "base64",
+                    "media_type": "image/png",
+                    "data": "YWJjMTIz",
+                },
+            },
+            {"type": "text", "text": "Describe this"},
+        ],
+    }]
+    result = anthropic_to_ollama_messages(messages)
+    assert result[0]["images"] == ["YWJjMTIz"]
+    assert "Describe this" in result[0]["content"]
+
+def test_converts_multi_turn():
+    messages = [
+        {"role": "user", "content": "Hi"},
+        {"role": "assistant", "content": "Hello!"},
+        {"role": "user", "content": "How are you?"},
+    ]
+    result = anthropic_to_ollama_messages(messages)
+    assert len(result) == 3
+    assert result[1]["role"] == "assistant"
+
+
+@pytest.mark.asyncio
+async def test_ollama_running_true():
+    mock_response = MagicMock()
+    mock_response.status_code = 200
+    with patch("ollama_provider.httpx.AsyncClient") as MockClient:
+        MockClient.return_value.__aenter__.return_value.get = AsyncMock(return_value=mock_response)
+        result = await check_ollama_running()
+    assert result is True
+
+
+@pytest.mark.asyncio
+async def test_ollama_running_false_on_exception():
+    with patch("ollama_provider.httpx.AsyncClient") as MockClient:
+        MockClient.return_value.__aenter__.return_value.get = AsyncMock(side_effect=Exception("refused"))
+        result = await check_ollama_running()
+    assert result is False
+
+
+@pytest.mark.asyncio
+async def test_list_models_returns_names():
+    mock_response = MagicMock()
+    mock_response.status_code = 200
+    mock_response.json.return_value = {"models": [{"name": "llama3:8b"}, {"name": "codellama:34b"}]}
+    mock_response.raise_for_status = MagicMock()
+    with patch("ollama_provider.httpx.AsyncClient") as MockClient:
+        MockClient.return_value.__aenter__.return_value.get = AsyncMock(return_value=mock_response)
+        models = await list_ollama_models()
+    assert "llama3:8b" in models
+
+
+@pytest.mark.asyncio
+async def test_ollama_chat_returns_anthropic_format():
+    mock_response = MagicMock()
+    mock_response.raise_for_status = MagicMock()
+    mock_response.json.return_value = {
+        "message": {"content": "42 is the answer."},
+        "created_at": "2026-01-01T00:00:00Z",
+        "prompt_eval_count": 10,
+        "eval_count": 8,
+    }
+    with patch("ollama_provider.httpx.AsyncClient") as MockClient:
+        MockClient.return_value.__aenter__.return_value.post = AsyncMock(return_value=mock_response)
+        result = await ollama_chat(
+            model="llama3:8b",
+            messages=[{"role": "user", "content": "What is 6*7?"}]
+        )
+    assert result["type"] == "message"
+    assert result["role"] == "assistant"
+    assert "42" in result["content"][0]["text"]
+
+
+@pytest.mark.asyncio
+async def test_ollama_chat_prepends_system():
+    captured = {}
+
+    async def mock_post(url, json=None, **kwargs):
+        captured.update(json or {})
+        m = MagicMock()
+        m.raise_for_status = MagicMock()
+        m.json.return_value = {
+            "message": {"content": "ok"},
+            "created_at": "",
+            "prompt_eval_count": 1,
+            "eval_count": 1
+        }
+        return m
+    with patch("ollama_provider.httpx.AsyncClient") as MockClient:
+        MockClient.return_value.__aenter__.return_value.post = mock_post
+        await ollama_chat(
+            model="llama3:8b",
+            messages=[{"role": "user", "content": "Hi"}],
+            system="Be helpful.",
+        )
+    assert captured["messages"][0]["role"] == "system"
+    assert "helpful" in captured["messages"][0]["content"]
+
+
+@pytest.mark.asyncio
+async def test_ollama_chat_includes_base64_images_in_payload():
+    captured = {}
+
+    async def mock_post(url, json=None, **kwargs):
+        captured.update(json or {})
+        m = MagicMock()
+        m.raise_for_status = MagicMock()
+        m.json.return_value = {
+            "message": {"content": "ok"},
+            "created_at": "",
+            "prompt_eval_count": 1,
+            "eval_count": 1,
+        }
+        return m
+
+    with patch("ollama_provider.httpx.AsyncClient") as MockClient:
+        MockClient.return_value.__aenter__.return_value.post = mock_post
+        await ollama_chat(
+            model="llama3:8b",
+            messages=[{
+                "role": "user",
+                "content": [
+                    {
+                        "type": "image",
+                        "source": {
+                            "type": "base64",
+                            "media_type": "image/jpeg",
+                            "data": "ZHVtbXk=",
+                        },
+                    },
+                    {"type": "text", "text": "What is in this image?"},
+                ],
+            }],
+        )
+
+    assert captured["messages"][0]["images"] == ["ZHVtbXk="]
+    assert "What is in this image?" in captured["messages"][0]["content"]

vendor/openclaude/python/tests/test_smart_router.py

@@ -0,0 +1,231 @@
+"""
+test_smart_router.py
+--------------------
+Tests for the SmartRouter.
+Run: pytest python/tests/test_smart_router.py -v
+"""
+
+import pytest
+import asyncio
+from unittest.mock import AsyncMock, MagicMock, patch
+from smart_router import SmartRouter, Provider
+
+
+# ── Fixtures ──────────────────────────────────────────────────────────────────
+
+
+@pytest.fixture(autouse=True)
+def fake_api_key(monkeypatch):
+    monkeypatch.setenv("FAKE_KEY", "test-key")
+
+
+def make_provider(name, healthy=True, configured=True,
+                  latency=100.0, cost=0.002, errors=0, requests=0):
+    p = Provider(
+        name=name,
+        ping_url=f"https://{name}.example.com/health",
+        api_key_env="FAKE_KEY",
+        cost_per_1k_tokens=cost,
+        big_model=f"{name}-big",
+        small_model=f"{name}-small",
+    )
+    p.healthy = healthy
+    p.avg_latency_ms = latency
+    p.error_count = errors
+    p.request_count = requests
+    if not configured:
+        p.api_key_env = ""  # makes is_configured False for non-local providers
+    return p
+
+
+def make_router(providers=None, strategy="balanced"):
+    r = SmartRouter(providers=providers, strategy=strategy)
+    r._initialized = True
+    return r
+
+
+# ── Provider.score() ──────────────────────────────────────────────────────────
+
+def test_score_unhealthy_is_inf():
+    p = make_provider("openai", healthy=False)
+    assert p.score() == float("inf")
+
+
+def test_score_unconfigured_is_inf():
+    p = make_provider("openai", configured=False)
+    assert p.score() == float("inf")
+
+
+def test_score_latency_strategy_prefers_faster():
+    fast = make_provider("fast", latency=50.0, cost=0.01)
+    slow = make_provider("slow", latency=500.0, cost=0.001)
+    assert fast.score("latency") < slow.score("latency")
+
+
+def test_score_cost_strategy_prefers_cheaper():
+    cheap = make_provider("cheap", latency=500.0, cost=0.0001)
+    expensive = make_provider("expensive", latency=50.0, cost=0.05)
+    assert cheap.score("cost") < expensive.score("cost")
+
+
+def test_score_balanced_strategy_uses_both():
+    p = make_provider("test", latency=200.0, cost=0.002)
+    s = p.score("balanced")
+    assert s > 0
+
+
+def test_score_error_rate_penalty():
+    clean = make_provider("clean", errors=0, requests=10)
+    dirty = make_provider("dirty", errors=8, requests=10)
+    assert clean.score() < dirty.score()
+
+
+# ── SmartRouter.is_large_request() ───────────────────────────────────────────
+
+def test_is_large_request_short():
+    r = make_router()
+    msgs = [{"role": "user", "content": "Hello!"}]
+    assert r.is_large_request(msgs) is False
+
+
+def test_is_large_request_long():
+    r = make_router()
+    msgs = [{"role": "user", "content": "x" * 3000}]
+    assert r.is_large_request(msgs) is True
+
+
+# ── SmartRouter.select_provider() ────────────────────────────────────────────
+
+def test_select_provider_picks_best_score():
+    p1 = make_provider("slow", latency=800.0)
+    p2 = make_provider("fast", latency=50.0)
+    r = make_router(providers=[p1, p2], strategy="latency")
+    selected = r.select_provider()
+    assert selected.name == "fast"
+
+
+def test_select_provider_skips_unhealthy():
+    p1 = make_provider("bad", healthy=False)
+    p2 = make_provider("good", healthy=True)
+    r = make_router(providers=[p1, p2])
+    selected = r.select_provider()
+    assert selected.name == "good"
+
+
+def test_select_provider_returns_none_when_all_down():
+    p1 = make_provider("a", healthy=False)
+    p2 = make_provider("b", healthy=False)
+    r = make_router(providers=[p1, p2])
+    assert r.select_provider() is None
+
+
+# ── SmartRouter.get_model_for_provider() ─────────────────────────────────────
+
+def test_get_model_large_request():
+    p = make_provider("openai")
+    r = make_router()
+    model = r.get_model_for_provider(p, "claude-sonnet")
+    assert model == "openai-big"
+
+
+def test_get_model_large_message_overrides_claude_label():
+    p = make_provider("openai")
+    r = make_router()
+    model = r.get_model_for_provider(p, "claude-haiku", is_large_request=True)
+    assert model == "openai-big"
+
+
+def test_get_model_small_request():
+    p = make_provider("openai")
+    r = make_router()
+    model = r.get_model_for_provider(p, "claude-haiku")
+    assert model == "openai-small"
+
+
+# ── SmartRouter.route() ───────────────────────────────────────────────────────
+
+@pytest.mark.asyncio
+async def test_route_returns_best_provider():
+    p1 = make_provider("expensive", cost=0.05, latency=50.0)
+    p2 = make_provider("cheap", cost=0.0005, latency=200.0)
+    r = make_router(providers=[p1, p2], strategy="cost")
+    result = await r.route([{"role": "user", "content": "Hi"}], "claude-haiku")
+    assert result["provider"] == "cheap"
+
+
+@pytest.mark.asyncio
+async def test_route_uses_big_model_for_large_message_bodies():
+    p = make_provider("openai")
+    r = make_router(providers=[p])
+    result = await r.route([
+        {"role": "user", "content": "x" * 3001},
+    ], "claude-haiku")
+    assert result["model"] == "openai-big"
+
+
+@pytest.mark.asyncio
+async def test_route_raises_when_no_providers():
+    p = make_provider("a", healthy=False)
+    r = make_router(providers=[p])
+    with pytest.raises(RuntimeError, match="no providers available"):
+        await r.route([{"role": "user", "content": "Hi"}])
+
+
+@pytest.mark.asyncio
+async def test_route_excludes_providers():
+    p1 = make_provider("openai", latency=50.0)
+    p2 = make_provider("gemini", latency=200.0)
+    r = make_router(providers=[p1, p2], strategy="latency")
+    result = await r.route(
+        [{"role": "user", "content": "Hi"}],
+        exclude_providers=["openai"]
+    )
+    assert result["provider"] == "gemini"
+
+
+# ── SmartRouter.record_result() ──────────────────────────────────────────────
+
+@pytest.mark.asyncio
+async def test_record_result_updates_latency():
+    p = make_provider("openai", latency=200.0)
+    r = make_router(providers=[p])
+    await r.record_result("openai", success=True, duration_ms=100.0)
+    assert p.avg_latency_ms < 200.0  # should decrease toward 100
+
+
+@pytest.mark.asyncio
+async def test_record_result_increments_requests():
+    p = make_provider("openai")
+    r = make_router(providers=[p])
+    await r.record_result("openai", success=True, duration_ms=100.0)
+    assert p.request_count == 1
+
+
+@pytest.mark.asyncio
+async def test_record_result_increments_errors():
+    p = make_provider("openai")
+    r = make_router(providers=[p])
+    await r.record_result("openai", success=False, duration_ms=0)
+    assert p.error_count == 1
+
+
+# ── SmartRouter.status() ─────────────────────────────────────────────────────
+
+def test_status_returns_all_providers():
+    p1 = make_provider("openai")
+    p2 = make_provider("gemini")
+    r = make_router(providers=[p1, p2])
+    status = r.status()
+    assert len(status) == 2
+    names = [s["provider"] for s in status]
+    assert "openai" in names
+    assert "gemini" in names
+
+
+def test_status_contains_required_fields():
+    p = make_provider("openai")
+    r = make_router(providers=[p])
+    status = r.status()[0]
+    for field in ["provider", "healthy", "latency_ms",
+                  "cost_per_1k", "requests", "errors", "score"]:
+        assert field in status

vendor/openclaude/release-please-config.json

@@ -0,0 +1,11 @@
+{
+  "$schema": "https://raw.githubusercontent.com/googleapis/release-please/main/schemas/config.json",
+  "packages": {
+    ".": {
+      "release-type": "node",
+      "package-name": "@gitlawb/openclaude",
+      "bump-minor-pre-major": true,
+      "include-v-in-tag": true
+    }
+  }
+}

vendor/openclaude/scripts/build.ts

@@ -0,0 +1,492 @@
+/**
+ * OpenClaude build script — bundles the TypeScript source into a single
+ * distributable JS file using Bun's bundler.
+ *
+ * Handles:
+ * - bun:bundle feature() flags for the open build
+ * - MACRO.* globals → inlined version/build-time constants
+ * - src/ path aliases
+ */
+
+import { readFileSync, readdirSync, writeFileSync } from 'fs'
+import { join } from 'path'
+import { noTelemetryPlugin } from './no-telemetry-plugin'
+
+const pkg = JSON.parse(readFileSync('./package.json', 'utf-8'))
+const version = pkg.version
+
+// Feature flags for the open build.
+// Most Anthropic-internal features stay off; open-build features can be
+// selectively enabled here when their full source exists in the mirror.
+const featureFlags: Record<string, boolean> = {
+  // ── Disabled: require Anthropic infrastructure or missing source ─────
+  VOICE_MODE: false,              // Push-to-talk STT via claude.ai OAuth endpoint
+  PROACTIVE: false,               // Autonomous agent mode (missing proactive/ module)
+  KAIROS: false,                  // Persistent assistant/session mode (cloud backend)
+  BRIDGE_MODE: false,             // Remote desktop bridge via CCR infrastructure
+  DAEMON: false,                  // Background daemon process (stubbed in open build)
+  AGENT_TRIGGERS: false,          // Scheduled remote agent triggers
+  ABLATION_BASELINE: false,       // A/B testing harness for eval experiments
+  CONTEXT_COLLAPSE: false,        // Context collapsing optimization (stubbed)
+  COMMIT_ATTRIBUTION: false,      // Co-Authored-By metadata in git commits
+  UDS_INBOX: false,               // Unix Domain Socket inter-session messaging
+  BG_SESSIONS: false,             // Background sessions via tmux (stubbed)
+  WEB_BROWSER_TOOL: false,        // Built-in browser automation (source not mirrored)
+  CHICAGO_MCP: false,             // Computer-use MCP (native Swift modules stubbed)
+  COWORKER_TYPE_TELEMETRY: false, // Telemetry for agent/coworker type classification
+
+  // ── Enabled: upstream defaults ──────────────────────────────────────
+  COORDINATOR_MODE: true,             // Multi-agent coordinator with worker delegation
+  BUILTIN_EXPLORE_PLAN_AGENTS: true,  // Built-in Explore/Plan specialized subagents
+  BUDDY: true,                        // Buddy mode for paired programming
+  MONITOR_TOOL: true,                 // MCP server monitoring/streaming tool
+  TEAMMEM: true,                      // Team memory management
+  MESSAGE_ACTIONS: true,              // Message action buttons in the UI
+
+  // ── Enabled: new activations ────────────────────────────────────────
+  DUMP_SYSTEM_PROMPT: true,           // --dump-system-prompt CLI flag for debugging
+  CACHED_MICROCOMPACT: true,          // Cache-aware tool result truncation optimization
+  AWAY_SUMMARY: true,                 // "While you were away" recap after 5min blur
+  TRANSCRIPT_CLASSIFIER: true,        // Auto-approval classifier for safe tool uses
+  ULTRATHINK: true,                   // Deep thinking mode — type "ultrathink" to boost reasoning
+  TOKEN_BUDGET: true,                 // Token budget tracking with usage warnings
+  HISTORY_PICKER: true,               // Enhanced interactive prompt history picker
+  QUICK_SEARCH: true,                 // Ctrl+G quick search across prompts
+  SHOT_STATS: true,                   // Shot distribution stats in session summary
+  EXTRACT_MEMORIES: true,             // Auto-extract durable memories from conversations
+  FORK_SUBAGENT: true,                // Implicit context-forking when omitting subagent_type
+  VERIFICATION_AGENT: true,           // Built-in read-only agent for test/verification
+  MCP_SKILLS: true,                   // Discover skills dynamically from MCP server resources
+  PROMPT_CACHE_BREAK_DETECTION: true, // Detect & log unexpected prompt cache invalidations
+  HOOK_PROMPTS: true,                 // Allow tools to request interactive user prompts
+}
+
+// ── Pre-process: replace feature() calls with boolean literals ──────
+// Bun v1.3.9+ resolves `import { feature } from 'bun:bundle'` natively
+// before plugins can intercept it via onResolve. The bun: namespace is
+// handled by Bun's C++ resolver which runs before the JS plugin phase,
+// so the previous onResolve/onLoad shim was silently ineffective — ALL
+// feature() calls evaluated to false regardless of the featureFlags map.
+//
+// Fix: pre-process source files to strip the bun:bundle import and
+// replace feature('FLAG') calls with their boolean literal. Files are
+// modified in-place before Bun.build() and restored in a finally block.
+
+// Match feature('FLAG') calls, including multi-line: feature(\n  'FLAG',\n)
+const featureCallRe = /\bfeature\(\s*['"](\w+)['"][,\s]*\)/gs
+const featureImportRe = /import\s*\{[^}]*\bfeature\b[^}]*\}\s*from\s*['"]bun:bundle['"];?\s*\n?/g
+const modifiedFiles = new Map<string, string>() // path → original content
+
+function preProcessFeatureFlags(dir: string) {
+  for (const ent of readdirSync(dir, { withFileTypes: true })) {
+    const full = join(dir, ent.name)
+    if (ent.isDirectory()) { preProcessFeatureFlags(full); continue }
+    if (!/\.(ts|tsx)$/.test(ent.name)) continue
+
+    const raw = readFileSync(full, 'utf-8')
+    if (!raw.includes('feature(')) continue
+
+    let contents = raw
+    contents = contents.replace(featureImportRe, '')
+    contents = contents.replace(featureCallRe, (_match, name) =>
+      String((featureFlags as Record<string, boolean>)[name] ?? false),
+    )
+
+    if (contents !== raw) {
+      modifiedFiles.set(full, raw)
+      writeFileSync(full, contents)
+    }
+  }
+}
+
+function restoreModifiedFiles() {
+  for (const [path, original] of modifiedFiles) {
+    writeFileSync(path, original)
+  }
+  modifiedFiles.clear()
+}
+
+preProcessFeatureFlags(join(import.meta.dir, '..', 'src'))
+const numModified = modifiedFiles.size
+
+// Restore source files on abrupt termination (Ctrl+C, kill, etc.)
+for (const signal of ['SIGINT', 'SIGTERM'] as const) {
+  process.on(signal, () => {
+    restoreModifiedFiles()
+    process.exit(signal === 'SIGINT' ? 130 : 143)
+  })
+}
+
+try {
+
+const result = await Bun.build({
+  entrypoints: ['./src/entrypoints/cli.tsx'],
+  outdir: './dist',
+  target: 'node',
+  format: 'esm',
+  splitting: false,
+  sourcemap: 'external',
+  minify: false,
+  naming: 'cli.mjs',
+  define: {
+    // MACRO.* build-time constants
+    // Keep the internal compatibility version high enough to pass
+    // first-party minimum-version guards, but expose the real package
+    // version separately in Open Claude branding.
+    'MACRO.VERSION': JSON.stringify('99.0.0'),
+    'MACRO.DISPLAY_VERSION': JSON.stringify(version),
+    'MACRO.BUILD_TIME': JSON.stringify(new Date().toISOString()),
+    'MACRO.ISSUES_EXPLAINER':
+      JSON.stringify('report the issue at https://github.com/anthropics/claude-code/issues'),
+    'MACRO.PACKAGE_URL': JSON.stringify('@gitlawb/openclaude'),
+    'MACRO.NATIVE_PACKAGE_URL': 'undefined',
+  },
+  plugins: [
+    noTelemetryPlugin,
+    {
+      name: 'bun-bundle-shim',
+      setup(build) {
+        const internalFeatureStubModules = new Map([
+          [
+            '../daemon/workerRegistry.js',
+            'export async function runDaemonWorker() { throw new Error("Daemon worker is unavailable in the open build."); }',
+          ],
+          [
+            '../daemon/main.js',
+            'export async function daemonMain() { throw new Error("Daemon mode is unavailable in the open build."); }',
+          ],
+          [
+            '../cli/bg.js',
+            `
+export async function psHandler() { throw new Error("Background sessions are unavailable in the open build."); }
+export async function logsHandler() { throw new Error("Background sessions are unavailable in the open build."); }
+export async function attachHandler() { throw new Error("Background sessions are unavailable in the open build."); }
+export async function killHandler() { throw new Error("Background sessions are unavailable in the open build."); }
+export async function handleBgFlag() { throw new Error("Background sessions are unavailable in the open build."); }
+`,
+          ],
+          [
+            '../cli/handlers/templateJobs.js',
+            'export async function templatesMain() { throw new Error("Template jobs are unavailable in the open build."); }',
+          ],
+          [
+            '../environment-runner/main.js',
+            'export async function environmentRunnerMain() { throw new Error("Environment runner is unavailable in the open build."); }',
+          ],
+          [
+            '../self-hosted-runner/main.js',
+            'export async function selfHostedRunnerMain() { throw new Error("Self-hosted runner is unavailable in the open build."); }',
+          ],
+        ] as const)
+
+        // bun:bundle feature() replacement is handled by the source
+        // pre-processing step above (see preProcessFeatureFlags).
+        // The previous onResolve/onLoad shim was ineffective in Bun
+        // v1.3.9+ because the bun: namespace is resolved natively
+        // before the JS plugin phase runs.
+
+        build.onResolve(
+          { filter: /^\.\.\/(daemon\/workerRegistry|daemon\/main|cli\/bg|cli\/handlers\/templateJobs|environment-runner\/main|self-hosted-runner\/main)\.js$/ },
+          args => {
+            if (!internalFeatureStubModules.has(args.path)) return null
+            return {
+              path: args.path,
+              namespace: 'internal-feature-stub',
+            }
+          },
+        )
+        build.onLoad(
+          { filter: /.*/, namespace: 'internal-feature-stub' },
+          args => ({
+            contents:
+              internalFeatureStubModules.get(args.path) ??
+              'export {}',
+            loader: 'js',
+          }),
+        )
+
+        // Resolve react/compiler-runtime to the standalone package
+        build.onResolve({ filter: /^react\/compiler-runtime$/ }, () => ({
+          path: 'react/compiler-runtime',
+          namespace: 'react-compiler-shim',
+        }))
+        build.onLoad(
+          { filter: /.*/, namespace: 'react-compiler-shim' },
+          () => ({
+            contents: `export function c(size) { return new Array(size).fill(Symbol.for('react.memo_cache_sentinel')); }`,
+            loader: 'js',
+          }),
+        )
+
+        // NOTE: @opentelemetry/* kept as external deps (too many named exports to stub)
+
+        // Resolve native addon and missing snapshot imports to stubs
+        for (const mod of [
+          'audio-capture-napi',
+          'audio-capture.node',
+          'image-processor-napi',
+          'modifiers-napi',
+          'url-handler-napi',
+          'color-diff-napi',
+          '@anthropic-ai/mcpb',
+          '@ant/claude-for-chrome-mcp',
+          '@anthropic-ai/sandbox-runtime',
+          'asciichart',
+          'plist',
+          'cacache',
+          'fuse',
+          'code-excerpt',
+          'stack-utils',
+        ]) {
+          build.onResolve({ filter: new RegExp(`^${mod}$`) }, () => ({
+            path: mod,
+            namespace: 'native-stub',
+          }))
+        }
+        build.onLoad(
+          { filter: /.*/, namespace: 'native-stub' },
+          () => ({
+            // Comprehensive stub that handles any named export via Proxy
+            contents: `
+const noop = () => null;
+const noopClass = class {};
+const handler = {
+  get(_, prop) {
+    if (prop === '__esModule') return true;
+    if (prop === 'default') return new Proxy({}, handler);
+    if (prop === 'ExportResultCode') return { SUCCESS: 0, FAILED: 1 };
+    if (prop === 'resourceFromAttributes') return () => ({});
+    if (prop === 'SandboxRuntimeConfigSchema') return { parse: () => ({}) };
+    return noop;
+  }
+};
+const stub = new Proxy(noop, handler);
+export default stub;
+export const __stub = true;
+// Named exports for all known imports
+export const SandboxViolationStore = null;
+export const SandboxManager = new Proxy({}, { get: () => noop });
+export const SandboxRuntimeConfigSchema = { parse: () => ({}) };
+export const BROWSER_TOOLS = [];
+export const getMcpConfigForManifest = noop;
+export const ColorDiff = null;
+export const ColorFile = null;
+export const getSyntaxTheme = noop;
+export const plot = noop;
+export const createClaudeForChromeMcpServer = noop;
+// OpenTelemetry exports
+export const ExportResultCode = { SUCCESS: 0, FAILED: 1 };
+export const resourceFromAttributes = noop;
+export const Resource = noopClass;
+export const SimpleSpanProcessor = noopClass;
+export const BatchSpanProcessor = noopClass;
+export const NodeTracerProvider = noopClass;
+export const BasicTracerProvider = noopClass;
+export const OTLPTraceExporter = noopClass;
+export const OTLPLogExporter = noopClass;
+export const OTLPMetricExporter = noopClass;
+export const PrometheusExporter = noopClass;
+export const LoggerProvider = noopClass;
+export const SimpleLogRecordProcessor = noopClass;
+export const BatchLogRecordProcessor = noopClass;
+export const MeterProvider = noopClass;
+export const PeriodicExportingMetricReader = noopClass;
+export const trace = { getTracer: () => ({ startSpan: () => ({ end: noop, setAttribute: noop, setStatus: noop, recordException: noop }) }) };
+export const context = { active: noop, with: (_, fn) => fn() };
+export const SpanStatusCode = { OK: 0, ERROR: 1, UNSET: 2 };
+export const ATTR_SERVICE_NAME = 'service.name';
+export const ATTR_SERVICE_VERSION = 'service.version';
+export const SEMRESATTRS_SERVICE_NAME = 'service.name';
+export const SEMRESATTRS_SERVICE_VERSION = 'service.version';
+export const AggregationTemporality = { CUMULATIVE: 0, DELTA: 1 };
+export const DataPointType = { HISTOGRAM: 0, SUM: 1, GAUGE: 2 };
+export const InstrumentType = { COUNTER: 0, HISTOGRAM: 1, UP_DOWN_COUNTER: 2 };
+export const PushMetricExporter = noopClass;
+export const SeverityNumber = {};
+`,
+            loader: 'js',
+          }),
+        )
+
+        // Resolve .md and .txt file imports to empty string stubs
+        build.onResolve({ filter: /\.(md|txt)$/ }, (args) => ({
+          path: args.path,
+          namespace: 'text-stub',
+        }))
+        build.onLoad(
+          { filter: /.*/, namespace: 'text-stub' },
+          () => ({
+            contents: `export default '';`,
+            loader: 'js',
+          }),
+        )
+
+        // Pre-scan: find all missing modules that need stubbing
+        // (Bun's onResolve corrupts module graph even when returning null,
+        //  so we use exact-match resolvers instead of catch-all patterns)
+        const fs = require('fs')
+        const pathMod = require('path')
+        const srcDir = pathMod.resolve(__dirname, '..', 'src')
+        const missingModules = new Set<string>()
+        const missingModuleExports = new Map<string, Set<string>>()
+
+        // Known missing external packages
+        for (const pkg of [
+          '@ant/computer-use-mcp',
+          '@ant/computer-use-mcp/sentinelApps',
+          '@ant/computer-use-mcp/types',
+          '@ant/computer-use-swift',
+          '@ant/computer-use-input',
+        ]) {
+          missingModules.add(pkg)
+        }
+
+        // Scan source to find imports that can't resolve
+        function scanForMissingImports() {
+          function checkAndRegister(specifier: string, fileDir: string, namedPart: string) {
+                const names = namedPart.split(',')
+                  .map((s: string) => s.trim().replace(/^type\s+/, ''))
+                  .filter((s: string) => s && !s.startsWith('type '))
+
+                // Check src/tasks/ non-relative imports
+                if (specifier.startsWith('src/tasks/')) {
+                  const resolved = pathMod.resolve(__dirname, '..', specifier)
+                  const candidates = [
+                    resolved,
+                    `${resolved}.ts`, `${resolved}.tsx`,
+                    resolved.replace(/\.js$/, '.ts'), resolved.replace(/\.js$/, '.tsx'),
+                    pathMod.join(resolved, 'index.ts'), pathMod.join(resolved, 'index.tsx'),
+                  ]
+                  if (!candidates.some((c: string) => fs.existsSync(c))) {
+                    missingModules.add(specifier)
+                  }
+                }
+                // Check relative .js imports
+                else if (specifier.endsWith('.js') && (specifier.startsWith('./') || specifier.startsWith('../'))) {
+                  const resolved = pathMod.resolve(fileDir, specifier)
+                  const tsVariant = resolved.replace(/\.js$/, '.ts')
+                  const tsxVariant = resolved.replace(/\.js$/, '.tsx')
+                  if (!fs.existsSync(resolved) && !fs.existsSync(tsVariant) && !fs.existsSync(tsxVariant)) {
+                    missingModules.add(specifier)
+                  }
+                }
+
+                // Track named exports for missing modules
+                if (names.length > 0) {
+                  if (!missingModuleExports.has(specifier)) missingModuleExports.set(specifier, new Set())
+                  for (const n of names) missingModuleExports.get(specifier)!.add(n)
+                }
+          }
+
+          function walk(dir: string) {
+            for (const ent of fs.readdirSync(dir, { withFileTypes: true })) {
+              const full = pathMod.join(dir, ent.name)
+              if (ent.isDirectory()) { walk(full); continue }
+              if (!/\.(ts|tsx)$/.test(ent.name)) continue
+              const rawCode: string = fs.readFileSync(full, 'utf-8')
+              const fileDir = pathMod.dirname(full)
+
+              // Strip comments before scanning for imports/requires.
+              // The regex scanner matches require()/import() patterns
+              // inside JSDoc comments, causing false-positive missing
+              // module detection that breaks the build with noop stubs.
+              const code = rawCode
+                .replace(/\/\*[\s\S]*?\*\//g, '')  // block comments
+                .replace(/\/\/.*$/gm, '')           // line comments
+
+              // Collect static imports: import { X } from '...'
+              for (const m of code.matchAll(/import\s+(?:\{([^}]*)\}|(\w+))?\s*(?:,\s*\{([^}]*)\})?\s*from\s+['"](.*?)['"]/g)) {
+                checkAndRegister(m[4], fileDir, m[1] || m[3] || '')
+              }
+
+              // Collect dynamic requires: require('...') — these are used
+              // behind feature() gates and become live when flags are enabled.
+              for (const m of code.matchAll(/require\(\s*['"](\.\.?\/[^'"]+)['"]\s*\)/g)) {
+                checkAndRegister(m[1], fileDir, '')
+              }
+
+              // Collect dynamic imports: import('...')
+              for (const m of code.matchAll(/import\(\s*['"](\.\.?\/[^'"]+)['"]\s*\)/g)) {
+                checkAndRegister(m[1], fileDir, '')
+              }
+            }
+          }
+          walk(srcDir)
+        }
+        scanForMissingImports()
+
+        // Register exact-match resolvers for each missing module
+        for (const mod of missingModules) {
+          const escaped = mod.replace(/[.*+?^${}()|[\]\\]/g, '\\$&')
+          build.onResolve({ filter: new RegExp(`^${escaped}$`) }, () => ({
+            path: mod,
+            namespace: 'missing-module-stub',
+          }))
+        }
+
+        build.onLoad(
+          { filter: /.*/, namespace: 'missing-module-stub' },
+          (args) => {
+            const names = missingModuleExports.get(args.path) ?? new Set()
+            const exports = [...names].map(n => `export const ${n} = noop;`).join('\n')
+            return {
+              contents: `
+const noop = () => null;
+export default noop;
+${exports}
+`,
+              loader: 'js',
+            }
+          },
+        )
+      },
+    },
+  ],
+  external: [
+    // OpenTelemetry — too many named exports to stub, kept external
+    '@opentelemetry/api',
+    '@opentelemetry/api-logs',
+    '@opentelemetry/core',
+    '@opentelemetry/exporter-trace-otlp-grpc',
+    '@opentelemetry/exporter-trace-otlp-http',
+    '@opentelemetry/exporter-trace-otlp-proto',
+    '@opentelemetry/exporter-logs-otlp-http',
+    '@opentelemetry/exporter-logs-otlp-proto',
+    '@opentelemetry/exporter-logs-otlp-grpc',
+    '@opentelemetry/exporter-metrics-otlp-proto',
+    '@opentelemetry/exporter-metrics-otlp-grpc',
+    '@opentelemetry/exporter-metrics-otlp-http',
+    '@opentelemetry/exporter-prometheus',
+    '@opentelemetry/resources',
+    '@opentelemetry/sdk-trace-base',
+    '@opentelemetry/sdk-trace-node',
+    '@opentelemetry/sdk-logs',
+    '@opentelemetry/sdk-metrics',
+    '@opentelemetry/semantic-conventions',
+    // Native image processing
+    'sharp',
+    // Cloud provider SDKs
+    '@aws-sdk/client-bedrock',
+    '@aws-sdk/client-bedrock-runtime',
+    '@aws-sdk/client-sts',
+    '@aws-sdk/credential-providers',
+    '@azure/identity',
+    'google-auth-library',
+  ],
+})
+
+if (!result.success) {
+  console.error('Build failed:')
+  for (const log of result.logs) {
+    console.error(log)
+  }
+  process.exitCode = 1
+} else {
+  console.log(`✓ Built openclaude v${version} → dist/cli.mjs`)
+}
+
+} finally {
+  // Always restore source files, even if Bun.build() throws
+  restoreModifiedFiles()
+  console.log(`  🔄 feature-flags: pre-processed ${numModified} files (restored)`)
+}

vendor/openclaude/scripts/grpc-cli.ts

@@ -0,0 +1,121 @@
+import * as grpc from '@grpc/grpc-js'
+import * as protoLoader from '@grpc/proto-loader'
+import path from 'path'
+import * as readline from 'readline'
+
+const PROTO_PATH = path.resolve(import.meta.dirname, '../src/proto/openclaude.proto')
+
+const packageDefinition = protoLoader.loadSync(PROTO_PATH, {
+  keepCase: true,
+  longs: String,
+  enums: String,
+  defaults: true,
+  oneofs: true,
+})
+
+const protoDescriptor = grpc.loadPackageDefinition(packageDefinition) as any
+const openclaudeProto = protoDescriptor.openclaude.v1
+
+const rl = readline.createInterface({
+  input: process.stdin,
+  output: process.stdout
+})
+
+function askQuestion(query: string): Promise<string> {
+  return new Promise(resolve => {
+    rl.question(query, resolve)
+  })
+}
+
+async function main() {
+  const host = process.env.GRPC_HOST || 'localhost'
+  const port = process.env.GRPC_PORT || '50051'
+  const client = new openclaudeProto.AgentService(
+    `${host}:${port}`,
+    grpc.credentials.createInsecure()
+  )
+
+  let call: grpc.ClientDuplexStream<any, any> | null = null
+
+  const startStream = () => {
+    call = client.Chat()
+    let textStreamed = false
+
+    call.on('data', async (serverMessage: any) => {
+      if (serverMessage.text_chunk) {
+        process.stdout.write(serverMessage.text_chunk.text)
+        textStreamed = true
+      } else if (serverMessage.tool_start) {
+        console.log(`\n\x1b[36m[Tool Call]\x1b[0m \x1b[1m${serverMessage.tool_start.tool_name}\x1b[0m`)
+        console.log(`\x1b[90m${serverMessage.tool_start.arguments_json}\x1b[0m\n`)
+      } else if (serverMessage.tool_result) {
+        console.log(`\n\x1b[32m[Tool Result]\x1b[0m \x1b[1m${serverMessage.tool_result.tool_name}\x1b[0m`)
+        const out = serverMessage.tool_result.output
+        if (out.length > 500) {
+          console.log(`\x1b[90m${out.substring(0, 500)}...\n(Output truncated, total length: ${out.length})\x1b[0m`)
+        } else {
+          console.log(`\x1b[90m${out}\x1b[0m`)
+        }
+      } else if (serverMessage.action_required) {
+        const action = serverMessage.action_required
+        console.log(`\n\x1b[33m[Action Required]\x1b[0m`)
+        const reply = await askQuestion(`\x1b[1m${action.question}\x1b[0m (y/n) > `)
+        
+        call?.write({
+          input: {
+            prompt_id: action.prompt_id,
+            reply: reply.trim()
+          }
+        })
+      } else if (serverMessage.done) {
+        if (!textStreamed && serverMessage.done.full_text) {
+          process.stdout.write(serverMessage.done.full_text)
+        }
+        textStreamed = false
+        console.log('\n\x1b[32m[Generation Complete]\x1b[0m')
+        promptUser()
+      } else if (serverMessage.error) {
+        console.error(`\n\x1b[31m[Server Error]\x1b[0m ${serverMessage.error.message}`)
+        promptUser()
+      }
+    })
+
+    call.on('end', () => {
+      console.log('\n\x1b[90m[Stream closed by server]\x1b[0m')
+      // Don't prompt user here, let 'done' or 'error' handlers do it
+    })
+
+    call.on('error', (err: Error) => {
+      console.error('\n\x1b[31m[Stream Error]\x1b[0m', err.message)
+      promptUser()
+    })
+  }
+
+  const promptUser = async () => {
+    const message = await askQuestion('\n\x1b[35m> \x1b[0m')
+    
+    if (message.trim().toLowerCase() === '/exit' || message.trim().toLowerCase() === '/quit') {
+      console.log('Bye!')
+      rl.close()
+      process.exit(0)
+    }
+
+    if (!call || call.destroyed) {
+      startStream()
+    }
+
+    call!.write({
+      request: {
+        session_id: 'cli-session-1',
+        message: message,
+        working_directory: process.cwd()
+      }
+    })
+  }
+
+  console.log('\x1b[32mOpenClaude gRPC CLI\x1b[0m')
+  console.log('\x1b[90mType /exit to quit.\x1b[0m')
+  promptUser()
+}
+
+main()

vendor/openclaude/scripts/no-telemetry-growthbook-stub.test.ts

@@ -0,0 +1,163 @@
+import { afterAll, beforeEach, describe, expect, test } from 'bun:test'
+import { mkdirSync, readFileSync, rmSync, unlinkSync, writeFileSync } from 'node:fs'
+import { join } from 'node:path'
+import { tmpdir } from 'node:os'
+
+// ---------------------------------------------------------------------------
+// Setup: extract the growthbook stub from no-telemetry-plugin.ts, write it to
+// a temp .mjs file, and dynamically import it so we can test the real code
+// that gets bundled.
+// ---------------------------------------------------------------------------
+
+const pluginSource = readFileSync(join(__dirname, 'no-telemetry-plugin.ts'), 'utf-8')
+const stubMatch = pluginSource.match(/'services\/analytics\/growthbook': `([\s\S]*?)`/)
+if (!stubMatch) throw new Error('Could not extract growthbook stub from no-telemetry-plugin.ts')
+
+const testDir = join(tmpdir(), `growthbook-stub-test-${process.pid}`)
+const stubFile = join(testDir, 'growthbook-stub.mjs')
+const flagsFile = join(testDir, 'test-flags.json')
+
+mkdirSync(testDir, { recursive: true })
+writeFileSync(stubFile, stubMatch[1])
+
+// Point the stub at our test flags file (checked by _loadFlags on first access)
+process.env.CLAUDE_FEATURE_FLAGS_FILE = flagsFile
+
+const stub = await import(stubFile)
+
+// ---------------------------------------------------------------------------
+// Tests
+// ---------------------------------------------------------------------------
+
+describe('growthbook stub — local feature flag overrides', () => {
+  beforeEach(() => {
+    stub.resetGrowthBook()
+    try { unlinkSync(flagsFile) } catch { /* may not exist */ }
+  })
+
+  afterAll(() => {
+    rmSync(testDir, { recursive: true, force: true })
+    delete process.env.CLAUDE_FEATURE_FLAGS_FILE
+  })
+
+  // ── File absent ──────────────────────────────────────────────────
+
+  test('returns defaultValue when flags file is absent', () => {
+    expect(stub.getFeatureValue_CACHED_MAY_BE_STALE('tengu_foo', 42)).toBe(42)
+  })
+
+  test('getAllGrowthBookFeatures returns {} when file is absent', () => {
+    expect(stub.getAllGrowthBookFeatures()).toEqual({})
+  })
+
+  // ── Open-build defaults (_openBuildDefaults) ────────────────────
+
+  test('returns open-build default when flags file is absent', () => {
+    // tengu_passport_quail is in _openBuildDefaults as true; without a
+    // flags file the stub should return the open-build override, not
+    // the call-site defaultValue.
+    expect(stub.getFeatureValue_CACHED_MAY_BE_STALE('tengu_passport_quail', false)).toBe(true)
+    expect(stub.getFeatureValue_CACHED_MAY_BE_STALE('tengu_coral_fern', false)).toBe(true)
+  })
+
+  test('flags file overrides open-build defaults', () => {
+    // User-provided feature-flags.json takes priority over _openBuildDefaults.
+    writeFileSync(flagsFile, JSON.stringify({ tengu_passport_quail: false }))
+
+    expect(stub.getFeatureValue_CACHED_MAY_BE_STALE('tengu_passport_quail', true)).toBe(false)
+  })
+
+  // ── Valid JSON object ────────────────────────────────────────────
+
+  test('loads and returns values from a valid JSON file', () => {
+    writeFileSync(flagsFile, JSON.stringify({ tengu_foo: true, tengu_bar: 'hello' }))
+
+    expect(stub.getFeatureValue_CACHED_MAY_BE_STALE('tengu_foo', false)).toBe(true)
+    expect(stub.getFeatureValue_CACHED_MAY_BE_STALE('tengu_bar', 'default')).toBe('hello')
+  })
+
+  test('returns defaultValue for keys not present in the file', () => {
+    writeFileSync(flagsFile, JSON.stringify({ tengu_foo: true }))
+
+    expect(stub.getFeatureValue_CACHED_MAY_BE_STALE('tengu_missing', 99)).toBe(99)
+  })
+
+  test('getAllGrowthBookFeatures returns the full flags object', () => {
+    const flags = { tengu_a: true, tengu_b: false, tengu_c: 42 }
+    writeFileSync(flagsFile, JSON.stringify(flags))
+
+    expect(stub.getAllGrowthBookFeatures()).toEqual(flags)
+  })
+
+  // ── Malformed / non-object JSON ──────────────────────────────────
+
+  test('falls back to defaults on malformed JSON', () => {
+    writeFileSync(flagsFile, '{not valid json!!!')
+
+    expect(stub.getFeatureValue_CACHED_MAY_BE_STALE('tengu_foo', 'fallback')).toBe('fallback')
+  })
+
+  test('falls back to defaults when JSON is a primitive (true)', () => {
+    writeFileSync(flagsFile, 'true')
+
+    expect(stub.getFeatureValue_CACHED_MAY_BE_STALE('tengu_foo', 'fallback')).toBe('fallback')
+  })
+
+  test('falls back to defaults when JSON is an array', () => {
+    writeFileSync(flagsFile, '["a", "b"]')
+
+    expect(stub.getFeatureValue_CACHED_MAY_BE_STALE('tengu_foo', 'fallback')).toBe('fallback')
+  })
+
+  // ── Cache invalidation ───────────────────────────────────────────
+
+  test('resetGrowthBook clears cache so the file is re-read', () => {
+    writeFileSync(flagsFile, JSON.stringify({ tengu_foo: 'first' }))
+    expect(stub.getFeatureValue_CACHED_MAY_BE_STALE('tengu_foo', 'x')).toBe('first')
+
+    // Update the file — cached value is still 'first'
+    writeFileSync(flagsFile, JSON.stringify({ tengu_foo: 'second' }))
+    expect(stub.getFeatureValue_CACHED_MAY_BE_STALE('tengu_foo', 'x')).toBe('first')
+
+    // After reset, the new value is picked up
+    stub.resetGrowthBook()
+    expect(stub.getFeatureValue_CACHED_MAY_BE_STALE('tengu_foo', 'x')).toBe('second')
+  })
+
+  test('refreshGrowthBookFeatures clears cache', async () => {
+    writeFileSync(flagsFile, JSON.stringify({ tengu_foo: 'v1' }))
+    expect(stub.getFeatureValue_CACHED_MAY_BE_STALE('tengu_foo', 'x')).toBe('v1')
+
+    writeFileSync(flagsFile, JSON.stringify({ tengu_foo: 'v2' }))
+    await stub.refreshGrowthBookFeatures()
+    expect(stub.getFeatureValue_CACHED_MAY_BE_STALE('tengu_foo', 'x')).toBe('v2')
+  })
+
+  // ── Multiple getter variants ─────────────────────────────────────
+
+  test('all getter functions read from local flags', async () => {
+    writeFileSync(flagsFile, JSON.stringify({ tengu_gate: true, tengu_config: { a: 1 } }))
+
+    expect(await stub.getFeatureValue_DEPRECATED('tengu_gate', false)).toBe(true)
+    stub.resetGrowthBook()
+    expect(stub.getFeatureValue_CACHED_WITH_REFRESH('tengu_gate', false)).toBe(true)
+    stub.resetGrowthBook()
+    expect(stub.checkStatsigFeatureGate_CACHED_MAY_BE_STALE('tengu_gate')).toBe(true)
+    stub.resetGrowthBook()
+    expect(await stub.checkGate_CACHED_OR_BLOCKING('tengu_gate')).toBe(true)
+    stub.resetGrowthBook()
+    expect(await stub.getDynamicConfig_BLOCKS_ON_INIT('tengu_config', {})).toEqual({ a: 1 })
+    stub.resetGrowthBook()
+    expect(stub.getDynamicConfig_CACHED_MAY_BE_STALE('tengu_config', {})).toEqual({ a: 1 })
+  })
+
+  // ── Security gate ────────────────────────────────────────────────
+
+  test('checkSecurityRestrictionGate always returns false regardless of flags', async () => {
+    writeFileSync(flagsFile, JSON.stringify({
+      tengu_disable_bypass_permissions_mode: true,
+    }))
+
+    expect(await stub.checkSecurityRestrictionGate()).toBe(false)
+  })
+})

vendor/openclaude/scripts/no-telemetry-plugin.ts

@@ -0,0 +1,459 @@
+/**
+ * No-Telemetry Build Plugin for OpenClaude
+ *
+ * Replaces all analytics, telemetry, and phone-home modules with no-op stubs
+ * at compile time. Zero runtime cost, zero network calls to Anthropic.
+ *
+ * This file is NOT tracked upstream — merge conflicts are impossible.
+ * Only build.ts needs a one-line import + one-line array entry.
+ *
+ * Kills:
+ *   - GrowthBook remote feature flags (api.anthropic.com)
+ *   - Datadog event intake
+ *   - 1P event logging (api.anthropic.com/api/event_logging/batch)
+ *   - BigQuery metrics exporter (api.anthropic.com/api/claude_code/metrics)
+ *   - Perfetto / OpenTelemetry session tracing
+ *   - Auto-updater (storage.googleapis.com, npm registry)
+ *   - Plugin fetch telemetry
+ *   - Transcript / feedback sharing
+ */
+
+import type { BunPlugin } from 'bun'
+
+// Module path (relative to src/, without extension) → stub source
+const stubs: Record<string, string> = {
+
+	// ─── Analytics core ─────────────────────────────────────────────
+
+	'services/analytics/index': `
+export function stripProtoFields(metadata) { return metadata; }
+export function attachAnalyticsSink() {}
+export function logEvent() {}
+export async function logEventAsync() {}
+export function _resetForTesting() {}
+`,
+
+	'services/analytics/growthbook': `
+import _fs from 'node:fs';
+import _path from 'node:path';
+import _os from 'node:os';
+
+let _flags = undefined;
+
+// ── Open-build GrowthBook overrides ───────────────────────────────────
+// Override upstream defaultValue for runtime gates tied to build-time
+// features. Only keys that DIFFER from upstream belong here — the
+// catalog below is pure documentation and does NOT affect resolution.
+//
+// Priority: ~/.claude/feature-flags.json > _openBuildDefaults > defaultValue
+//
+// To override at runtime, create ~/.claude/feature-flags.json:
+//   { "tengu_some_flag": true }
+const _openBuildDefaults = {
+  'tengu_sedge_lantern': true,  // AWAY_SUMMARY — "while you were away" recap (upstream: false)
+  'tengu_hive_evidence': true,  // VERIFICATION_AGENT — read-only test/verification agent (upstream: false)
+  'tengu_passport_quail': true, // EXTRACT_MEMORIES — enable memory extraction (upstream: false)
+  'tengu_coral_fern': true,     // EXTRACT_MEMORIES — enable memory search in past context (upstream: false)
+};
+
+/* ── Known runtime feature keys (reference only) ───────────────────────
+ * This catalog does NOT participate in flag resolution. It documents
+ * the known GrowthBook keys and their upstream default values, scraped
+ * from src/ call sites. It is NOT exhaustive — new keys may be added
+ * upstream between catalog updates.
+ *
+ * Some keys have different defaults at different call sites — this is
+ * intentional upstream (the server unifies the value at runtime).
+ *
+ * To activate any of these, add them to ~/.claude/feature-flags.json
+ * or to _openBuildDefaults above.
+ *
+ * ── Reasoning & thinking ──────────────────────────────────────────────
+ *   tengu_turtle_carbon            = true       ULTRATHINK deep thinking runtime gate
+ *   tengu_thinkback                = gate       /thinkback replay command
+ *
+ * ── Agents & orchestration ────────────────────────────────────────────
+ *   tengu_amber_flint              = true       Agent swarms coordination
+ *   tengu_amber_stoat              = true       Built-in agent availability (Explore, Plan, etc.)
+ *   tengu_agent_list_attach        = true       Attach file context to agent list
+ *   tengu_auto_background_agents   = false      Auto-spawn background agents
+ *   tengu_slim_subagent_claudemd   = true       Lighter ClaudeMD for subagents
+ *   tengu_hive_evidence            = false      Verification agent / evidence tracking (4 call sites)
+ *   tengu_ultraplan_model          = model cfg  ULTRAPLAN model selection (dynamic config)
+ *
+ * ── Memory & context ──────────────────────────────────────────────────
+ *   tengu_passport_quail           = false      EXTRACT_MEMORIES main gate (isExtractModeActive)
+ *   tengu_coral_fern               = false      EXTRACT_MEMORIES search in past context
+ *   tengu_slate_thimble            = false      Memory dir paths (non-interactive sessions)
+ *   tengu_herring_clock            = true/false Team memory paths (varies by call site)
+ *   tengu_bramble_lintel           = null       Extract memories throttle (null → every turn)
+ *   tengu_sedge_lantern            = false      AWAY_SUMMARY "while you were away" recap
+ *   tengu_session_memory           = false      Session memory service
+ *   tengu_sm_config                = {}         Session memory config (dynamic)
+ *   tengu_sm_compact_config        = {}         Session memory compaction config (dynamic)
+ *   tengu_cobalt_raccoon           = false      Reactive compaction (suppress auto-compact)
+ *   tengu_pebble_leaf_prune        = false      Session storage pruning
+ *
+ * ── Kairos & cron ─────────────────────────────────────────────────────
+ *   tengu_kairos_brief             = false      Brief layout mode (KAIROS)
+ *   tengu_kairos_brief_config      = {}         Brief config (dynamic)
+ *   tengu_kairos_cron              = true       Cron scheduler enable
+ *   tengu_kairos_cron_durable      = true       Durable (disk-persistent) cron tasks
+ *   tengu_kairos_cron_config       = {}         Cron jitter config (dynamic)
+ *
+ * ── Bridge & remote (require Anthropic infra) ─────────────────────────
+ *   tengu_ccr_bridge               = false      CCR bridge connection
+ *   tengu_ccr_bridge_multi_session = gate       Multi-session spawn mode
+ *   tengu_ccr_mirror               = false      CCR session mirroring
+ *   tengu_ccr_bundle_seed_enabled  = gate       Git bundle seeding for CCR
+ *   tengu_ccr_bundle_max_bytes     = null       Bundle size limit (null → default)
+ *   tengu_bridge_repl_v2           = false      Environment-less REPL bridge v2
+ *   tengu_bridge_repl_v2_cse_shim_enabled = true CSE→Session tag retag shim
+ *   tengu_bridge_min_version       = {min:'0'}  Min CLI version for bridge (dynamic)
+ *   tengu_bridge_initial_history_cap = 200      Initial history cap for bridge
+ *   tengu_bridge_system_init       = false      Bridge system initialization
+ *   tengu_cobalt_harbor            = false      Auto-connect CCR at startup
+ *   tengu_cobalt_lantern           = false      Remote setup preconditions
+ *   tengu_remote_backend           = false      Remote TUI backend
+ *   tengu_surreal_dali             = false      Remote agent tasks / triggers
+ *
+ * ── Prompt & API ──────────────────────────────────────────────────────
+ *   tengu_attribution_header       = true       Attribution header in API requests
+ *   tengu_basalt_3kr               = true       MCP instructions delta
+ *   tengu_slate_prism              = true/false Message formatting (varies by call site)
+ *   tengu_amber_prism              = false      Message content formatting
+ *   tengu_amber_json_tools         = false      JSON format for tool schemas
+ *   tengu_fgts                     = false      API feature gates
+ *   tengu_otk_slot_v1              = false      One-time key slots for API auth
+ *   tengu_cicada_nap_ms            = 0          Background GrowthBook refresh throttle (ms)
+ *   tengu_miraculo_the_bard        = false      Service initialization gate
+ *   tengu_immediate_model_command  = false      Immediate /model command execution
+ *   tengu_chomp_inflection         = false      Prompt suggestions after responses
+ *   tengu_tool_pear                = gate       API betas for tool use
+ *   tengu-off-switch               = {act:false} Service kill switch (dynamic; uses dash)
+ *
+ * ── Permissions & security ────────────────────────────────────────────
+ *   tengu_birch_trellis            = true       Bash auto-mode permissions config
+ *   tengu_auto_mode_config         = {}         Auto-mode configuration (dynamic, many call sites)
+ *   tengu_iron_gate_closed         = true       Permission iron gate (with refresh)
+ *   tengu_destructive_command_warning = false    Warning for destructive bash commands
+ *   tengu_disable_bypass_permissions_mode = security Security killswitch (always false in open build)
+ *
+ * ── UI & UX ───────────────────────────────────────────────────────────
+ *   tengu_willow_mode              = 'off'      REPL rendering mode
+ *   tengu_terminal_panel           = false      Terminal panel keybinding
+ *   tengu_terminal_sidebar         = false      Terminal sidebar in REPL/config
+ *   tengu_marble_sandcastle        = false      Fast mode gate
+ *   tengu_jade_anvil_4             = false      Rate limit options UI ordering
+ *   tengu_collage_kaleidoscope     = true       Native clipboard image paste (macOS)
+ *   tengu_lapis_finch              = false      Plugin/hint recommendation
+ *   tengu_lodestone_enabled        = false      Deep links claude-cli:// protocol
+ *   tengu_copper_panda             = false      Skill improvement suggestions
+ *   tengu_desktop_upsell           = {}         Desktop app upsell config (dynamic)
+ *   tengu-top-of-feed-tip          = {}         Emergency tip of feed (dynamic; uses dash)
+ *
+ * ── File operations ───────────────────────────────────────────────────
+ *   tengu_quartz_lantern           = false      File read/write dedup optimization
+ *   tengu_moth_copse               = false      Attachments handling (variant A)
+ *   tengu_marble_fox               = false      Attachments handling (variant B)
+ *   tengu_scratch                  = gate       Scratchpad filesystem access / coordinator
+ *
+ * ── MCP & plugins ─────────────────────────────────────────────────────
+ *   tengu_harbor                   = false      MCP channel allowlist verification
+ *   tengu_harbor_permissions       = false      MCP channel permissions enforcement
+ *   tengu_copper_bridge            = false      Chrome MCP bridge
+ *   tengu_chrome_auto_enable       = false      Auto-enable Chrome MCP on startup
+ *   tengu_glacier_2xr              = false      Enhanced tool search / ToolSearchTool
+ *   tengu_malort_pedway            = {}         Computer-use (Chicago) config (dynamic)
+ *
+ * ── VSCode / IDE ──────────────────────────────────────────────────────
+ *   tengu_quiet_fern               = false      VSCode browser support
+ *   tengu_vscode_cc_auth           = false      VSCode in-band OAuth via claude_authenticate
+ *   tengu_vscode_review_upsell     = gate       VSCode review upsell
+ *   tengu_vscode_onboarding        = gate       VSCode onboarding experience
+ *
+ * ── Voice ─────────────────────────────────────────────────────────────
+ *   tengu_amber_quartz_disabled    = false      VOICE_MODE kill-switch (false = voice allowed)
+ *
+ * ── Auto-updater (stubbed in open build) ──────────────────────────────
+ *   tengu_version_config           = {min:'0'}  Min version enforcement (dynamic)
+ *   tengu_max_version_config       = {}         Max version / deprecation config (dynamic)
+ *
+ * ── Telemetry & tracing ───────────────────────────────────────────────
+ *   tengu_trace_lantern            = false      Beta session tracing
+ *   tengu_chair_sermon             = gate       Analytics / message formatting gate
+ *   tengu_strap_foyer              = false      Settings sync to cloud
+ */
+
+function _loadFlags() {
+  if (_flags !== undefined) return;
+  try {
+    const flagsPath = process.env.CLAUDE_FEATURE_FLAGS_FILE
+      || _path.join(_os.homedir(), '.claude', 'feature-flags.json');
+    const parsed = JSON.parse(_fs.readFileSync(flagsPath, 'utf-8'));
+    _flags = (parsed && typeof parsed === 'object' && !Array.isArray(parsed)) ? parsed : null;
+  } catch {
+    _flags = null;
+  }
+}
+
+function _getFlagValue(key, defaultValue) {
+  _loadFlags();
+  if (_flags != null && Object.hasOwn(_flags, key)) return _flags[key];
+  if (Object.hasOwn(_openBuildDefaults, key)) return _openBuildDefaults[key];
+  return defaultValue;
+}
+
+const noop = () => {};
+export function onGrowthBookRefresh() { return noop; }
+export function hasGrowthBookEnvOverride() { return false; }
+export function getAllGrowthBookFeatures() { _loadFlags(); return _flags || {}; }
+export function getGrowthBookConfigOverrides() { return {}; }
+export function setGrowthBookConfigOverride() {}
+export function clearGrowthBookConfigOverrides() {}
+export function getApiBaseUrlHost() { return undefined; }
+export const initializeGrowthBook = async () => null;
+export async function getFeatureValue_DEPRECATED(feature, defaultValue) { return _getFlagValue(feature, defaultValue); }
+export function getFeatureValue_CACHED_MAY_BE_STALE(feature, defaultValue) { return _getFlagValue(feature, defaultValue); }
+export function getFeatureValue_CACHED_WITH_REFRESH(feature, defaultValue) { return _getFlagValue(feature, defaultValue); }
+export function checkStatsigFeatureGate_CACHED_MAY_BE_STALE(gate) { return Boolean(_getFlagValue(gate, false)); }
+// Security killswitch — always false in the open build. Anthropic uses this
+// gate to remotely disable bypassPermissions mode; exposing it via local flags
+// would let users accidentally lock themselves out of --dangerously-skip-permissions.
+export async function checkSecurityRestrictionGate(gate) { return false; }
+export async function checkGate_CACHED_OR_BLOCKING(gate) { return Boolean(_getFlagValue(gate, false)); }
+export function refreshGrowthBookAfterAuthChange() {}
+export function resetGrowthBook() { _flags = undefined; }
+export async function refreshGrowthBookFeatures() { _flags = undefined; }
+export function setupPeriodicGrowthBookRefresh() {}
+export function stopPeriodicGrowthBookRefresh() {}
+export async function getDynamicConfig_BLOCKS_ON_INIT(configName, defaultValue) { return _getFlagValue(configName, defaultValue); }
+export function getDynamicConfig_CACHED_MAY_BE_STALE(configName, defaultValue) { return _getFlagValue(configName, defaultValue); }
+`,
+
+	'services/analytics/sink': `
+export function initializeAnalyticsGates() {}
+export function initializeAnalyticsSink() {}
+`,
+
+	'services/analytics/config': `
+export function isAnalyticsDisabled() { return true; }
+export function isFeedbackSurveyDisabled() { return true; }
+`,
+
+	'services/analytics/datadog': `
+export const initializeDatadog = async () => false;
+export async function shutdownDatadog() {}
+export async function trackDatadogEvent() {}
+`,
+
+	'services/analytics/firstPartyEventLogger': `
+export function getEventSamplingConfig() { return {}; }
+export function shouldSampleEvent() { return null; }
+export async function shutdown1PEventLogging() {}
+export function is1PEventLoggingEnabled() { return false; }
+export function logEventTo1P() {}
+export function logGrowthBookExperimentTo1P() {}
+export function initialize1PEventLogging() {}
+export async function reinitialize1PEventLoggingIfConfigChanged() {}
+`,
+
+	'services/analytics/firstPartyEventLoggingExporter': `
+export class FirstPartyEventLoggingExporter {
+	constructor() {}
+	async export(logs, resultCallback) { resultCallback({ code: 0 }); }
+	async getQueuedEventCount() { return 0; }
+	async shutdown() {}
+	async forceFlush() {}
+}
+`,
+
+	'services/analytics/metadata': `
+export function sanitizeToolNameForAnalytics(toolName) { return toolName; }
+export function isToolDetailsLoggingEnabled() { return false; }
+export function isAnalyticsToolDetailsLoggingEnabled() { return false; }
+export function mcpToolDetailsForAnalytics() { return {}; }
+export function extractMcpToolDetails() { return undefined; }
+export function extractSkillName() { return undefined; }
+export function extractToolInputForTelemetry() { return undefined; }
+export function getFileExtensionForAnalytics() { return undefined; }
+export function getFileExtensionsFromBashCommand() { return undefined; }
+export async function getEventMetadata() { return {}; }
+export function to1PEventFormat() { return {}; }
+`,
+
+	// ─── Telemetry subsystems ───────────────────────────────────────
+
+	'utils/telemetry/bigqueryExporter': `
+export class BigQueryMetricsExporter {
+	constructor() {}
+	async export(metrics, resultCallback) { resultCallback({ code: 0 }); }
+	async shutdown() {}
+	async forceFlush() {}
+	selectAggregationTemporality() { return 0; }
+}
+`,
+
+	'utils/telemetry/perfettoTracing': `
+export function initializePerfettoTracing() {}
+export function isPerfettoTracingEnabled() { return false; }
+export function registerAgent() {}
+export function unregisterAgent() {}
+export function startLLMRequestPerfettoSpan() { return ''; }
+export function endLLMRequestPerfettoSpan() {}
+export function startToolPerfettoSpan() { return ''; }
+export function endToolPerfettoSpan() {}
+export function startUserInputPerfettoSpan() { return ''; }
+export function endUserInputPerfettoSpan() {}
+export function emitPerfettoInstant() {}
+export function emitPerfettoCounter() {}
+export function startInteractionPerfettoSpan() { return ''; }
+export function endInteractionPerfettoSpan() {}
+export function getPerfettoEvents() { return []; }
+export function resetPerfettoTracer() {}
+export async function triggerPeriodicWriteForTesting() {}
+export function evictStaleSpansForTesting() {}
+export const MAX_EVENTS_FOR_TESTING = 0;
+export function evictOldestEventsForTesting() {}
+`,
+
+	'utils/telemetry/sessionTracing': `
+const noopSpan = {
+	end() {}, setAttribute() {}, setStatus() {},
+	recordException() {}, addEvent() {}, isRecording() { return false; },
+};
+export function isBetaTracingEnabled() { return false; }
+export function isEnhancedTelemetryEnabled() { return false; }
+export function startInteractionSpan() { return noopSpan; }
+export function endInteractionSpan() {}
+export function startLLMRequestSpan() { return noopSpan; }
+export function endLLMRequestSpan() {}
+export function startToolSpan() { return noopSpan; }
+export function startToolBlockedOnUserSpan() { return noopSpan; }
+export function endToolBlockedOnUserSpan() {}
+export function startToolExecutionSpan() { return noopSpan; }
+export function endToolExecutionSpan() {}
+export function endToolSpan() {}
+export function addToolContentEvent() {}
+export function getCurrentSpan() { return null; }
+export async function executeInSpan(spanName, fn) { return fn(noopSpan); }
+export function startHookSpan() { return noopSpan; }
+export function endHookSpan() {}
+`,
+
+	// ─── Auto-updater (phones home to GCS + npm) ──────────────────
+
+	'utils/autoUpdater': `
+export async function assertMinVersion() {}
+export async function getMaxVersion() { return undefined; }
+export async function getMaxVersionMessage() { return undefined; }
+export function shouldSkipVersion() { return true; }
+export function getLockFilePath() { return '/tmp/openclaude-update.lock'; }
+export async function checkGlobalInstallPermissions() { return { hasPermissions: false, npmPrefix: null }; }
+export async function getLatestVersion() { return null; }
+export async function getNpmDistTags() { return { latest: null, stable: null }; }
+export async function getLatestVersionFromGcs() { return null; }
+export async function getGcsDistTags() { return { latest: null, stable: null }; }
+export async function getVersionHistory() { return []; }
+export async function installGlobalPackage() { return 'success'; }
+`,
+
+	// ─── Plugin fetch telemetry (not the marketplace itself) ───────
+
+	'utils/plugins/fetchTelemetry': `
+export function logPluginFetch() {}
+export function classifyFetchError() { return 'disabled'; }
+`,
+
+	// ─── Transcript / feedback sharing ─────────────────────────────
+
+	'components/FeedbackSurvey/submitTranscriptShare': `
+export async function submitTranscriptShare() { return { success: false }; }
+`,
+
+	// ─── Internal employee logging (not needed in the external build) ─────
+
+	'services/internalLogging': `
+export async function logPermissionContextForAnts() {}
+export const getContainerId = async () => null;
+`,
+
+	// ─── Deleted Anthropic-internal modules ───────────────────────────────
+
+	'services/api/dumpPrompts': `
+export function createDumpPromptsFetch() { return undefined; }
+export function getDumpPromptsPath() { return ''; }
+export function getLastApiRequests() { return []; }
+export function clearApiRequestCache() {}
+export function clearDumpState() {}
+export function clearAllDumpState() {}
+export function addApiRequestToCache() {}
+`,
+
+	'utils/undercover': `
+export function isUndercover() { return false; }
+export function getUndercoverInstructions() { return ''; }
+export function shouldShowUndercoverAutoNotice() { return false; }
+`,
+
+	'types/generated/events_mono/claude_code/v1/claude_code_internal_event': `
+export const ClaudeCodeInternalEvent = {
+  fromJSON: value => value,
+  toJSON: value => value,
+  create: value => value ?? {},
+  fromPartial: value => value ?? {},
+};
+`,
+
+	'types/generated/events_mono/growthbook/v1/growthbook_experiment_event': `
+export const GrowthbookExperimentEvent = {
+  fromJSON: value => value,
+  toJSON: value => value,
+  create: value => value ?? {},
+  fromPartial: value => value ?? {},
+};
+`,
+
+	'types/generated/events_mono/common/v1/auth': `
+export const PublicApiAuth = {
+  fromJSON: value => value,
+  toJSON: value => value,
+  create: value => value ?? {},
+  fromPartial: value => value ?? {},
+};
+`,
+
+	'types/generated/google/protobuf/timestamp': `
+export const Timestamp = {
+  fromJSON: value => value,
+  toJSON: value => value,
+  create: value => value ?? {},
+  fromPartial: value => value ?? {},
+};
+`,
+}
+
+function escapeForResolvedPathRegex(modulePath: string): string {
+	return modulePath
+		.replace(/[|\\{}()[\]^$+*?.]/g, '\\$&')
+		.replace(/\//g, '[/\\\\]')
+}
+
+export const noTelemetryPlugin: BunPlugin = {
+	name: 'no-telemetry',
+	setup(build) {
+		for (const [modulePath, contents] of Object.entries(stubs)) {
+			// Build regex that matches the resolved file path on any OS
+			// e.g. "services/analytics/growthbook" → /services[/\\]analytics[/\\]growthbook\.(ts|js)$/
+			const escaped = escapeForResolvedPathRegex(modulePath)
+			const filter = new RegExp(`${escaped}\\.(ts|js)$`)
+
+			build.onLoad({ filter }, () => ({
+				contents,
+				loader: 'js',
+			}))
+		}
+
+		console.log(`  🔇 no-telemetry: stubbed ${Object.keys(stubs).length} modules`)
+	},
+}