Enhance Dockerfile: Implement BuildKit cache for pip install to optimize dependency fetching

Refactor code structure for improved readability and maintainability

Update Dockerfile to use python3 for virtual environment creation in newer PyTorch images

Refactor code structure for improved readability and maintainability

Bump train Space base to pytorch:2.10.0-cuda12.6 (torchao>=0.16 needs torch>=2.7's register_constant)

Comprehensive fix: newer dep stack + HF-dataset cache + verified locally

Allow shipping pre-built council cache in env Space for instant UI

Council fallback to mock with annotated rationale on real-client failure

Drop Gradio for FastAPI+HTML to break websockets conflict chain

Comprehensive audit: layered cache + version-coherent train deps + fail-fast

Wire /council endpoint to real OpenAI client when API_BASE_URL is set

Merge branch 'main' of github-personal:CRIMSONHydra/stocker

Add target=_blank to external links so iframe-embedded README clicks escape the Space iframe

Merge pull request #2 from CRIMSONHydra/final-changes

Bust Docker cache for git clone via ADD of latest commit SHA URL

Pin train deps to torch-2.5-compatible versions

Bump training Space base image to pytorch 2.7.0 for torchao>=0.16 compat (torch.int1 added in 2.6+)

Fix broken README links on the Space iframe + nav cleanup

Update .gitignore to exclude Vite build artifacts and enhance App component with chart range filtering and current price indicators

Add USE_MOCK_SPECIALISTS env toggle for endpoint-independent training

Drop base_path from Space frontmatter so iframe loads / (new blog landing) by default

Render README as blog landing page at / + move demo to /web

Restore multi-stage Dockerfile: rebuild React frontend in stage 1, mount in stage 2

Rename env Space to stocker-env to avoid conflict with stocker Inference Endpoint

ci: hardcode non-secret config in workflow env (only HF_TOKEN needs to be a GitHub secret now)

Remove colab_launcher.ipynb (HF security scanner flag, cloudflared tunnel no longer needed since we moved to HF Spaces)

ci: auto-sync HF Space secrets via HfApi.add_space_secret

Revert root Dockerfile to single-stage (drop React build until lockfile/build is verified)

ci: auto-create HF Spaces if missing; sync README to current setup

Refactor notebook to use hosted HF Inference Endpoint for Gemma 4, enhancing modularity and performance

Build React frontend in Docker, drop --mock from training Space evals

ci: GitHub Actions workflow to deploy both HF Spaces on push to main

Fix training Space Dockerfile: clone repo instead of COPY (HF Spaces needs root Dockerfile)

Add HF endpoint wiring, training Space, submission-ready README + blog

Add notebook for tuning reward on task_easy with Gemma 4 E4B IT

Merge branch 'main' of github-personal:CRIMSONHydra/stocker

Add corpus functionality: implement endpoints, loader, and build script for expanded dataset

Merge pull request #1 from CRIMSONHydra/frontend

Refactor environment checks and improve data handling in API responses; update .gitignore for clarity

Refactor code structure and remove redundant changes

Enhance training and smoke test notebooks; update dependencies and configurations

Refactor Council.run to support Jupyter environments; add smoke test notebook and related files

Enhance TransformersLLMClient initialization with improved memory management and diagnostics; update Colab notebook execution counts and repository URL

Add Pyright configuration, enhance dataset building, and update training notebook

Update Colab launcher to use the correct repository URL and streamline server startup process

Enhance Colab integration with VSCode; add colab_launcher notebook and update train_grpo workflow for repo detection

Add scripts for rendering charts and serving vLLM; enhance task validation and introduce council tests

init