Feat/sunday 3 (#16)

* ui polish plan

* ui polish

* Polish Studio chat view layout and export UX.

Widen centered Chat view, add slide demo tip styling, and simplify export to Download PPTX.

Co-authored-by: Cursor <cursoragent@cursor.com>

* ui polish

* ui polish

* ui polish

---------

Co-authored-by: Cursor <cursoragent@cursor.com>

.cursor/plans/last_sprint_demo_polish_2f148411.plan.md

@@ -0,0 +1,146 @@
+---
+name: Last Sprint Demo Polish
+overview: Rename remaining "Debug" user-facing labels to "Chat", then polish Studio UI/UX for hackathon submission day — loading states, trace visibility, deploy/cache fix — without new backend features like slides-from-chat or quiz.
+todos:
+  - id: chat-labels
+    content: "Update user-facing strings: Chat panel header, Agent trace summaries, Classic tab label in index.html + app.py"
+    status: completed
+  - id: cache-bust-deploy
+    content: Add static asset version query param and redeploy HF Space so sidebar Chat is visible
+    status: completed
+  - id: ux-polish
+    content: Export button clarity, honest slide loading hints, clear debug chat on new session
+    status: completed
+  - id: chat-view-css
+    content: Polish full-page Chat view layout (centered/wider card)
+    status: completed
+  - id: submission
+    content: Run pytest, record demo video, upload trace via scripts/upload_trace.py, submit
+    status: completed
+isProject: false
+---
+
+# Last sprint: Chat rename + demo polish
+
+Submission is **today (June 15)**. You chose **demo & submission polish** over the larger slides-from-chat / quiz feature plan ([`.cursor/plans/slides_from_chat_+_quiz_f53701a3.plan.md`](.cursor/plans/slides_from_chat_+_quiz_f53701a3.plan.md)).
+
+## Current state
+
+Studio at `/` is feature-complete for the hackathon narrative: Research ingest + RAG Q&A, Slides generation with exports, Language lessons (voice + RAG), Settings, Classic UI fallback. Classic parity work is marked completed in [`.cursor/plans/studio_classic_parity_48cdd684.plan.md`](.cursor/plans/studio_classic_parity_48cdd684.plan.md).
+
+Your screenshots still show sidebar **Debug** with a bug icon, but the working tree already has **Chat** + `chat` icon in [`apps/gradio-space/static/studio/index.html`](apps/gradio-space/static/studio/index.html) line 40. Likely causes: **HF Space not redeployed** or **browser cache** on `/static/studio/*` (no cache-bust query param on CSS/JS).
+
+```mermaid
+flowchart LR
+  subgraph done [Already in repo]
+    SidebarChat["Sidebar: Chat + chat icon"]
+    APIs["research_chat, generate_slides, language_lesson_turn, debug_chat"]
+    Traces["Trace JSON in collapsed details"]
+  end
+  subgraph polish [This sprint]
+    Labels["Panel labels: Chat not debug"]
+    UX["Loading + export UX"]
+    Deploy["Push Space + cache bust"]
+    Submit["Trace upload + demo video"]
+  end
+  done --> polish
+```
+
+---
+
+## Part 1 — Debug → Chat (user-facing labels)
+
+**Scope:** HTML/CSS string changes only. Keep internal `data-view="debug"`, `debug_chat` API, and JS function names — no risky refactor.
+
+| Location | Current | Target |
+|----------|---------|--------|
+| Sidebar nav | Already `Chat` + chat icon | Verify after deploy |
+| Chat panel header | `Chat (debug)` | `Chat` |
+| Chat empty state | "test the active local model" | "Ask the local model — turn on RAG to ground answers in your library" |
+| Trace collapsibles (all views) | `Debug trace` | `Agent trace` (better for **Sharing is Caring** jury story) |
+| Classic tab | `Chat (debug)` in [`app.py`](apps/gradio-space/src/gradio_space/app.py) | `Chat` (optional `Developer` badge via existing CSS in Classic) |
+
+Files: [`index.html`](apps/gradio-space/static/studio/index.html), optionally [`app.py`](apps/gradio-space/src/gradio_space/app.py).
+
+---
+
+## Part 2 — UI/UX polish (no new APIs)
+
+### 2a. Static asset cache bust
+
+Add a build/version suffix to CSS/JS in `index.html` (e.g. `?v=20260615` or read from a single `STUDIO_ASSET_VERSION` constant in [`server.py`](apps/gradio-space/src/gradio_space/server.py) injected into the HTML route). Ensures judges see Chat label after one deploy.
+
+### 2b. Export button clarity
+
+[`studio.js`](apps/gradio-space/static/studio/studio.js) `btn-export` only opens PPTX. Options (pick one, ~15 min):
+
+- Change label to **Download PPTX** when slides exist, or
+- Small export menu: PPTX / DOCX / HTML (links already rendered in `#downloads`)
+
+### 2c. Loading expectations (slides bottleneck)
+
+Your screenshot shows **~451s** for outline on CPU. Code already caps tokens via [`outline_max_tokens`](libs/agent/src/agent/prompts.py) and shows overlay hint *"30–90 seconds"* — misleading on CPU.
+
+- Update overlay hint to: *"First run may take several minutes on CPU; use GPU Space or fewer slides for a quick demo."*
+- Add a one-line tip under **Generate slides**: *"Demo tip: 3 slides, GPU hardware."*
+
+No model changes required for submission; avoids over-promising speed.
+
+### 2d. Research chat parity with Language lessons
+
+Language lessons render `rag_references` under assistant bubbles ([`studio.js`](apps/gradio-space/static/studio/studio.js) ~407). Research chat only renders `msg.content` — citations are inline in text today, which is acceptable for demo. **Optional quick win:** if `ask_question` history includes a references field, mirror the lessons pattern (check [`research_mind.py`](apps/gradio-space/src/gradio_space/tabs/research_mind.py) return shape first).
+
+### 2e. Chat full-page view polish
+
+[`studio.css`](apps/gradio-space/static/studio/studio.css) already hides other columns on `data-view="debug"`. Minor CSS: center the chat card, slightly wider max-width on desktop so the dedicated Chat view feels intentional (not a leftover debug panel).
+
+---
+
+## Part 3 — Bug fixes (small, high-impact)
+
+| Issue | Fix |
+|-------|-----|
+| Stale "Debug" in production | Deploy + cache bust (Part 2a) |
+| Misleading slide timing hint | Part 2c |
+| New session clears research history but not debug chat | In `btn-new-session` handler, also clear `state.debugChatHistory` + `renderDebugChat()` |
+| Export disabled until slides generated | Expected; document in demo script |
+
+Run existing test suite before push: `uv run pytest` (100 tests).
+
+---
+
+## Part 4 — Submission checklist (non-code, same day)
+
+From [README hackathon checklist](README.md):
+
+1. **Space live** — push branch, confirm `/` loads Studio with Chat sidebar
+2. **Demo video** (2–3 min): Research ingest → Ask with citations → Slides (3 slides, GPU if possible) → Language lesson voice turn → expand **Agent trace** → download PPTX
+3. **Sharing is Caring** — generate one slide run, then:
+   ```bash
+   uv run python scripts/upload_trace.py --repo-id YOUR_USER/build-small-agent-traces
+   ```
+4. **Social post** + formal submission on hackathon page
+
+Badge targets already mapped in README: Best Agent, Tiny Titan, OpenBMB, Off-the-Grid, Sharing is Caring.
+
+---
+
+## Explicitly deferred (post-submission or next hackathon)
+
+- Generate slides from chat + Quiz maker ([slides/quiz plan](.cursor/plans/slides_from_chat_+_quiz_f53701a3.plan.md))
+- EchoCoach charts in Studio (Classic only today)
+- Outline speed optimizations (separate outline-only smaller model, streaming tokens)
+
+---
+
+## Estimated effort
+
+| Block | Time |
+|-------|------|
+| Chat / trace labels | ~30 min |
+| Cache bust + deploy | ~30 min |
+| Export + loading hints + new-session bug | ~45 min |
+| Chat view CSS + optional research refs | ~45 min |
+| Demo video + trace upload | ~1–2 hr |
+
+**Total:** ~3–4 hours to a polished submission.

apps/gradio-space/src/gradio_space/app.py

@@ -69,7 +69,7 @@ def build_demo() -> gr.Blocks:
                 build_echo_coach_tab()
             with gr.Tab("TeacherVoice"):
                 build_teacher_voice_tab(workspace)
-            with gr.Tab("Chat (debug)"):
+            with gr.Tab("Chat"):
                 build_chat_tab(workspace)
 
     return demo

apps/gradio-space/src/gradio_space/server.py

@@ -5,7 +5,7 @@ from pathlib import Path
 
 import gradio as gr
 from fastapi import Request
-from fastapi.responses import FileResponse, RedirectResponse
+from fastapi.responses import FileResponse, HTMLResponse, RedirectResponse
 from fastapi.staticfiles import StaticFiles
 
 from gradio import mount_gradio_app
@@ -23,6 +23,15 @@ from gradio_space.ui.theme import get_theme, load_css
 _PKG_ROOT = Path(__file__).resolve().parent
 _APP_ROOT = _PKG_ROOT.parents[1]
 _STATIC_DIR = _APP_ROOT / "static" / "studio"
+_STUDIO_ASSET_VERSION = "20260615"
+_STUDIO_INDEX_HTML = _STATIC_DIR / "index.html"
+
+
+def _studio_index_html() -> str:
+    return _STUDIO_INDEX_HTML.read_text().replace(
+        "{{STUDIO_ASSET_VERSION}}",
+        _STUDIO_ASSET_VERSION,
+    )
 
 
 def _all_allowed_paths() -> list[str]:
@@ -66,13 +75,13 @@ def create_server() -> gr.Server:
         if _wants_classic_ui(request):
             # Relative path keeps huggingface.co/spaces/.../classic on HTTPS (not http hf.space).
             return RedirectResponse(url="classic", status_code=302)
-        return FileResponse(_STATIC_DIR / "index.html")
+        return HTMLResponse(_studio_index_html())
 
     @server.get("/studio")
     async def studio_alias(request: Request):
         if _wants_classic_ui(request):
             return RedirectResponse(url="classic", status_code=302)
-        return FileResponse(_STATIC_DIR / "index.html")
+        return HTMLResponse(_studio_index_html())
 
     demo = build_demo()
     mount_gradio_app(

apps/gradio-space/static/studio/index.html

@@ -15,7 +15,7 @@
         document.documentElement.dataset.theme = "dark";
     })();
   </script>
-  <link rel="stylesheet" href="/static/studio/studio.css" />
+  <link rel="stylesheet" href="/static/studio/studio.css?v={{STUDIO_ASSET_VERSION}}" />
 </head>
 <body>
   <div id="studio-error" class="studio-banner studio-banner-error hidden" role="alert"></div>
@@ -37,7 +37,7 @@
       <button type="button" class="nav-item" data-view="research"><span class="material-symbols-outlined">search</span>Research</button>
       <button type="button" class="nav-item active" data-view="slides"><span class="material-symbols-outlined">present_to_all</span>Slides</button>
       <button type="button" class="nav-item" data-view="language-lessons"><span class="material-symbols-outlined">translate</span>Language lessons</button>
-      <button type="button" class="nav-item" data-view="debug"><span class="material-symbols-outlined">bug_report</span>Debug</button>
+      <button type="button" class="nav-item" data-view="debug"><span class="material-symbols-outlined">chat</span>Chat</button>
       <button type="button" id="btn-open-settings" class="nav-item"><span class="material-symbols-outlined">settings</span>Settings</button>
       <a href="?classic" class="nav-item nav-link"><span class="material-symbols-outlined">open_in_new</span>Classic UI</a>
     </nav>
@@ -184,7 +184,7 @@
             </button>
             <p id="research-chat-status" class="status-text"></p>
             <details class="studio-debug-trace" id="research-trace-details">
-              <summary>Debug trace</summary>
+              <summary>Agent trace</summary>
               <div id="research-trace-panel"></div>
             </details>
           </div>
@@ -268,6 +268,7 @@
               Generate Slides
             </button>
           </div>
+          <p class="status-text slide-demo-tip">Demo tip: 3 slides on GPU hardware for a quick run.</p>
           <p id="generate-status" class="status-text">Ready to generate.</p>
           <div id="progress-panel" class="progress-panel">
             <div class="progress-panel-head">
@@ -281,7 +282,7 @@
             <ol id="progress-steps" class="progress-steps"></ol>
             <div id="progress-log" class="progress-log hidden" aria-live="polite"></div>
             <details class="studio-debug-trace" id="slides-trace-details">
-              <summary>Debug trace</summary>
+              <summary>Agent trace</summary>
               <div id="slides-trace-panel"></div>
             </details>
           </div>
@@ -291,7 +292,7 @@
             <div class="region-loading-inner">
               <span class="studio-spinner" aria-hidden="true"></span>
               <p id="canvas-overlay-text" class="region-loading-text">Generating slides…</p>
-              <p class="region-loading-hint canvas-overlay-hint">Local CPU models often take 30–90 seconds.</p>
+              <p class="region-loading-hint canvas-overlay-hint">First run may take several minutes on CPU; use GPU Space or fewer slides for a quick demo.</p>
             </div>
           </div>
           <div id="slide-canvas-content" class="slide-canvas-content">
@@ -417,8 +418,8 @@
     <section class="col col-debug">
       <div class="card card-tall coach-debug-card">
         <div class="coach-card-head">
-          <h2 class="section-label">Chat (debug)</h2>
-          <p class="status-text view-debug-only">Plain chat or corpus-grounded answers — traces appear below when RAG is on.</p>
+          <h2 class="section-label">Chat</h2>
+          <p class="status-text view-debug-only">Ask the local model — turn on RAG to ground answers in your library.</p>
         </div>
         <label class="toggle-row">
           <span>Use ResearchMind RAG</span>
@@ -445,7 +446,7 @@
           </label>
         </div>
         <div id="debug-chat-messages" class="research-chat-messages debug-chat-messages">
-          <p class="research-chat-empty">Send a message to test the active local model.</p>
+          <p class="research-chat-empty">Ask the local model — turn on RAG to ground answers in your library.</p>
         </div>
         <div class="coach-debug-compose">
           <label class="field">
@@ -455,7 +456,7 @@
           <button type="button" id="btn-debug-send" class="btn btn-primary btn-block">Send</button>
         </div>
         <details class="studio-debug-trace" id="debug-trace-details">
-          <summary>Debug trace</summary>
+          <summary>Agent trace</summary>
           <div id="debug-trace-panel"></div>
         </details>
       </div>
@@ -494,6 +495,6 @@
     </aside>
   </div>
 
-  <script type="module" src="/static/studio/studio.js"></script>
+  <script type="module" src="/static/studio/studio.js?v={{STUDIO_ASSET_VERSION}}"></script>
 </body>
 </html>

apps/gradio-space/static/studio/studio.css

@@ -436,6 +436,12 @@ body.sidebar-open {
 
 .controls-actions { display: flex; justify-content: flex-end; margin-top: 0.75rem; }
 
+.slide-demo-tip {
+  margin-top: 0.5rem;
+  font-size: 0.8125rem;
+  color: var(--on-surface-variant);
+}
+
 .region-loading-host,
 .card-ingest,
 .card-chat,
@@ -1779,14 +1785,24 @@ body.sidebar-open {
 
 .workspace[data-view="debug"] {
   grid-template-columns: 1fr;
-  max-width: 640px;
+  max-width: 48rem;
   margin-left: auto;
   margin-right: auto;
+  padding-left: 1rem;
+  padding-right: 1rem;
+}
+
+.workspace[data-view="debug"] .col-debug {
+  width: 100%;
+}
+
+.workspace[data-view="debug"] .coach-debug-card {
+  min-height: calc(100vh - var(--topbar-h) - var(--context-bar-h) - 2rem);
 }
 
 .debug-chat-messages {
-  min-height: 200px;
-  max-height: 320px;
+  min-height: 280px;
+  max-height: min(480px, 50vh);
 }
 
 .lessons-rail-card .field + .field {

apps/gradio-space/static/studio/studio.js

@@ -603,7 +603,7 @@ function renderDebugChat() {
   const container = $("#debug-chat-messages");
   if (!state.debugChatHistory.length) {
     container.innerHTML =
-      '<p class="research-chat-empty">Send a message to test the active local model.</p>';
+      '<p class="research-chat-empty">Ask the local model — turn on RAG to ground answers in your library.</p>';
     return;
   }
   container.innerHTML = state.debugChatHistory
@@ -1283,6 +1283,8 @@ async function generateSlides() {
       <a href="${fileUrl(data.downloads.docx)}" download>DOCX</a>
       <a href="${fileUrl(data.downloads.html)}" download>HTML</a>`;
         $("#btn-export").disabled = false;
+        const exportBtn = $("#btn-export");
+        if (exportBtn) exportBtn.textContent = "Download PPTX";
         syncLayoutOffsets();
       }
 
@@ -1298,7 +1300,7 @@ async function generateSlides() {
     },
     {
       overlayEl: $("#canvas-overlay"),
-      hint: "Local CPU models often take 30–90 seconds.",
+      hint: "First run may take several minutes on CPU; use GPU Space or fewer slides for a quick demo.",
     }
   );
 }
@@ -1656,9 +1658,11 @@ function bindUi() {
   $("#btn-new-session").addEventListener("click", () => {
     state.workspaceSessionId = "";
     state.researchChatHistory = [];
+    state.debugChatHistory = [];
     state.discoveredUrls = [];
     state.selectedUrls = [];
     renderResearchChat();
+    renderDebugChat();
     renderResearchUrlChoices([], []);
     $("#workspace-session").value = "";
     $("#ingest-status").textContent =

results/lm_eval/ensemble-jepa__manual/summary.md

@@ -1,15 +0,0 @@
-# lm-eval summary
-
-- experiment: `ensemble-jepa__manual`
-- model backend: `ensemble-lm`
-- base model: `/home/msg-encrypted/Documents/dev/hackathon/small-model-hackathon/models/ensemble/custom-jepa-pretrain`
-- ensemble checkpoint: `/home/msg-encrypted/Documents/dev/hackathon/small-model-hackathon/models/ensemble/custom-jepa-pretrain`
-- tasks: arc_easy, hellaswag
-- num_fewshot: 0
-- limit: 25
-- seed: 42
-
-| task | metric | score |
-| --- | --- | ---: |
-| arc_easy | acc,none | 0.5200 |
-| hellaswag | acc,none | 0.2800 |