Spaces:

build-small-hackathon
/

lesson-agent

Running on Zero

App Files Files Community

MSG commited on 16 days ago

Commit

196a48f

1 Parent(s): 871f869

Feat/fix stuff and space basics (#13)

Browse files

* dark theme wip

* loading

* wip loader spinner

* wip research and rag

* benchmark wip

* clean coach page

Files changed (17) hide show

apps/gradio-space/src/gradio_space/api/studio.py +2 -1
apps/gradio-space/src/gradio_space/research_helpers.py +13 -2
apps/gradio-space/src/gradio_space/ui/studio_html.py +1 -1
apps/gradio-space/static/studio/index.html +60 -45
apps/gradio-space/static/studio/studio.css +145 -95
apps/gradio-space/static/studio/studio.js +347 -206
libs/agent/src/agent/research_prompts.py +4 -1
libs/agent/src/agent/runner.py +7 -0
libs/agent/src/agent/tools/research_tools.py +21 -1
libs/echocoach/src/echocoach/prompts.py +1 -1
libs/echocoach/src/echocoach/teacher_voice.py +29 -11
libs/echocoach/tests/test_teacher_voice.py +27 -1
libs/inference/src/inference/response_clean.py +36 -1
libs/inference/tests/test_response_clean.py +12 -0
libs/researchmind/src/researchmind/citations.py +1 -1
libs/researchmind/tests/test_citations.py +8 -0
scripts/benchmark_rag_chat.py +156 -0

apps/gradio-space/src/gradio_space/api/studio.py CHANGED Viewed

@@ -579,9 +579,9 @@ def api_teacher_voice_turn(
         assistant=result.assistant_text,
         status=result.rag_status or "Turn complete.",
         voiceout_path=result.voiceout_path,
     )
 def api_teacher_voice_audio_turn(
     audio_path: str,
     mode: TeacherVoiceMode = "lesson",
@@ -627,6 +627,7 @@ def api_teacher_voice_audio_turn(
         status=result.rag_status or "Turn complete.",
         voiceout_path=result.voiceout_path,
         user_text=result.user_text,
     )

         assistant=result.assistant_text,
         status=result.rag_status or "Turn complete.",
         voiceout_path=result.voiceout_path,
+        rag_references=result.rag_references,
     )
 def api_teacher_voice_audio_turn(
     audio_path: str,
     mode: TeacherVoiceMode = "lesson",
         status=result.rag_status or "Turn complete.",
         voiceout_path=result.voiceout_path,
         user_text=result.user_text,
+        rag_references=result.rag_references,
     )

apps/gradio-space/src/gradio_space/research_helpers.py CHANGED Viewed

@@ -126,6 +126,17 @@ def trace_summary_markdown(trace_path: str) -> str:
         "",
     ]
     for step in data.get("steps", []):
         if step.get("type") != "note":
             continue
         msg = step.get("message", "")
@@ -193,8 +204,8 @@ def format_citations_markdown(trace_json: str) -> str:
         return ""
     lines = ["", "---", "**Sources:**"]
     for i, cite in enumerate(citations[:5], start=1):
-        title = cite.get("title") or cite.get("uri") or "Source"
-        uri = cite.get("uri") or ""
         lines.append(f"{i}. [{title}]({uri})" if uri else f"{i}. {title}")
     if len(citations) > 5:
         lines.append(f"_…and {len(citations) - 5} more (see Advanced trace)._")

         "",
     ]
     for step in data.get("steps", []):
+        if step.get("type") == "step":
+            label = step.get("label") or step.get("name") or "step"
+            dur = step.get("duration_ms")
+            extra = {k: v for k, v in step.items() if k not in ("type", "name", "label", "duration_ms", "elapsed_ms")}
+            detail = ""
+            if dur is not None:
+                detail = f" ({dur} ms)"
+            if extra:
+                detail += " — " + ", ".join(f"{k}={v!r}" for k, v in extra.items())
+            lines.append(f"- **{label}**{detail}")
+            continue
         if step.get("type") != "note":
             continue
         msg = step.get("message", "")
         return ""
     lines = ["", "---", "**Sources:**"]
     for i, cite in enumerate(citations[:5], start=1):
+        title = cite.get("doc_title") or cite.get("title") or cite.get("doc_uri") or cite.get("uri") or "Source"
+        uri = cite.get("doc_uri") or cite.get("uri") or ""
         lines.append(f"{i}. [{title}]({uri})" if uri else f"{i}. {title}")
     if len(citations) > 5:
         lines.append(f"_…and {len(citations) - 5} more (see Advanced trace)._")

apps/gradio-space/src/gradio_space/ui/studio_html.py CHANGED Viewed

@@ -127,7 +127,7 @@ def render_echo_coach_panel(
     if pace_score is None and not tip and not report_md and not transcript_html:
         return """
 <div class="studio-coach-panel studio-coach-idle">
-  <p class="studio-coach-hint">Record a pitch in the Coach view, then click <strong>Analyze pitch</strong> for metrics.</p>
 </div>"""
     score = pace_score if pace_score is not None else "—"

     if pace_score is None and not tip and not report_md and not transcript_html:
         return """
 <div class="studio-coach-panel studio-coach-idle">
+  <p class="studio-coach-hint">Record a pitch below, then click <strong>Analyze pitch</strong> for metrics.</p>
 </div>"""
     score = pace_score if pace_score is not None else "—"

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

@@ -8,6 +8,13 @@
   <link rel="preconnect" href="https://fonts.gstatic.com" crossorigin />
   <link href="https://fonts.googleapis.com/css2?family=Hanken+Grotesk:ital,wght@0,100..900;1,100..900&display=swap" rel="stylesheet" />
   <link href="https://fonts.googleapis.com/css2?family=Material+Symbols+Outlined:opsz,wght,FILL,GRAD@24,400,0,0&display=swap" rel="stylesheet" />
   <link rel="stylesheet" href="/static/studio/studio.css" />
 </head>
 <body>
@@ -28,7 +35,6 @@
       <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="voice"><span class="material-symbols-outlined">mic</span>Voice</button>
-      <button type="button" class="nav-item" data-view="coach"><span class="material-symbols-outlined">school</span>Coach</button>
       <button type="button" class="nav-item" data-view="debug"><span class="material-symbols-outlined">bug_report</span>Debug</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>
@@ -47,6 +53,9 @@
       <strong id="project-title">Small Model Finetuning</strong>
     </nav>
     <div class="topbar-actions">
       <a href="/classic" class="btn btn-ghost">Classic UI</a>
       <button type="button" id="btn-export" class="btn btn-primary" disabled>Export</button>
     </div>
@@ -266,11 +275,11 @@
           </div>
         </div>
         <div id="slide-canvas" class="slide-canvas">
-          <div id="canvas-overlay" class="canvas-overlay hidden" aria-live="polite">
-            <div class="canvas-overlay-inner">
-              <span class="canvas-spinner" aria-hidden="true"></span>
-              <p id="canvas-overlay-text">Generating slides…</p>
-              <p class="canvas-overlay-hint">Local CPU models often take 30–90 seconds.</p>
             </div>
           </div>
           <div id="slide-canvas-content" class="slide-canvas-content">
@@ -296,16 +305,16 @@
           <div class="card voice-rag-card">
             <p class="card-title">RAG Scope</p>
             <label class="toggle-row">
-              <span>Cross-Reference Sources</span>
               <input id="use-rag" type="checkbox" checked />
             </label>
-            <p class="status-text">Uses workspace session and documents unless overridden below.</p>
           </div>
           <div class="card voice-rail-controls">
             <p class="card-title">Mode</p>
             <div class="mode-cards voice-mode-cards" id="voice-modes">
               <button type="button" class="mode-card" data-mode="explain">Explain</button>
-              <button type="button" class="mode-card active" data-mode="lesson">Coach</button>
               <button type="button" class="mode-card" data-mode="pitch">Practice</button>
             </div>
             <label class="field voice-topic-wrap" id="voice-topic-wrap">
@@ -371,43 +380,46 @@
               <div id="voice-audio-out" class="voice-audio-out"></div>
             </div>
           </div>
-        </div>
-      </div>
-      <div class="card coach-panel-wrap view-coach-only">
-        <div class="coach-card-head">
-          <h2 class="section-label">EchoCoach · Pitch analysis</h2>
-          <p class="coach-card-desc">Record or upload a short pitch for pace, filler highlights, and spoken feedback.</p>
-        </div>
-        <div class="coach-capture-row">
-          <div class="coach-capture-controls">
-            <div class="recording-row coach-recording-row">
-              <button type="button" id="btn-coach-record-start" class="btn btn-secondary">Start mic</button>
-              <button type="button" id="btn-coach-record-stop" class="btn btn-secondary" disabled>Stop mic</button>
-              <button type="button" id="btn-coach-sample" class="btn btn-ghost">Load sample</button>
             </div>
-            <p id="coach-record-status" class="status-text coach-record-status"></p>
-          </div>
-          <label class="field coach-upload-field">
-            <span>Upload pitch (WAV)</span>
-            <input id="coach-audio" type="file" accept="audio/*" />
-          </label>
-        </div>
-        <div class="controls-grid coach-presets">
-          <label class="field">
-            <span>Language</span>
-            <select id="coach-language" class="input"></select>
-          </label>
-          <label class="field">
-            <span>ASR preset</span>
-            <select id="coach-asr" class="input"></select>
-          </label>
         </div>
-        <label class="toggle-row coach-voiceout-toggle">
-          <span>Speak full rewrite (VoiceOut)</span>
-          <input id="coach-speak-rewrite" type="checkbox" />
-        </label>
-        <button type="button" id="btn-analyze" class="btn btn-primary btn-block coach-analyze-btn">Analyze pitch</button>
-        <div id="coach-panel" class="coach-results-panel"></div>
       </div>
     </section>
@@ -415,7 +427,6 @@
       <div class="card card-tall coach-debug-card">
         <div class="coach-card-head">
           <h2 class="section-label">Chat (debug)</h2>
-          <p class="coach-card-desc view-coach-only">Plain chat or corpus-grounded answers — traces appear below when RAG is on.</p>
           <p class="status-text view-debug-only">Plain chat or corpus-grounded answers — traces appear below when RAG is on.</p>
         </div>
         <label class="toggle-row">
@@ -467,6 +478,10 @@
         <h2 id="settings-title">Settings</h2>
         <button type="button" id="btn-close-settings" class="btn btn-ghost btn-icon material-symbols-outlined" aria-label="Close">close</button>
       </div>
       <div id="settings-model-wrap">
         <p id="settings-active-model" class="status-text"></p>
         <label class="field hidden" id="settings-model-select-wrap">

   <link rel="preconnect" href="https://fonts.gstatic.com" crossorigin />
   <link href="https://fonts.googleapis.com/css2?family=Hanken+Grotesk:ital,wght@0,100..900;1,100..900&display=swap" rel="stylesheet" />
   <link href="https://fonts.googleapis.com/css2?family=Material+Symbols+Outlined:opsz,wght,FILL,GRAD@24,400,0,0&display=swap" rel="stylesheet" />
+  <script>
+    (function () {
+      var t = localStorage.getItem("studio-theme");
+      if (t === "dark" || (!t && matchMedia("(prefers-color-scheme: dark)").matches))
+        document.documentElement.dataset.theme = "dark";
+    })();
+  </script>
   <link rel="stylesheet" href="/static/studio/studio.css" />
 </head>
 <body>
       <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="voice"><span class="material-symbols-outlined">mic</span>Voice</button>
       <button type="button" class="nav-item" data-view="debug"><span class="material-symbols-outlined">bug_report</span>Debug</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>
       <strong id="project-title">Small Model Finetuning</strong>
     </nav>
     <div class="topbar-actions">
+      <button type="button" id="theme-toggle-btn" class="btn btn-ghost btn-icon" aria-label="Toggle light/dark theme" title="Toggle theme">
+        <span class="material-symbols-outlined" id="theme-icon">dark_mode</span>
+      </button>
       <a href="/classic" class="btn btn-ghost">Classic UI</a>
       <button type="button" id="btn-export" class="btn btn-primary" disabled>Export</button>
     </div>
           </div>
         </div>
         <div id="slide-canvas" class="slide-canvas">
+          <div id="canvas-overlay" class="region-loading hidden" aria-live="polite">
+            <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>
             </div>
           </div>
           <div id="slide-canvas-content" class="slide-canvas-content">
           <div class="card voice-rag-card">
             <p class="card-title">RAG Scope</p>
             <label class="toggle-row">
+              <span>Answer from my indexed sources</span>
               <input id="use-rag" type="checkbox" checked />
             </label>
+            <p class="status-text">Ground teacher replies in your workspace documents when enabled.</p>
           </div>
           <div class="card voice-rail-controls">
             <p class="card-title">Mode</p>
             <div class="mode-cards voice-mode-cards" id="voice-modes">
               <button type="button" class="mode-card" data-mode="explain">Explain</button>
+              <button type="button" class="mode-card active" data-mode="lesson">Lesson</button>
               <button type="button" class="mode-card" data-mode="pitch">Practice</button>
             </div>
             <label class="field voice-topic-wrap" id="voice-topic-wrap">
               <div id="voice-audio-out" class="voice-audio-out"></div>
             </div>
           </div>
+          <details class="card voice-pitch-analysis hidden" id="voice-pitch-analysis" open>
+            <summary class="voice-pitch-summary">
+              <span class="section-label">Deep pitch analysis</span>
+              <span class="voice-pitch-summary-hint">Pace, fillers, charts, and spoken rewrite</span>
+            </summary>
+            <div class="coach-panel-wrap">
+              <p class="coach-card-desc">Record or upload a short monologue (up to 30s), then analyze for metrics and feedback.</p>
+              <div class="coach-capture-row">
+                <div class="coach-capture-controls">
+                  <div class="recording-row coach-recording-row">
+                    <button type="button" id="btn-coach-record-start" class="btn btn-secondary">Start mic</button>
+                    <button type="button" id="btn-coach-record-stop" class="btn btn-secondary" disabled>Stop mic</button>
+                    <button type="button" id="btn-coach-sample" class="btn btn-ghost">Load sample</button>
+                  </div>
+                  <p id="coach-record-status" class="status-text coach-record-status"></p>
+                </div>
+                <label class="field coach-upload-field">
+                  <span>Upload pitch (WAV)</span>
+                  <input id="coach-audio" type="file" accept="audio/*" />
+                </label>
+              </div>
+              <div class="controls-grid coach-presets">
+                <label class="field">
+                  <span>Language</span>
+                  <select id="coach-language" class="input"></select>
+                </label>
+                <label class="field">
+                  <span>ASR preset</span>
+                  <select id="coach-asr" class="input"></select>
+                </label>
+              </div>
+              <label class="toggle-row coach-voiceout-toggle">
+                <span>Speak full rewrite (VoiceOut)</span>
+                <input id="coach-speak-rewrite" type="checkbox" />
+              </label>
+              <button type="button" id="btn-analyze" class="btn btn-primary btn-block coach-analyze-btn">Analyze pitch</button>
+              <div id="coach-panel" class="coach-results-panel"></div>
             </div>
+          </details>
         </div>
       </div>
     </section>
       <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>
         </div>
         <label class="toggle-row">
         <h2 id="settings-title">Settings</h2>
         <button type="button" id="btn-close-settings" class="btn btn-ghost btn-icon material-symbols-outlined" aria-label="Close">close</button>
       </div>
+      <label class="toggle-row theme-toggle">
+        <span>Dark theme</span>
+        <input id="theme-toggle" type="checkbox" role="switch" aria-label="Dark theme" />
+      </label>
       <div id="settings-model-wrap">
         <p id="settings-active-model" class="status-text"></p>
         <label class="field hidden" id="settings-model-select-wrap">

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

@@ -5,6 +5,7 @@
   --on-primary: #ffffff;
   --background: #f7f9fb;
   --surface: #f7f9fb;
   --surface-container-low: #f2f4f6;
   --surface-container: #eceef0;
   --secondary: #565e74;
@@ -15,6 +16,14 @@
   --inverse-surface: #2d3133;
   --inverse-on-surface: #eff1f3;
   --error: #ba1a1a;
   --sidebar-w: 280px;
   --topbar-h: 64px;
   --context-bar-h: 72px;
@@ -23,6 +32,30 @@
   --shadow-sm: 0 1px 2px rgba(0, 0, 0, 0.06);
 }
 * { box-sizing: border-box; }
 body {
@@ -52,7 +85,7 @@ body {
 }
 .studio-banner-loading { background: var(--secondary-container); color: var(--on-secondary-container); }
-.studio-banner-error { background: #ffdad6; color: #93000a; }
 .sidebar {
   position: fixed;
@@ -160,7 +193,7 @@ body {
   left: var(--sidebar-w);
   right: 0;
   z-index: 35;
-  background: #fff;
   border-bottom: 1px solid var(--outline-variant);
   padding: 0.5rem 1.5rem;
 }
@@ -244,7 +277,7 @@ body {
 }
 .card {
-  background: #fff;
   border: 1px solid var(--outline-variant);
   border-radius: var(--radius-xl);
   padding: 1rem;
@@ -288,7 +321,7 @@ body {
 .btn:active { transform: scale(0.98); }
 .btn:disabled { opacity: 0.5; cursor: not-allowed; }
 .btn-primary { background: var(--primary); color: var(--on-primary); }
-.btn-secondary { background: #fff; border-color: var(--outline-variant); color: var(--on-surface); }
 .btn-ghost { background: transparent; border-color: var(--outline-variant); color: var(--secondary); text-decoration: none; }
 .btn-block { width: 100%; margin-top: 0.5rem; }
@@ -299,13 +332,14 @@ body {
   border-radius: 8px;
   font: inherit;
   font-size: 0.9rem;
-  background: #fff;
 }
 .input:focus {
   outline: none;
   border-color: var(--primary);
-  box-shadow: 0 0 0 2px rgba(168, 51, 0, 0.15);
 }
 .upload-zone {
@@ -350,13 +384,24 @@ body {
 .controls-actions { display: flex; justify-content: flex-end; margin-top: 0.75rem; }
 .slide-canvas {
   position: relative;
   flex: 1;
   min-height: 320px;
   border: 2px dashed var(--outline-variant);
   border-radius: var(--radius-xl);
-  background: rgba(216, 218, 220, 0.15);
   overflow: auto;
   padding: 1rem;
 }
@@ -365,39 +410,52 @@ body {
   min-height: 280px;
 }
-.canvas-overlay {
   position: absolute;
   inset: 0;
   z-index: 5;
   display: flex;
   align-items: center;
   justify-content: center;
-  background: rgba(247, 249, 251, 0.88);
-  backdrop-filter: blur(2px);
-  border-radius: var(--radius-lg);
 }
-.canvas-overlay-inner {
   text-align: center;
-  padding: 1.5rem;
 }
 .canvas-spinner,
 .lesson-running-spinner {
   display: inline-block;
-  width: 36px;
-  height: 36px;
   border: 3px solid var(--outline-variant);
   border-top-color: var(--primary);
   border-radius: 50%;
   animation: studio-spin 0.9s linear infinite;
-  margin-bottom: 0.75rem;
 }
-.canvas-overlay-hint {
-  margin: 0.35rem 0 0;
   font-size: 0.85rem;
   color: var(--secondary);
 }
 @keyframes studio-spin {
@@ -457,7 +515,7 @@ body {
 .progress-step.pending { color: var(--secondary); }
 .progress-step.active { color: var(--primary); font-weight: 600; }
-.progress-step.done { color: #2d6a4f; }
 .progress-log {
   margin: 0.75rem 0 0;
@@ -510,7 +568,7 @@ body {
   text-transform: uppercase;
 }
-.studio-badge-rag { background: var(--primary-fixed); color: #390c00; }
 .studio-badge-muted { background: var(--surface-container); color: var(--secondary); }
 .studio-doc-header { margin-bottom: 0.5rem; }
@@ -522,7 +580,7 @@ body {
   padding: 0.5rem;
   border: 1px solid var(--outline-variant);
   border-radius: 8px;
-  background: #fff;
 }
 .studio-doc-card:hover { border-color: var(--primary); }
@@ -550,10 +608,11 @@ body {
   padding: 0.6rem 0.75rem;
   border: 1px solid var(--outline-variant);
   border-radius: 8px;
-  background: #fff;
   font: inherit;
   cursor: pointer;
   text-align: left;
 }
 .mode-card.active {
@@ -870,7 +929,7 @@ body {
   border-radius: 6px;
 }
-.url-choice-item:hover { background: rgba(255, 255, 255, 0.65); }
 .url-choice-item span {
   word-break: break-all;
@@ -941,7 +1000,7 @@ body {
 .research-chat-assistant {
   align-self: flex-start;
-  background: #fff;
   border: 1px solid var(--outline-variant);
 }
@@ -961,8 +1020,7 @@ body {
 .workspace[data-view="voice"] .col-research,
 .workspace[data-view="voice"] .col-slides { display: none; }
-.workspace[data-view="voice"] .col-debug,
-.workspace[data-view="voice"] .view-coach-only { display: none; }
 .view-voice-only { display: none; }
@@ -995,6 +1053,54 @@ body {
 .workspace[data-view="voice"] .voice-main {
   min-width: 0;
 }
 .workspace[data-view="voice"] .voice-main-card {
@@ -1115,69 +1221,9 @@ body {
   }
 }
-.workspace[data-view="coach"] .col-research,
-.workspace[data-view="coach"] .col-slides { display: none; }
-.workspace[data-view="coach"] .view-voice-only { display: none; }
 .workspace[data-view="slides"] .col-studio,
 .workspace[data-view="research"] .col-debug { display: none; }
-.workspace[data-view="coach"] {
-  grid-template-columns: minmax(0, 1.05fr) minmax(0, 0.95fr);
-  max-width: 1280px;
-  gap: 1.25rem;
-  align-items: start;
-}
-.workspace[data-view="coach"] .coach-panel-wrap,
-.workspace[data-view="coach"] .coach-debug-card {
-  display: flex;
-  flex-direction: column;
-}
-.workspace[data-view="coach"] .coach-results-panel {
-  flex: 1;
-  min-height: 120px;
-  margin-top: 0.75rem;
-  overflow-y: auto;
-}
-.workspace[data-view="coach"] .coach-results-panel:not(:empty) {
-  border-top: 1px solid var(--outline-variant);
-  padding-top: 0.75rem;
-}
-.workspace[data-view="coach"] .debug-chat-messages {
-  min-height: 140px;
-  max-height: min(240px, 30vh);
-  margin-bottom: 0.5rem;
-}
-.workspace[data-view="coach"] .coach-debug-compose {
-  padding-top: 0;
-  border-top: none;
-}
-.workspace[data-view="coach"] .coach-debug-compose textarea {
-  min-height: 3.5rem;
-  resize: vertical;
-}
-.workspace[data-view="coach"] .coach-debug-card .studio-debug-trace {
-  flex-shrink: 0;
-  margin-top: 0.5rem;
-}
-.workspace[data-view="coach"] .coach-debug-card .toggle-row,
-.workspace[data-view="coach"] .coach-debug-card .debug-rag-scope {
-  flex-shrink: 0;
-}
-.view-coach-only { display: none; }
-.workspace[data-view="coach"] .view-coach-only:not(.coach-panel-wrap) { display: block; }
-.workspace[data-view="coach"] .view-debug-only { display: none; }
 .coach-card-head {
   margin-bottom: 0.85rem;
 }
@@ -1231,13 +1277,6 @@ body {
 }
 @media (max-width: 960px) {
-  .workspace[data-view="coach"] {
-    grid-template-columns: 1fr;
-    max-width: 640px;
-    margin-left: auto;
-    margin-right: auto;
-  }
   .coach-capture-row {
     grid-template-columns: 1fr;
   }
@@ -1297,13 +1336,17 @@ body {
   padding: 0.75rem;
   border: 1px solid var(--outline-variant);
   border-radius: 8px;
-  background: #fff;
   font-size: 0.875rem;
   line-height: 1.5;
   max-height: 240px;
   overflow-y: auto;
 }
 .debug-rag-scope {
   margin: 0.75rem 0;
 }
@@ -1399,6 +1442,14 @@ body {
   font-size: 0.875rem;
 }
 .voice-replay-row {
   display: flex;
   flex-wrap: wrap;
@@ -1509,7 +1560,6 @@ body {
   margin-top: 0;
 }
-.workspace[data-view="debug"] .view-coach-only { display: none; }
 .workspace[data-view="debug"] .view-debug-only { display: block; }
 .workspace[data-view="debug"] .col-research,

   --on-primary: #ffffff;
   --background: #f7f9fb;
   --surface: #f7f9fb;
+  --surface-elevated: #ffffff;
   --surface-container-low: #f2f4f6;
   --surface-container: #eceef0;
   --secondary: #565e74;
   --inverse-surface: #2d3133;
   --inverse-on-surface: #eff1f3;
   --error: #ba1a1a;
+  --banner-error-bg: #ffdad6;
+  --banner-error-text: #93000a;
+  --canvas-overlay-bg: rgba(247, 249, 251, 0.88);
+  --slide-canvas-bg: rgba(216, 218, 220, 0.15);
+  --hover-overlay: rgba(255, 255, 255, 0.65);
+  --badge-rag-text: #390c00;
+  --progress-done: #2d6a4f;
+  --input-focus-ring: rgba(168, 51, 0, 0.15);
   --sidebar-w: 280px;
   --topbar-h: 64px;
   --context-bar-h: 72px;
   --shadow-sm: 0 1px 2px rgba(0, 0, 0, 0.06);
 }
+[data-theme="dark"] {
+  color-scheme: dark;
+  --background: #121416;
+  --surface: #1a1c1e;
+  --surface-elevated: #242729;
+  --surface-container-low: #1e2022;
+  --surface-container: #2a2d30;
+  --secondary: #b8bcc8;
+  --secondary-container: #3d4454;
+  --on-secondary-container: #dae2fd;
+  --outline-variant: #4a403c;
+  --on-surface: #e3e2e6;
+  --primary-fixed: #5c2a12;
+  --banner-error-bg: #93000a;
+  --banner-error-text: #ffdad6;
+  --canvas-overlay-bg: rgba(18, 20, 22, 0.88);
+  --slide-canvas-bg: rgba(255, 255, 255, 0.04);
+  --hover-overlay: rgba(255, 255, 255, 0.06);
+  --badge-rag-text: #ffdbd0;
+  --progress-done: #6fcf97;
+  --input-focus-ring: rgba(203, 74, 24, 0.25);
+  --shadow-sm: 0 1px 2px rgba(0, 0, 0, 0.35);
+}
 * { box-sizing: border-box; }
 body {
 }
 .studio-banner-loading { background: var(--secondary-container); color: var(--on-secondary-container); }
+.studio-banner-error { background: var(--banner-error-bg); color: var(--banner-error-text); }
 .sidebar {
   position: fixed;
   left: var(--sidebar-w);
   right: 0;
   z-index: 35;
+  background: var(--surface-elevated);
   border-bottom: 1px solid var(--outline-variant);
   padding: 0.5rem 1.5rem;
 }
 }
 .card {
+  background: var(--surface-elevated);
   border: 1px solid var(--outline-variant);
   border-radius: var(--radius-xl);
   padding: 1rem;
 .btn:active { transform: scale(0.98); }
 .btn:disabled { opacity: 0.5; cursor: not-allowed; }
 .btn-primary { background: var(--primary); color: var(--on-primary); }
+.btn-secondary { background: var(--surface-elevated); border-color: var(--outline-variant); color: var(--on-surface); }
 .btn-ghost { background: transparent; border-color: var(--outline-variant); color: var(--secondary); text-decoration: none; }
 .btn-block { width: 100%; margin-top: 0.5rem; }
   border-radius: 8px;
   font: inherit;
   font-size: 0.9rem;
+  background: var(--surface-elevated);
+  color: var(--on-surface);
 }
 .input:focus {
   outline: none;
   border-color: var(--primary);
+  box-shadow: 0 0 0 2px var(--input-focus-ring);
 }
 .upload-zone {
 .controls-actions { display: flex; justify-content: flex-end; margin-top: 0.75rem; }
+.region-loading-host,
+.card-ingest,
+.card-chat,
+.voice-main-card,
+.coach-panel-wrap,
+.coach-debug-card,
+.controls-panel,
+.voice-rail-controls {
+  position: relative;
+}
 .slide-canvas {
   position: relative;
   flex: 1;
   min-height: 320px;
   border: 2px dashed var(--outline-variant);
   border-radius: var(--radius-xl);
+  background: var(--slide-canvas-bg);
   overflow: auto;
   padding: 1rem;
 }
   min-height: 280px;
 }
+.region-loading {
   position: absolute;
   inset: 0;
   z-index: 5;
   display: flex;
   align-items: center;
   justify-content: center;
+  background: transparent;
+  pointer-events: none;
 }
+.region-loading-inner {
+  display: flex;
+  flex-direction: column;
+  align-items: center;
+  gap: 0.5rem;
   text-align: center;
 }
+.studio-spinner,
 .canvas-spinner,
 .lesson-running-spinner {
   display: inline-block;
+  width: 32px;
+  height: 32px;
   border: 3px solid var(--outline-variant);
   border-top-color: var(--primary);
   border-radius: 50%;
   animation: studio-spin 0.9s linear infinite;
+  flex-shrink: 0;
 }
+.region-loading-text,
+.canvas-overlay-text {
+  margin: 0;
   font-size: 0.85rem;
+  font-weight: 500;
+  color: var(--secondary);
+}
+.region-loading-hint,
+.canvas-overlay-hint {
+  margin: 0;
+  font-size: 0.78rem;
   color: var(--secondary);
+  opacity: 0.85;
 }
 @keyframes studio-spin {
 .progress-step.pending { color: var(--secondary); }
 .progress-step.active { color: var(--primary); font-weight: 600; }
+.progress-step.done { color: var(--progress-done); }
 .progress-log {
   margin: 0.75rem 0 0;
   text-transform: uppercase;
 }
+.studio-badge-rag { background: var(--primary-fixed); color: var(--badge-rag-text); }
 .studio-badge-muted { background: var(--surface-container); color: var(--secondary); }
 .studio-doc-header { margin-bottom: 0.5rem; }
   padding: 0.5rem;
   border: 1px solid var(--outline-variant);
   border-radius: 8px;
+  background: var(--surface-elevated);
 }
 .studio-doc-card:hover { border-color: var(--primary); }
   padding: 0.6rem 0.75rem;
   border: 1px solid var(--outline-variant);
   border-radius: 8px;
+  background: var(--surface-elevated);
   font: inherit;
   cursor: pointer;
   text-align: left;
+  color: var(--on-surface);
 }
 .mode-card.active {
   border-radius: 6px;
 }
+.url-choice-item:hover { background: var(--hover-overlay); }
 .url-choice-item span {
   word-break: break-all;
 .research-chat-assistant {
   align-self: flex-start;
+  background: var(--surface-elevated);
   border: 1px solid var(--outline-variant);
 }
 .workspace[data-view="voice"] .col-research,
 .workspace[data-view="voice"] .col-slides { display: none; }
+.workspace[data-view="voice"] .col-debug { display: none; }
 .view-voice-only { display: none; }
 .workspace[data-view="voice"] .voice-main {
   min-width: 0;
+  display: flex;
+  flex-direction: column;
+  gap: 1rem;
+}
+.workspace[data-view="voice"] .voice-pitch-analysis {
+  margin: 0;
+}
+.workspace[data-view="voice"] .voice-pitch-analysis[open] .voice-pitch-summary {
+  margin-bottom: 0.75rem;
+}
+.workspace[data-view="voice"] .voice-pitch-summary {
+  cursor: pointer;
+  list-style: none;
+  display: flex;
+  flex-direction: column;
+  gap: 0.2rem;
+}
+.workspace[data-view="voice"] .voice-pitch-summary::-webkit-details-marker {
+  display: none;
+}
+.workspace[data-view="voice"] .voice-pitch-summary-hint {
+  font-size: 0.84rem;
+  color: var(--secondary);
+  font-weight: 400;
+}
+.workspace[data-view="voice"] .voice-pitch-analysis .coach-panel-wrap {
+  padding-top: 0.25rem;
+}
+.workspace[data-view="voice"] .voice-discuss-btn {
+  margin-top: 0.75rem;
+}
+.workspace[data-view="voice"] .coach-results-panel {
+  min-height: 80px;
+  margin-top: 0.75rem;
+  overflow-y: auto;
+}
+.workspace[data-view="voice"] .coach-results-panel:not(:empty) {
+  border-top: 1px solid var(--outline-variant);
+  padding-top: 0.75rem;
 }
 .workspace[data-view="voice"] .voice-main-card {
   }
 }
 .workspace[data-view="slides"] .col-studio,
 .workspace[data-view="research"] .col-debug { display: none; }
 .coach-card-head {
   margin-bottom: 0.85rem;
 }
 }
 @media (max-width: 960px) {
   .coach-capture-row {
     grid-template-columns: 1fr;
   }
   padding: 0.75rem;
   border: 1px solid var(--outline-variant);
   border-radius: 8px;
+  background: var(--surface-elevated);
   font-size: 0.875rem;
   line-height: 1.5;
   max-height: 240px;
   overflow-y: auto;
 }
+.theme-toggle {
+  margin-bottom: 0.25rem;
+}
 .debug-rag-scope {
   margin: 0.75rem 0;
 }
   font-size: 0.875rem;
 }
+.voice-rag-refs {
+  margin-top: 0.65rem;
+  padding-top: 0.55rem;
+  border-top: 1px solid var(--outline-variant);
+  font-size: 0.82rem;
+  color: var(--on-surface-variant);
+}
 .voice-replay-row {
   display: flex;
   flex-wrap: wrap;
   margin-top: 0;
 }
 .workspace[data-view="debug"] .view-debug-only { display: block; }
 .workspace[data-view="debug"] .col-research,

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

@@ -1,6 +1,30 @@
 import { Client } from "https://cdn.jsdelivr.net/npm/@gradio/client@1.14.0/+esm";
 const $ = (sel) => document.querySelector(sel);
 const SLIDE_PIPELINE_STEPS = [
   "Load language model",
   "Gather lesson sources",
@@ -33,6 +57,7 @@ const state = {
   browserRecordChunks: [],
   pendingVoiceAudioPath: null,
   pendingCoachAudioPath: null,
   useBrowserMic: true,
 };
@@ -249,14 +274,16 @@ async function discoverVoiceSources() {
     showError("Set a focus or workspace topic before discovering sources.");
     return;
   }
-  const data = await callApi("discover_sources", [topic, state.workspaceSessionId]);
-  $("#voice-ingest-status").textContent = stripMd(data.status || "Discovery complete.");
-  renderVoiceUrlChoices(data.urls || [], data.selected_urls || data.urls || []);
-  if (data.session_id) {
-    state.workspaceSessionId = data.session_id;
-    $("#workspace-session").value = data.session_id;
-  }
-  await refreshWorkspaceSessions(state.workspaceSessionId);
 }
 async function autoVoiceIngest() {
@@ -265,46 +292,53 @@ async function autoVoiceIngest() {
     showError("Set a focus or workspace topic before auto-ingest.");
     return;
   }
-  const data = await callApi("auto_search_ingest", [topic, state.workspaceSessionId]);
-  applyVoiceIngestResult(data);
-  state.voiceDiscoveredUrls = [];
-  state.voiceSelectedUrls = [];
-  renderVoiceUrlChoices([], []);
-  await refreshWorkspaceSessions(state.workspaceSessionId);
 }
 async function ingestVoiceSources() {
   const topic = voiceEffectiveTopic();
   const pasted = $("#voice-urls-text")?.value.trim() || "";
   const selected = getSelectedDiscoveredUrls("#voice-url-choices-list");
-  const paths = [];
   const files = $("#voice-ingest-file")?.files;
-  if (files?.length) {
-    for (const file of files) {
-      paths.push(await uploadFile(file));
-    }
-  }
-  if (!pasted && !selected.length && !paths.length) {
     showError("Add URLs, select suggested sources, or upload a file — then ingest.");
     return;
   }
-  const data = await callApi("ingest_sources", [
-    topic,
-    state.workspaceSessionId,
-    pasted,
-    selected,
-    paths,
-  ]);
-  applyVoiceIngestResult(data);
-  if (pasted) $("#voice-urls-text").value = "";
-  if (files?.length) $("#voice-ingest-file").value = "";
-  await refreshWorkspaceSessions(state.workspaceSessionId);
 }
 function syncVoiceModeUi() {
   const ragMode = state.voiceMode === "explain" || state.voiceMode === "lesson";
   $("#voice-topic-wrap")?.classList.toggle("hidden", !ragMode);
   $("#voice-rag-sources")?.classList.toggle("hidden", !ragMode);
   const placeholders = {
     explain: "e.g. How does finetuning differ from pretraining?",
     lesson: "What is the difference between pretraining and finetuning a small model?",
@@ -327,7 +361,10 @@ function renderVoiceChat() {
     if (item && typeof item === "object" && item.role) {
       const role = item.role === "user" ? "user" : "assistant";
       const label = role === "user" ? "You" : "Teacher";
-      const body = renderMarkdownLite(voiceMessageText(item.content));
       parts.push(
         `<div class="research-chat-bubble research-chat-${role}"><div class="research-chat-role">${label}</div><div class="research-chat-body">${body}</div></div>`
       );
@@ -396,15 +433,17 @@ async function discoverSources() {
     showError("Set a workspace topic before discovering sources.");
     return;
   }
-  const data = await callApi("discover_sources", [topic, state.workspaceSessionId]);
-  $("#ingest-status").textContent = stripMd(data.status || "Discovery complete.");
-  renderResearchUrlChoices(data.urls || [], data.selected_urls || data.urls || []);
-  if (data.session_id) {
-    state.workspaceSessionId = data.session_id;
-    $("#workspace-session").value = data.session_id;
-  }
-  setTracePanel("#research-trace-panel", data);
-  await refreshWorkspaceSessions(state.workspaceSessionId);
 }
 async function discoverSlideSources() {
@@ -413,8 +452,10 @@ async function discoverSlideSources() {
     showError("Set a topic before discovering sources.");
     return;
   }
-  const data = await callApi("discover_sources", [topic, state.workspaceSessionId]);
-  renderSlideUrlChoices(data.urls || [], data.selected_urls || data.urls || []);
 }
 async function autoSearchIngest() {
@@ -423,12 +464,14 @@ async function autoSearchIngest() {
     showError("Set a workspace topic before auto-ingest.");
     return;
   }
-  const data = await callApi("auto_search_ingest", [topic, state.workspaceSessionId]);
-  applyIngestResult(data);
-  state.discoveredUrls = [];
-  state.selectedUrls = [];
-  renderResearchUrlChoices([], []);
-  await refreshWorkspaceSessions(state.workspaceSessionId);
 }
 async function ingestSources({ urlsText = "", selectedUrls = [], pendingFiles = null } = {}) {
@@ -437,30 +480,32 @@ async function ingestSources({ urlsText = "", selectedUrls = [], pendingFiles =
   let selected = selectedUrls;
   if (workflow === "select") selected = getSelectedDiscoveredUrls();
   const pasted = workflow === "direct" ? urlsText : urlsText || $("#ingest-url").value.trim();
-  const paths = [];
   const files = pendingFiles || $("#ingest-file").files;
-  if (files?.length) {
-    for (const file of files) {
-      const b64 = await fileToBase64(file);
-      const saved = await callApi("save_upload", [file.name, b64]);
-      paths.push(saved.path);
-    }
-  }
-  if (!pasted && !selected.length && !paths.length) {
     showError("Add URLs, select suggested sources, or upload a file — then ingest.");
     return;
   }
-  const data = await callApi("ingest_sources", [
-    topic,
-    state.workspaceSessionId,
-    pasted,
-    selected,
-    paths,
-  ]);
-  applyIngestResult(data);
-  if (pasted) $("#ingest-url").value = "";
-  if (files?.length) $("#ingest-file").value = "";
-  await refreshWorkspaceSessions(state.workspaceSessionId);
 }
 function renderResearchChat() {
@@ -512,18 +557,20 @@ async function askResearchQuestion() {
     return;
   }
   const docIds = effectiveDocIds([]);
-  const data = await callApi("research_chat", [
-    question,
-    state.workspaceSessionId,
-    docIds,
-    state.researchChatHistory,
-  ]);
-  state.researchChatHistory = data.history || [];
-  renderResearchChat();
-  $("#research-question").value = "";
-  $("#research-chat-status").textContent = stripMd(data.rag_hint || "");
-  setTracePanel("#research-trace-panel", data);
-  updateResearchRagBadge();
 }
 async function sendDebugMessage() {
@@ -537,23 +584,25 @@ async function sendDebugMessage() {
   const debugDocIds = selectedDebugDocIds();
   const workspaceDocIds = selectedWorkspaceDocIds();
   const modelKey = $("#debug-model-key")?.value || "";
-  const data = await callApi("debug_chat", [
-    message,
-    state.debugChatHistory,
-    useRag,
-    debugSession,
-    debugDocIds,
-    modelKey,
-    state.workspaceSessionId,
-    workspaceDocIds,
-  ]);
-  state.debugChatHistory = data.history || [];
-  renderDebugChat();
-  $("#debug-message").value = "";
-  if (data.rag_hint) {
-    $("#debug-rag-hint").textContent = stripMd(data.rag_hint);
-  }
-  setTracePanel("#debug-trace-panel", data);
 }
 function effectiveDebugSessionId() {
@@ -651,10 +700,55 @@ async function getClient() {
   return state.client;
 }
 function setLoading(on) {
   $("#studio-loading").classList.toggle("hidden", !on);
 }
 function startProgressPanel() {
   const panel = $("#progress-panel");
   const stepsEl = $("#progress-steps");
@@ -968,91 +1062,107 @@ async function generateSlides() {
   const urlsText = $("#slide-urls-text")?.value.trim() || "";
   const selectedUrls = getSelectedDiscoveredUrls("#slide-url-choices-list");
-  const filePaths = [];
-  const slideFiles = $("#slide-source-files")?.files;
-  if (slideFiles?.length) {
-    for (const file of slideFiles) {
-      filePaths.push(await uploadFile(file));
-    }
-  }
-  startProgressPanel();
-  const waitTimer = advanceProgressWhileWaiting();
-  let data;
-  try {
-    data = await callApi("generate_slides", [
-      topic,
-      grade,
-      slideCount,
-      state.workspaceSessionId,
-      useRag,
-      docIds,
-      sourceMode,
-      searchWorkflow,
-      urlsText,
-      selectedUrls,
-      filePaths,
-    ]);
-  } catch (_err) {
-    $("#progress-eta").textContent = "Failed";
-    throw _err;
-  } finally {
-    clearInterval(waitTimer);
-    if (state.progressTimer) {
-      clearInterval(state.progressTimer);
-      state.progressTimer = null;
-    }
-  }
-  finishProgressPanel(data);
-  $("#generate-status").textContent = stripMd(data.status || "Slides generated.");
-  const canvasHtml =
-    data.canvas_html ||
-    (data.preview_html ? `<div class="studio-canvas-inner">${data.preview_html}</div>` : "");
-  $("#slide-canvas").innerHTML =
-    canvasHtml || '<div class="studio-canvas-empty"><p>Preview unavailable.</p></div>';
-  const galleryEl = $("#slide-gallery");
-  if (data.gallery_html) {
-    galleryEl.innerHTML = data.gallery_html;
-    galleryEl.classList.remove("hidden");
-  } else if (data.gallery?.length) {
-    galleryEl.innerHTML = data.gallery
-      .map(
-        (path, i) =>
-          `<a class="studio-gallery-item" href="${fileUrl(path)}" target="_blank" rel="noopener"><img src="${fileUrl(path)}" alt="Slide ${i + 1}" loading="lazy" /></a>`
-      )
-      .join("");
-    galleryEl.classList.remove("hidden");
-  } else {
-    galleryEl.classList.add("hidden");
-    galleryEl.innerHTML = "";
-  }
-  state.downloads = data.downloads;
-  const dl = $("#downloads");
-  if (data.downloads?.pptx) {
-    dl.classList.remove("hidden");
-    dl.innerHTML = `
       <a href="${fileUrl(data.downloads.pptx)}" download>PPTX</a>
       <a href="${fileUrl(data.downloads.docx)}" download>DOCX</a>
       <a href="${fileUrl(data.downloads.html)}" download>HTML</a>`;
-    $("#btn-export").disabled = false;
-  }
-  const outlineDetails = $("#slide-outline-details");
-  const outlineEl = $("#slide-outline");
-  if (data.outline_md) {
-    outlineEl.innerHTML = renderMarkdownLite(data.outline_md);
-    outlineDetails?.classList.remove("hidden");
-  } else {
-    outlineEl.innerHTML = "";
-    outlineDetails?.classList.add("hidden");
-  }
 }
 function renderVoiceReply(data, { keepAudio = false } = {}) {
   state.history = data.history ?? state.history;
   renderVoiceChat();
   if (data.status) {
     $("#voice-turn-status").textContent = stripMd(data.status);
@@ -1075,19 +1185,21 @@ async function sendVoiceTurn() {
   const useRag = voiceUseRag();
   const docIds = effectiveDocIds([]);
   const language = state.voicePresets?.default_language || "en";
-  const data = await callApi("teacher_voice_turn", [
-    message,
-    state.voiceMode,
-    topic,
-    state.workspaceSessionId,
-    useRag,
-    state.history,
-    docIds,
-    language,
-    null,
-  ]);
-  $("#voice-message").value = "";
-  renderVoiceReply(data);
 }
 async function sendVoiceAudioTurn(audioPath) {
@@ -1096,19 +1208,21 @@ async function sendVoiceAudioTurn(audioPath) {
   const docIds = effectiveDocIds([]);
   const language = state.voicePresets?.default_language || "en";
   const asr = state.voicePresets?.default_asr || null;
-  const data = await callApi("teacher_voice_audio_turn", [
-    audioPath,
-    state.voiceMode,
-    topic,
-    state.workspaceSessionId,
-    useRag,
-    state.history,
-    docIds,
-    language,
-    asr,
-  ]);
-  if (data.user_text) $("#voice-message").value = data.user_text;
-  renderVoiceReply(data);
 }
 async function speakVoiceReply(firstSentenceOnly) {
@@ -1139,13 +1253,36 @@ async function analyzePitchWithPath(audioPath) {
   const language = $("#coach-language")?.value || "en";
   const asr = $("#coach-asr")?.value || null;
   const speakRewrite = $("#coach-speak-rewrite")?.checked || false;
-  $("#coach-panel").innerHTML = `
-    <div class="studio-coach-panel studio-coach-live">
-      <div class="studio-coach-header"><span class="studio-coach-dot"></span>
-      <span class="studio-coach-label">Analyzing…</span></div>
-    </div>`;
-  const data = await callApi("analyze_pitch", [audioPath, language, asr, speakRewrite]);
-  $("#coach-panel").innerHTML = data.coach_panel_html || "";
 }
 async function analyzePitch() {
@@ -1265,6 +1402,8 @@ function bindUi() {
   $("#btn-open-settings")?.addEventListener("click", openSettingsDrawer);
   $("#btn-close-settings")?.addEventListener("click", closeSettingsDrawer);
   $("#settings-backdrop")?.addEventListener("click", closeSettingsDrawer);
   $("#btn-reload-model")?.addEventListener("click", () => reloadModelFromSettings().catch(() => {}));
   $("#btn-open-research-view")?.addEventListener("click", openResearchView);
@@ -1375,6 +1514,8 @@ function bindUi() {
       syncVoiceModeUi();
     });
   });
 }
 bindUi();

 import { Client } from "https://cdn.jsdelivr.net/npm/@gradio/client@1.14.0/+esm";
 const $ = (sel) => document.querySelector(sel);
+const THEME_KEY = "studio-theme";
+function getPreferredTheme() {
+  const saved = localStorage.getItem(THEME_KEY);
+  if (saved === "light" || saved === "dark") return saved;
+  return window.matchMedia("(prefers-color-scheme: dark)").matches ? "dark" : "light";
+}
+function applyTheme(theme) {
+  document.documentElement.dataset.theme = theme === "dark" ? "dark" : "light";
+  localStorage.setItem(THEME_KEY, theme);
+  const icon = $("#theme-icon");
+  if (icon) icon.textContent = theme === "dark" ? "light_mode" : "dark_mode";
+  const checkbox = $("#theme-toggle");
+  if (checkbox) checkbox.checked = theme === "dark";
+}
+function toggleTheme() {
+  const current = document.documentElement.dataset.theme === "dark" ? "dark" : "light";
+  applyTheme(current === "dark" ? "light" : "dark");
+}
+applyTheme(getPreferredTheme());
 const SLIDE_PIPELINE_STEPS = [
   "Load language model",
   "Gather lesson sources",
   browserRecordChunks: [],
   pendingVoiceAudioPath: null,
   pendingCoachAudioPath: null,
+  lastPitchAnalysis: null,
   useBrowserMic: true,
 };
     showError("Set a focus or workspace topic before discovering sources.");
     return;
   }
+  await withRegionLoading($(".voice-rail-controls"), "Discovering sources…", async () => {
+    const data = await callApi("discover_sources", [topic, state.workspaceSessionId]);
+    $("#voice-ingest-status").textContent = stripMd(data.status || "Discovery complete.");
+    renderVoiceUrlChoices(data.urls || [], data.selected_urls || data.urls || []);
+    if (data.session_id) {
+      state.workspaceSessionId = data.session_id;
+      $("#workspace-session").value = data.session_id;
+    }
+    await refreshWorkspaceSessions(state.workspaceSessionId);
+  });
 }
 async function autoVoiceIngest() {
     showError("Set a focus or workspace topic before auto-ingest.");
     return;
   }
+  await withRegionLoading($(".voice-rail-controls"), "Auto-ingesting sources…", async () => {
+    const data = await callApi("auto_search_ingest", [topic, state.workspaceSessionId]);
+    applyVoiceIngestResult(data);
+    state.voiceDiscoveredUrls = [];
+    state.voiceSelectedUrls = [];
+    renderVoiceUrlChoices([], []);
+    await refreshWorkspaceSessions(state.workspaceSessionId);
+  });
 }
 async function ingestVoiceSources() {
   const topic = voiceEffectiveTopic();
   const pasted = $("#voice-urls-text")?.value.trim() || "";
   const selected = getSelectedDiscoveredUrls("#voice-url-choices-list");
   const files = $("#voice-ingest-file")?.files;
+  if (!pasted && !selected.length && !files?.length) {
     showError("Add URLs, select suggested sources, or upload a file — then ingest.");
     return;
   }
+  await withRegionLoading($(".voice-rail-controls"), "Ingesting sources…", async () => {
+    const paths = [];
+    if (files?.length) {
+      for (const file of files) {
+        paths.push(await uploadFile(file));
+      }
+    }
+    const data = await callApi("ingest_sources", [
+      topic,
+      state.workspaceSessionId,
+      pasted,
+      selected,
+      paths,
+    ]);
+    applyVoiceIngestResult(data);
+    if (pasted) $("#voice-urls-text").value = "";
+    if (files?.length) $("#voice-ingest-file").value = "";
+    await refreshWorkspaceSessions(state.workspaceSessionId);
+  });
 }
 function syncVoiceModeUi() {
   const ragMode = state.voiceMode === "explain" || state.voiceMode === "lesson";
+  const practiceMode = state.voiceMode === "pitch";
   $("#voice-topic-wrap")?.classList.toggle("hidden", !ragMode);
   $("#voice-rag-sources")?.classList.toggle("hidden", !ragMode);
+  $(".voice-rag-card")?.classList.toggle("hidden", practiceMode);
+  $("#voice-pitch-analysis")?.classList.toggle("hidden", !practiceMode);
   const placeholders = {
     explain: "e.g. How does finetuning differ from pretraining?",
     lesson: "What is the difference between pretraining and finetuning a small model?",
     if (item && typeof item === "object" && item.role) {
       const role = item.role === "user" ? "user" : "assistant";
       const label = role === "user" ? "You" : "Teacher";
+      let body = renderMarkdownLite(voiceMessageText(item.content));
+      if (role === "assistant" && item.rag_references) {
+        body += `<div class="voice-rag-refs">${renderMarkdownLite(item.rag_references)}</div>`;
+      }
       parts.push(
         `<div class="research-chat-bubble research-chat-${role}"><div class="research-chat-role">${label}</div><div class="research-chat-body">${body}</div></div>`
       );
     showError("Set a workspace topic before discovering sources.");
     return;
   }
+  await withRegionLoading($(".card-ingest"), "Discovering sources…", async () => {
+    const data = await callApi("discover_sources", [topic, state.workspaceSessionId]);
+    $("#ingest-status").textContent = stripMd(data.status || "Discovery complete.");
+    renderResearchUrlChoices(data.urls || [], data.selected_urls || data.urls || []);
+    if (data.session_id) {
+      state.workspaceSessionId = data.session_id;
+      $("#workspace-session").value = data.session_id;
+    }
+    setTracePanel("#research-trace-panel", data);
+    await refreshWorkspaceSessions(state.workspaceSessionId);
+  });
 }
 async function discoverSlideSources() {
     showError("Set a topic before discovering sources.");
     return;
   }
+  await withRegionLoading($(".controls-panel"), "Discovering sources…", async () => {
+    const data = await callApi("discover_sources", [topic, state.workspaceSessionId]);
+    renderSlideUrlChoices(data.urls || [], data.selected_urls || data.urls || []);
+  });
 }
 async function autoSearchIngest() {
     showError("Set a workspace topic before auto-ingest.");
     return;
   }
+  await withRegionLoading($(".card-ingest"), "Auto-ingesting sources…", async () => {
+    const data = await callApi("auto_search_ingest", [topic, state.workspaceSessionId]);
+    applyIngestResult(data);
+    state.discoveredUrls = [];
+    state.selectedUrls = [];
+    renderResearchUrlChoices([], []);
+    await refreshWorkspaceSessions(state.workspaceSessionId);
+  });
 }
 async function ingestSources({ urlsText = "", selectedUrls = [], pendingFiles = null } = {}) {
   let selected = selectedUrls;
   if (workflow === "select") selected = getSelectedDiscoveredUrls();
   const pasted = workflow === "direct" ? urlsText : urlsText || $("#ingest-url").value.trim();
   const files = pendingFiles || $("#ingest-file").files;
+  if (!pasted && !selected.length && !files?.length) {
     showError("Add URLs, select suggested sources, or upload a file — then ingest.");
     return;
   }
+  await withRegionLoading($(".card-ingest"), "Ingesting sources…", async () => {
+    const paths = [];
+    if (files?.length) {
+      for (const file of files) {
+        const b64 = await fileToBase64(file);
+        const saved = await callApi("save_upload", [file.name, b64]);
+        paths.push(saved.path);
+      }
+    }
+    const data = await callApi("ingest_sources", [
+      topic,
+      state.workspaceSessionId,
+      pasted,
+      selected,
+      paths,
+    ]);
+    applyIngestResult(data);
+    if (pasted) $("#ingest-url").value = "";
+    if (files?.length) $("#ingest-file").value = "";
+    await refreshWorkspaceSessions(state.workspaceSessionId);
+  });
 }
 function renderResearchChat() {
     return;
   }
   const docIds = effectiveDocIds([]);
+  await withRegionLoading($("#research-chat-panel .card-chat"), "Searching sources…", async () => {
+    const data = await callApi("research_chat", [
+      question,
+      state.workspaceSessionId,
+      docIds,
+      state.researchChatHistory,
+    ]);
+    state.researchChatHistory = data.history || [];
+    renderResearchChat();
+    $("#research-question").value = "";
+    $("#research-chat-status").textContent = stripMd(data.rag_hint || "");
+    setTracePanel("#research-trace-panel", data);
+    updateResearchRagBadge();
+  });
 }
 async function sendDebugMessage() {
   const debugDocIds = selectedDebugDocIds();
   const workspaceDocIds = selectedWorkspaceDocIds();
   const modelKey = $("#debug-model-key")?.value || "";
+  await withRegionLoading($(".coach-debug-card"), "Thinking…", async () => {
+    const data = await callApi("debug_chat", [
+      message,
+      state.debugChatHistory,
+      useRag,
+      debugSession,
+      debugDocIds,
+      modelKey,
+      state.workspaceSessionId,
+      workspaceDocIds,
+    ]);
+    state.debugChatHistory = data.history || [];
+    renderDebugChat();
+    $("#debug-message").value = "";
+    if (data.rag_hint) {
+      $("#debug-rag-hint").textContent = stripMd(data.rag_hint);
+    }
+    setTracePanel("#debug-trace-panel", data);
+  });
 }
 function effectiveDebugSessionId() {
   return state.client;
 }
+let globalLoadingSuppress = 0;
 function setLoading(on) {
+  if (on && globalLoadingSuppress > 0) return;
   $("#studio-loading").classList.toggle("hidden", !on);
 }
+function setRegionLoading(container, on, message = "Working…", { overlayEl = null, hint = "" } = {}) {
+  if (!container) return;
+  let overlay = overlayEl || container.querySelector(":scope > .region-loading");
+  if (!overlay) {
+    overlay = document.createElement("div");
+    overlay.className = "region-loading hidden";
+    overlay.setAttribute("aria-live", "polite");
+    overlay.innerHTML = `
+      <div class="region-loading-inner">
+        <span class="studio-spinner" aria-hidden="true"></span>
+        <p class="region-loading-text"></p>
+        <p class="region-loading-hint hidden"></p>
+      </div>`;
+    container.insertBefore(overlay, container.firstChild);
+    if (getComputedStyle(container).position === "static") {
+      container.classList.add("region-loading-host");
+    }
+  }
+  const textEl =
+    overlay.querySelector(".region-loading-text") || overlay.querySelector("#canvas-overlay-text");
+  if (textEl) textEl.textContent = message;
+  const hintEl =
+    overlay.querySelector(".region-loading-hint") || overlay.querySelector(".canvas-overlay-hint");
+  if (hintEl) {
+    hintEl.textContent = hint;
+    hintEl.classList.toggle("hidden", !hint);
+  }
+  overlay.classList.toggle("hidden", !on);
+  container.setAttribute("aria-busy", on ? "true" : "false");
+}
+async function withRegionLoading(container, message, fn, options = {}) {
+  globalLoadingSuppress += 1;
+  setRegionLoading(container, true, message, options);
+  try {
+    return await fn();
+  } finally {
+    globalLoadingSuppress -= 1;
+    setRegionLoading(container, false, message, options);
+  }
+}
 function startProgressPanel() {
   const panel = $("#progress-panel");
   const stepsEl = $("#progress-steps");
   const urlsText = $("#slide-urls-text")?.value.trim() || "";
   const selectedUrls = getSelectedDiscoveredUrls("#slide-url-choices-list");
+  await withRegionLoading(
+    $("#slide-canvas"),
+    "Generating slides…",
+    async () => {
+      const filePaths = [];
+      const slideFiles = $("#slide-source-files")?.files;
+      if (slideFiles?.length) {
+        for (const file of slideFiles) {
+          filePaths.push(await uploadFile(file));
+        }
+      }
+      startProgressPanel();
+      const waitTimer = advanceProgressWhileWaiting();
+      let data;
+      try {
+        data = await callApi("generate_slides", [
+          topic,
+          grade,
+          slideCount,
+          state.workspaceSessionId,
+          useRag,
+          docIds,
+          sourceMode,
+          searchWorkflow,
+          urlsText,
+          selectedUrls,
+          filePaths,
+        ]);
+      } catch (_err) {
+        $("#progress-eta").textContent = "Failed";
+        throw _err;
+      } finally {
+        clearInterval(waitTimer);
+        if (state.progressTimer) {
+          clearInterval(state.progressTimer);
+          state.progressTimer = null;
+        }
+      }
+      finishProgressPanel(data);
+      $("#generate-status").textContent = stripMd(data.status || "Slides generated.");
+      const canvasHtml =
+        data.canvas_html ||
+        (data.preview_html ? `<div class="studio-canvas-inner">${data.preview_html}</div>` : "");
+      $("#slide-canvas-content").innerHTML =
+        canvasHtml || '<div class="studio-canvas-empty"><p>Preview unavailable.</p></div>';
+      const galleryEl = $("#slide-gallery");
+      if (data.gallery_html) {
+        galleryEl.innerHTML = data.gallery_html;
+        galleryEl.classList.remove("hidden");
+      } else if (data.gallery?.length) {
+        galleryEl.innerHTML = data.gallery
+          .map(
+            (path, i) =>
+              `<a class="studio-gallery-item" href="${fileUrl(path)}" target="_blank" rel="noopener"><img src="${fileUrl(path)}" alt="Slide ${i + 1}" loading="lazy" /></a>`
+          )
+          .join("");
+        galleryEl.classList.remove("hidden");
+      } else {
+        galleryEl.classList.add("hidden");
+        galleryEl.innerHTML = "";
+      }
+      state.downloads = data.downloads;
+      const dl = $("#downloads");
+      if (data.downloads?.pptx) {
+        dl.classList.remove("hidden");
+        dl.innerHTML = `
       <a href="${fileUrl(data.downloads.pptx)}" download>PPTX</a>
       <a href="${fileUrl(data.downloads.docx)}" download>DOCX</a>
       <a href="${fileUrl(data.downloads.html)}" download>HTML</a>`;
+        $("#btn-export").disabled = false;
+      }
+      const outlineDetails = $("#slide-outline-details");
+      const outlineEl = $("#slide-outline");
+      if (data.outline_md) {
+        outlineEl.innerHTML = renderMarkdownLite(data.outline_md);
+        outlineDetails?.classList.remove("hidden");
+      } else {
+        outlineEl.innerHTML = "";
+        outlineDetails?.classList.add("hidden");
+      }
+    },
+    {
+      overlayEl: $("#canvas-overlay"),
+      hint: "Local CPU models often take 30–90 seconds.",
+    }
+  );
 }
 function renderVoiceReply(data, { keepAudio = false } = {}) {
   state.history = data.history ?? state.history;
+  if (data.rag_references && state.history.length) {
+    const last = state.history[state.history.length - 1];
+    if (last && typeof last === "object" && last.role === "assistant") {
+      last.rag_references = data.rag_references;
+    }
+  }
   renderVoiceChat();
   if (data.status) {
     $("#voice-turn-status").textContent = stripMd(data.status);
   const useRag = voiceUseRag();
   const docIds = effectiveDocIds([]);
   const language = state.voicePresets?.default_language || "en";
+  await withRegionLoading($(".voice-main-card"), "Teacher is thinking…", async () => {
+    const data = await callApi("teacher_voice_turn", [
+      message,
+      state.voiceMode,
+      topic,
+      state.workspaceSessionId,
+      useRag,
+      state.history,
+      docIds,
+      language,
+      null,
+    ]);
+    $("#voice-message").value = "";
+    renderVoiceReply(data);
+  });
 }
 async function sendVoiceAudioTurn(audioPath) {
   const docIds = effectiveDocIds([]);
   const language = state.voicePresets?.default_language || "en";
   const asr = state.voicePresets?.default_asr || null;
+  await withRegionLoading($(".voice-main-card"), "Processing voice…", async () => {
+    const data = await callApi("teacher_voice_audio_turn", [
+      audioPath,
+      state.voiceMode,
+      topic,
+      state.workspaceSessionId,
+      useRag,
+      state.history,
+      docIds,
+      language,
+      asr,
+    ]);
+    if (data.user_text) $("#voice-message").value = data.user_text;
+    renderVoiceReply(data);
+  });
 }
 async function speakVoiceReply(firstSentenceOnly) {
   const language = $("#coach-language")?.value || "en";
   const asr = $("#coach-asr")?.value || null;
   const speakRewrite = $("#coach-speak-rewrite")?.checked || false;
+  await withRegionLoading($("#voice-pitch-analysis"), "Analyzing pitch…", async () => {
+    const data = await callApi("analyze_pitch", [audioPath, language, asr, speakRewrite]);
+    state.lastPitchAnalysis = data;
+    const panel = $("#coach-panel");
+    panel.innerHTML = data.coach_panel_html || "";
+    const discussBtn = document.createElement("button");
+    discussBtn.type = "button";
+    discussBtn.className = "btn btn-secondary voice-discuss-btn";
+    discussBtn.textContent = "Discuss in chat";
+    discussBtn.addEventListener("click", () => discussPitchInChat().catch(() => {}));
+    if (data.transcript_html || data.report_md || data.tip) {
+      panel.appendChild(discussBtn);
+    }
+  });
+}
+function discussPitchInChat() {
+  const data = state.lastPitchAnalysis;
+  if (!data) return;
+  const parts = [];
+  if (data.tip) parts.push(`Coach tip: ${stripMd(data.tip)}`);
+  if (data.report_md) parts.push(stripMd(data.report_md).slice(0, 800));
+  const prompt =
+    parts.length > 0
+      ? `Here is my pitch analysis. Help me improve based on this feedback:\n\n${parts.join("\n\n")}`
+      : "I just ran pitch analysis — what should I work on next?";
+  $("#voice-message").value = prompt;
+  $("#voice-message").focus();
+  const chat = $("#voice-chat-messages");
+  if (chat) chat.scrollIntoView({ behavior: "smooth", block: "nearest" });
 }
 async function analyzePitch() {
   $("#btn-open-settings")?.addEventListener("click", openSettingsDrawer);
   $("#btn-close-settings")?.addEventListener("click", closeSettingsDrawer);
   $("#settings-backdrop")?.addEventListener("click", closeSettingsDrawer);
+  $("#theme-toggle")?.addEventListener("change", toggleTheme);
+  $("#theme-toggle-btn")?.addEventListener("click", toggleTheme);
   $("#btn-reload-model")?.addEventListener("click", () => reloadModelFromSettings().catch(() => {}));
   $("#btn-open-research-view")?.addEventListener("click", openResearchView);
       syncVoiceModeUi();
     });
   });
+  syncVoiceModeUi();
 }
 bindUi();

libs/agent/src/agent/research_prompts.py CHANGED Viewed

@@ -18,6 +18,8 @@ def research_answer_system(skill_body: str, skill_path: Path) -> str:
         "Each context block is numbered [1], [2], … — one number per source document.",
         "Cite with those numbers only (e.g. [1]). Use at most a few citations per answer.",
         "Ignore any [n] markers inside source text; never list citation numbers in a row.",
         skill_body,
     ]
     if citation_ref:
@@ -33,4 +35,5 @@ Question: {question}
 Write a concise answer with inline [n] citations (one index per source document).
 Do not append a References section — it is added automatically.
-If context is insufficient, say so."""

         "Each context block is numbered [1], [2], … — one number per source document.",
         "Cite with those numbers only (e.g. [1]). Use at most a few citations per answer.",
         "Ignore any [n] markers inside source text; never list citation numbers in a row.",
+        "Write the final answer directly in 2–4 sentences. Do not use thinking tags, "
+        "chain-of-thought, or planning text.",
         skill_body,
     ]
     if citation_ref:
 Write a concise answer with inline [n] citations (one index per source document).
 Do not append a References section — it is added automatically.
+Do not use redacted_thinking, think tags, or step-by-step planning — only the answer.
+If context is insufficient, say so in one sentence."""

libs/agent/src/agent/runner.py CHANGED Viewed

@@ -906,7 +906,13 @@ class AgentRunner:
             model=model_key,
             user_input=req.model_dump(),
         )
         backend.load()
         answer_tool = self._tools.get("research_answer")
         raw_answer, citations, refs = answer_tool.handler(
@@ -916,6 +922,7 @@ class AgentRunner:
             skill_path=skill.path,
             session_id=req.session_id,
             doc_ids=req.doc_ids or None,
         )
         trace.log_llm(req.question, raw_answer)
         trace.log_note(

             model=model_key,
             user_input=req.model_dump(),
         )
+        load_started = monotonic()
         backend.load()
+        trace.log_step(
+            "load_model",
+            "Load language model",
+            duration_ms=int((monotonic() - load_started) * 1000),
+        )
         answer_tool = self._tools.get("research_answer")
         raw_answer, citations, refs = answer_tool.handler(
             skill_path=skill.path,
             session_id=req.session_id,
             doc_ids=req.doc_ids or None,
+            trace=trace,
         )
         trace.log_llm(req.question, raw_answer)
         trace.log_note(

libs/agent/src/agent/tools/research_tools.py CHANGED Viewed

@@ -1,6 +1,7 @@
 from __future__ import annotations
 from pathlib import Path
 from typing import Any
 from researchmind.citations import Citation, clean_model_answer, format_context_block, format_references
@@ -51,10 +52,12 @@ def tool_research_answer(
     skill_path: Path,
     session_id: str | None = None,
     doc_ids: list[str] | None = None,
 ) -> tuple[str, list[Citation], str]:
     cfg = get_config()
     store = get_store()
     scope_session, scope_docs = resolve_retrieve_scope(session_id, doc_ids)
     chunks = retrieve(
         question,
         store,
@@ -62,6 +65,15 @@ def tool_research_answer(
         session_id=scope_session,
         doc_ids=scope_docs,
     )
     if not chunks:
         hint = rag_scope_warning(session_id=session_id, doc_ids=doc_ids)
         return hint, [], ""
@@ -73,9 +85,17 @@ def tool_research_answer(
         {"role": "system", "content": system},
         {"role": "user", "content": user},
     ]
     answer = clean_model_answer(
-        backend.chat(messages, max_tokens=1024, temperature=0.3)
     )
     refs = format_references(citations)
     if session_id:
         store.add_message(session_id, "user", question, [c.chunk_id for c in citations])

 from __future__ import annotations
 from pathlib import Path
+from time import monotonic
 from typing import Any
 from researchmind.citations import Citation, clean_model_answer, format_context_block, format_references
     skill_path: Path,
     session_id: str | None = None,
     doc_ids: list[str] | None = None,
+    trace: Any | None = None,
 ) -> tuple[str, list[Citation], str]:
     cfg = get_config()
     store = get_store()
     scope_session, scope_docs = resolve_retrieve_scope(session_id, doc_ids)
+    retrieve_started = monotonic()
     chunks = retrieve(
         question,
         store,
         session_id=scope_session,
         doc_ids=scope_docs,
     )
+    if trace is not None:
+        trace.log_step(
+            "retrieve",
+            "Retrieve passages",
+            duration_ms=int((monotonic() - retrieve_started) * 1000),
+            chunks=len(chunks),
+            session_id=scope_session or "",
+            doc_ids=scope_docs or [],
+        )
     if not chunks:
         hint = rag_scope_warning(session_id=session_id, doc_ids=doc_ids)
         return hint, [], ""
         {"role": "system", "content": system},
         {"role": "user", "content": user},
     ]
+    generate_started = monotonic()
     answer = clean_model_answer(
+        backend.chat(messages, max_tokens=512, temperature=0.2)
     )
+    if trace is not None:
+        trace.log_step(
+            "generate",
+            "Generate cited answer",
+            duration_ms=int((monotonic() - generate_started) * 1000),
+            citations=len(citations),
+        )
     refs = format_references(citations)
     if session_id:
         store.add_message(session_id, "user", question, [c.chunk_id for c in citations])

libs/echocoach/src/echocoach/prompts.py CHANGED Viewed

@@ -27,7 +27,7 @@ If a lesson topic is set, stay focused on it. When source excerpts are provided,
 PITCH_SYSTEM = """You are TeacherVoice, a supportive public-speaking coach in a live conversation.
 Give brief, actionable feedback on what the student just said (opening, clarity, energy, structure).
 Do not produce JSON or long reports — speak naturally in 2-4 sentences.
-Suggest one concrete improvement for their next attempt. For charts and pace analysis, they can use the EchoCoach tab."""
 _MODE_SYSTEM: dict[TeacherVoiceMode, str] = {
     "explain": EXPLAIN_SYSTEM,

 PITCH_SYSTEM = """You are TeacherVoice, a supportive public-speaking coach in a live conversation.
 Give brief, actionable feedback on what the student just said (opening, clarity, energy, structure).
 Do not produce JSON or long reports — speak naturally in 2-4 sentences.
+Suggest one concrete improvement for their next attempt. For charts and pace analysis, expand **Deep pitch analysis** below the chat."""
 _MODE_SYSTEM: dict[TeacherVoiceMode, str] = {
     "explain": EXPLAIN_SYSTEM,

libs/echocoach/src/echocoach/teacher_voice.py CHANGED Viewed

@@ -12,7 +12,7 @@ from agent.trace import TraceRecorder
 from inference.base import InferenceBackend
 from inference.response_clean import (
     needs_teacher_compaction,
-    prepare_display_reply,
     strip_reasoning_output,
 )
 from researchmind.ingest import IngestPipeline
@@ -161,6 +161,7 @@ def fetch_rag_context(
 def _rag_turn_via_agent(
     user_text: str,
     *,
     topic: str | None,
     session_id: str,
     doc_ids: list[str] | None,
@@ -198,8 +199,13 @@ def _rag_turn_via_agent(
         research_trace=result.trace_path,
     )
-    assistant_text = result.answer.strip()
-    display_reply = prepare_display_reply(assistant_text)
     rag_refs = result.references_markdown or None
     return assistant_text, rag_refs, rag_status, display_reply
@@ -251,25 +257,36 @@ def _compact_teacher_reply(
     return compact or seed
-def _finalize_non_rag_reply(
     raw_reply: str,
     *,
     mode: TeacherVoiceMode,
     backend: InferenceBackend,
     trace: TraceRecorder,
 ) -> tuple[str, str]:
     assistant_text = strip_reasoning_output(raw_reply).strip()
-    if needs_teacher_compaction(raw_reply) or not assistant_text:
         assistant_text = _compact_teacher_reply(
             raw_reply,
             mode=mode,
             backend=backend,
             trace=trace,
         )
-    display_reply = prepare_display_reply(raw_reply)
-    if needs_teacher_compaction(display_reply):
-        display_reply = prepare_display_reply(assistant_text)
-    return assistant_text, display_reply
 def build_teacher_messages(
@@ -320,6 +337,7 @@ def _generate_teacher_reply(
     if use_rag and mode in RAG_MODES:
         assistant_text, rag_refs, rag_status, display_reply = _rag_turn_via_agent(
             user_text,
             topic=topic,
             session_id=session_id,
             doc_ids=doc_ids,
@@ -334,8 +352,8 @@ def _generate_teacher_reply(
             user_text=user_text,
             topic=topic,
         )
-        raw_reply = backend.chat(messages, max_tokens=256, temperature=0.2)
-        assistant_text, display_reply = _finalize_non_rag_reply(
             raw_reply,
             mode=mode,
             backend=backend,

 from inference.base import InferenceBackend
 from inference.response_clean import (
     needs_teacher_compaction,
+    reply_ends_complete_sentence,
     strip_reasoning_output,
 )
 from researchmind.ingest import IngestPipeline
 def _rag_turn_via_agent(
     user_text: str,
     *,
+    mode: TeacherVoiceMode,
     topic: str | None,
     session_id: str,
     doc_ids: list[str] | None,
         research_trace=result.trace_path,
     )
+    raw_answer = strip_references_for_tts(result.answer.strip())
+    assistant_text, display_reply = _finalize_voice_reply(
+        raw_answer,
+        mode=mode,
+        backend=backend,
+        trace=trace,
+    )
     rag_refs = result.references_markdown or None
     return assistant_text, rag_refs, rag_status, display_reply
     return compact or seed
+def _finalize_voice_reply(
     raw_reply: str,
     *,
     mode: TeacherVoiceMode,
     backend: InferenceBackend,
     trace: TraceRecorder,
 ) -> tuple[str, str]:
+    """Normalize model output into a complete spoken reply and chat display text."""
     assistant_text = strip_reasoning_output(raw_reply).strip()
+    needs_fix = (
+        not assistant_text
+        or needs_teacher_compaction(raw_reply)
+        or needs_teacher_compaction(assistant_text)
+        or not reply_ends_complete_sentence(assistant_text)
+    )
+    if needs_fix:
         assistant_text = _compact_teacher_reply(
             raw_reply,
             mode=mode,
             backend=backend,
             trace=trace,
         )
+    if not reply_ends_complete_sentence(assistant_text):
+        assistant_text = _compact_teacher_reply(
+            assistant_text or raw_reply,
+            mode=mode,
+            backend=backend,
+            trace=trace,
+        )
+    return assistant_text, assistant_text
 def build_teacher_messages(
     if use_rag and mode in RAG_MODES:
         assistant_text, rag_refs, rag_status, display_reply = _rag_turn_via_agent(
             user_text,
+            mode=mode,
             topic=topic,
             session_id=session_id,
             doc_ids=doc_ids,
             user_text=user_text,
             topic=topic,
         )
+        raw_reply = backend.chat(messages, max_tokens=512, temperature=0.2)
+        assistant_text, display_reply = _finalize_voice_reply(
             raw_reply,
             mode=mode,
             backend=backend,

libs/echocoach/tests/test_teacher_voice.py CHANGED Viewed

@@ -6,6 +6,7 @@ import numpy as np
 import pytest
 import soundfile as sf
 from echocoach.prompts import PITCH_SYSTEM, system_prompt_for_mode
 from echocoach.teacher_voice import (
     RagContext,
@@ -131,7 +132,7 @@ def test_build_teacher_messages_includes_topic_and_rag():
 def test_pitch_mode_system_prompt():
-    assert "EchoCoach" in system_prompt_for_mode("pitch")
     assert PITCH_SYSTEM == system_prompt_for_mode("pitch")
@@ -221,6 +222,7 @@ def test_rag_turn_via_agent_mock(monkeypatch, tmp_path):
     trace = TraceRecorder(skill="teacher-voice", model="test", user_input={})
     text, refs, status, display = _rag_turn_via_agent(
         "How do plants eat?",
         topic="Photosynthesis",
         session_id="",
         doc_ids=None,
@@ -251,6 +253,30 @@ def research_env(tmp_path, monkeypatch):
     monkeypatch.setenv("AGENT_OUTPUTS_DIR", str(tmp_path / "outputs"))
 def test_run_teacher_voice_text_turn_mock(monkeypatch, tmp_path):
     from echocoach.teacher_voice import run_teacher_voice_text_turn

 import pytest
 import soundfile as sf
+from inference.response_clean import reply_ends_complete_sentence
 from echocoach.prompts import PITCH_SYSTEM, system_prompt_for_mode
 from echocoach.teacher_voice import (
     RagContext,
 def test_pitch_mode_system_prompt():
+    assert "Deep pitch analysis" in system_prompt_for_mode("pitch")
     assert PITCH_SYSTEM == system_prompt_for_mode("pitch")
     trace = TraceRecorder(skill="teacher-voice", model="test", user_input={})
     text, refs, status, display = _rag_turn_via_agent(
         "How do plants eat?",
+        mode="explain",
         topic="Photosynthesis",
         session_id="",
         doc_ids=None,
     monkeypatch.setenv("AGENT_OUTPUTS_DIR", str(tmp_path / "outputs"))
+def test_finalize_voice_reply_compacts_incomplete_sentence():
+    from echocoach.teacher_voice import _finalize_voice_reply
+    from agent.trace import TraceRecorder
+    class _Backend:
+        def chat(self, messages, *, max_tokens=512, temperature=0.2):
+            return (
+                "Finetuning adapts a pretrained small model to your task using extra labeled data. "
+                "You keep most of the base weights and train on a focused dataset. "
+                "That usually beats prompting alone for domain-specific work."
+            )
+    trace = TraceRecorder(skill="teacher-voice", model="test", user_input={})
+    text, display = _finalize_voice_reply(
+        "The lesson aims to teach how to fine-tune small",
+        mode="lesson",
+        backend=_Backend(),
+        trace=trace,
+    )
+    assert reply_ends_complete_sentence(text)
+    assert "fine-tune" in text.lower() or "finetun" in text.lower()
+    assert text == display
 def test_run_teacher_voice_text_turn_mock(monkeypatch, tmp_path):
     from echocoach.teacher_voice import run_teacher_voice_text_turn

libs/inference/src/inference/response_clean.py CHANGED Viewed

@@ -50,8 +50,10 @@ _COMPLETE_SENTENCE = re.compile(r"[.!?][\"')\]]*\s*$")
 _LIST_OUTLINE = re.compile(r"^\d+\.\s", re.MULTILINE)
 _REASONING_OPENERS = (
     "we need to",
     "first,",
     "first, the",
     "next,",
     "the user",
     "let me",
@@ -62,6 +64,11 @@ _REASONING_OPENERS = (
     "i should",
     "i recall",
 )
 def _normalize_extracted(text: str) -> str:
@@ -167,6 +174,14 @@ def looks_like_reasoning_only(text: str) -> bool:
     return bool(_SENTENCE_PART.search(text) and len(text) > 120)
 def needs_teacher_compaction(text: str) -> bool:
     cleaned = text.strip()
     if not cleaned:
@@ -203,12 +218,31 @@ def prepare_display_reply(text: str) -> str:
     return cleaned
 def strip_thinking_blocks(text: str) -> str:
     """Remove chain-of-thought wrapper tags; keep remaining text (e.g. JSON) intact."""
     cleaned = text.strip()
     if not cleaned:
         return ""
-    return _THINK_BLOCKS.sub("", cleaned).strip()
 def strip_reasoning_output(text: str) -> str:
@@ -218,6 +252,7 @@ def strip_reasoning_output(text: str) -> str:
         return ""
     cleaned = _THINK_BLOCKS.sub("", cleaned).strip()
     if cleaned and not _THINK_BLOCKS.search(text):
         extracted = _extract_best_answer(cleaned)
         if extracted:

 _LIST_OUTLINE = re.compile(r"^\d+\.\s", re.MULTILINE)
 _REASONING_OPENERS = (
     "we need to",
+    "we are given",
     "first,",
     "first, the",
+    "first, let's",
     "next,",
     "the user",
     "let me",
     "i should",
     "i recall",
 )
+_THINK_TAG_PAIRS = (
+    (_RT_OPEN, _RT_CLOSE),
+    (_THINK_OPEN, _THINK_CLOSE),
+    ("<thinking>", "</thinking>"),
+)
 def _normalize_extracted(text: str) -> str:
     return bool(_SENTENCE_PART.search(text) and len(text) > 120)
+def reply_ends_complete_sentence(text: str) -> bool:
+    """True when visible reply text ends with sentence-ending punctuation."""
+    cleaned = strip_reasoning_output(text).strip()
+    if not cleaned:
+        return False
+    return bool(_COMPLETE_SENTENCE.search(cleaned))
 def needs_teacher_compaction(text: str) -> bool:
     cleaned = text.strip()
     if not cleaned:
     return cleaned
+def _strip_unclosed_think_tags(text: str) -> str:
+    """Drop thinking blocks that never closed (common when generation hits max_tokens)."""
+    cleaned = text.strip()
+    if not cleaned:
+        return ""
+    for open_tag, close_tag in _THINK_TAG_PAIRS:
+        lower = cleaned.lower()
+        open_lower = open_tag.lower()
+        close_lower = close_tag.lower()
+        start = lower.find(open_lower)
+        if start == -1:
+            continue
+        if close_lower in lower[start + len(open_tag) :]:
+            continue
+        cleaned = cleaned[:start].strip()
+    return cleaned
 def strip_thinking_blocks(text: str) -> str:
     """Remove chain-of-thought wrapper tags; keep remaining text (e.g. JSON) intact."""
     cleaned = text.strip()
     if not cleaned:
         return ""
+    cleaned = _THINK_BLOCKS.sub("", cleaned).strip()
+    return _strip_unclosed_think_tags(cleaned)
 def strip_reasoning_output(text: str) -> str:
         return ""
     cleaned = _THINK_BLOCKS.sub("", cleaned).strip()
+    cleaned = _strip_unclosed_think_tags(cleaned)
     if cleaned and not _THINK_BLOCKS.search(text):
         extracted = _extract_best_answer(cleaned)
         if extracted:

libs/inference/tests/test_response_clean.py CHANGED Viewed

@@ -2,6 +2,7 @@ from __future__ import annotations
 from inference.response_clean import (
     prepare_display_reply,
     strip_reasoning_output,
     strip_thinking_blocks,
 )
@@ -17,6 +18,12 @@ def test_strips_redacted_thinking_block():
     assert strip_reasoning_output(raw) == "The capital of France is Paris."
 def test_strips_think_block():
     raw = f"{_THINK_OPEN}\nplanning...\n{_THINK_CLOSE}\n\nAgents use memory [1]."
     assert strip_reasoning_output(raw) == "Agents use memory [1]."
@@ -107,6 +114,11 @@ So, three"""
     assert "Sentence 1:" not in out
 def test_prepare_display_reply_wraps_malformed_think_prefix():
     raw = "think> We need to plan the answer.\n\nThe answer is 42."
     out = prepare_display_reply(raw)

 from inference.response_clean import (
     prepare_display_reply,
+    reply_ends_complete_sentence,
     strip_reasoning_output,
     strip_thinking_blocks,
 )
     assert strip_reasoning_output(raw) == "The capital of France is Paris."
+def test_strips_unclosed_redacted_thinking_block():
+    raw = f"{_RT_OPEN}\nplanning without a final answer that never closes"
+    assert strip_reasoning_output(raw) == ""
+    assert strip_thinking_blocks(raw) == ""
 def test_strips_think_block():
     raw = f"{_THINK_OPEN}\nplanning...\n{_THINK_CLOSE}\n\nAgents use memory [1]."
     assert strip_reasoning_output(raw) == "Agents use memory [1]."
     assert "Sentence 1:" not in out
+def test_reply_ends_complete_sentence():
+    assert reply_ends_complete_sentence("Finetuning teaches a small model to specialize.")
+    assert not reply_ends_complete_sentence("The lesson aims to teach how to fine-tune small")
 def test_prepare_display_reply_wraps_malformed_think_prefix():
     raw = "think> We need to plan the answer.\n\nThe answer is 42."
     out = prepare_display_reply(raw)

libs/researchmind/src/researchmind/citations.py CHANGED Viewed

@@ -89,4 +89,4 @@ def clean_model_answer(answer: str) -> str:
             "The model returned planning text without a final answer. "
             "Try asking again or switch to a non-reasoning model preset."
         )
-    return text

             "The model returned planning text without a final answer. "
             "Try asking again or switch to a non-reasoning model preset."
         )
+    return text

libs/researchmind/tests/test_citations.py CHANGED Viewed

@@ -65,3 +65,11 @@ def test_clean_model_answer_strips_thinking_block():
     raw = f"{think_open}\nplan\n{think_close}\n\nAgents use tools and memory [1]."
     cleaned = clean_model_answer(raw)
     assert cleaned == "Agents use tools and memory [1]."

     raw = f"{think_open}\nplan\n{think_close}\n\nAgents use tools and memory [1]."
     cleaned = clean_model_answer(raw)
     assert cleaned == "Agents use tools and memory [1]."
+def test_clean_model_answer_rejects_unclosed_thinking():
+    rt_open = "<" + "redacted_thinking" + ">"
+    raw = f"{rt_open}\nWe are given a context and need to plan the answer."
+    cleaned = clean_model_answer(raw)
+    assert "redacted_thinking" not in cleaned
+    assert "planning text without a final answer" in cleaned

scripts/benchmark_rag_chat.py ADDED Viewed

	@@ -0,0 +1,156 @@

+#!/usr/bin/env python3
+"""Benchmark ResearchMind RAG retrieval and optional full chat latency."""
+from __future__ import annotations
+import argparse
+import json
+import statistics
+import sys
+import time
+from pathlib import Path
+from researchmind.config import get_config
+from researchmind.embeddings import embed_texts
+from researchmind.ingest import IngestPipeline
+from researchmind.retrieve import retrieve
+def _load_sessions() -> list[tuple[str, str]]:
+    store = IngestPipeline().store
+    return [(s.id, s.topic or "Untitled") for s in store.list_sessions()]
+def benchmark_retrieve(
+    question: str,
+    *,
+    session_id: str,
+    runs: int,
+) -> dict[str, object]:
+    cfg = get_config()
+    store = IngestPipeline().store
+    chunks_in_scope = store.get_chunks_with_embeddings(session_id=session_id or None)
+    timings: list[float] = []
+    retrieved = 0
+    for _ in range(runs):
+        started = time.perf_counter()
+        chunks = retrieve(question, store, config=cfg, session_id=session_id or None)
+        timings.append((time.perf_counter() - started) * 1000)
+        retrieved = len(chunks)
+    warm = timings[1:] if len(timings) > 1 else timings
+    embed_started = time.perf_counter()
+    embed_texts(["warmup query"], model_name=cfg.embed_model)
+    embed_warm_ms = (time.perf_counter() - embed_started) * 1000
+    return {
+        "question": question,
+        "session_id": session_id,
+        "chunks_in_scope": len(chunks_in_scope),
+        "retrieved_chunks": retrieved,
+        "top_k": cfg.top_k,
+        "max_context_chunks": cfg.max_context_chunks,
+        "embed_model": cfg.embed_model,
+        "embedder_warm_ms": round(embed_warm_ms, 1),
+        "retrieve_ms_cold": round(timings[0], 1) if timings else 0.0,
+        "retrieve_ms_mean": round(statistics.mean(warm), 1),
+        "retrieve_ms_stdev": round(statistics.stdev(warm), 1) if len(warm) > 1 else 0.0,
+        "retrieve_ms_min": round(min(warm), 1),
+        "retrieve_ms_max": round(max(warm), 1),
+    }
+def benchmark_chat(
+    question: str,
+    *,
+    session_id: str,
+    model_key: str | None,
+) -> dict[str, object]:
+    from agent.runner import AgentRunner
+    from gradio_space.model_loading import ensure_model_loaded, get_active_model_key
+    from inference.factory import get_backend
+    key = model_key or get_active_model_key()
+    load_err = ensure_model_loaded(key)
+    if load_err:
+        return {"error": load_err, "model": key}
+    backend = get_backend(key)
+    runner = AgentRunner()
+    started = time.perf_counter()
+    result = runner.run_researchmind_chat(
+        question=question,
+        session_id=session_id,
+        model_key=key,
+        backend=backend,
+        doc_ids=None,
+    )
+    total_ms = (time.perf_counter() - started) * 1000
+    trace = json.loads(Path(result.trace_path).read_text(encoding="utf-8"))
+    steps = [
+        {
+            "name": step.get("name"),
+            "label": step.get("label"),
+            "duration_ms": step.get("duration_ms"),
+        }
+        for step in trace.get("steps", [])
+        if step.get("type") == "step"
+    ]
+    return {
+        "model": key,
+        "question": question,
+        "session_id": session_id,
+        "total_ms": round(total_ms, 1),
+        "citations": len(result.citations),
+        "answer_preview": result.answer[:240],
+        "steps": steps,
+        "trace_path": result.trace_path,
+    }
+def main() -> int:
+    parser = argparse.ArgumentParser(description="Benchmark ResearchMind RAG chat")
+    parser.add_argument(
+        "--question",
+        default="how we can finetune model",
+        help="Question to benchmark",
+    )
+    parser.add_argument("--session-id", default="", help="Research session id")
+    parser.add_argument("--runs", type=int, default=5, help="Retrieve benchmark repetitions")
+    parser.add_argument(
+        "--full-chat",
+        action="store_true",
+        help="Also run one full RAG chat (loads local LLM)",
+    )
+    parser.add_argument("--model-key", default="", help="Override ACTIVE_MODEL preset")
+    args = parser.parse_args()
+    sessions = _load_sessions()
+    session_id = args.session_id.strip()
+    if not session_id:
+        session_id = sessions[0][0] if sessions else ""
+    if not session_id:
+        print("No indexed session found. Ingest sources first.")
+        return 1
+    retrieve_report = benchmark_retrieve(
+        args.question,
+        session_id=session_id,
+        runs=max(1, args.runs),
+    )
+    print(json.dumps({"retrieve": retrieve_report}, indent=2))
+    if args.full_chat:
+        chat_report = benchmark_chat(
+            args.question,
+            session_id=session_id,
+            model_key=args.model_key or None,
+        )
+        print(json.dumps({"chat": chat_report}, indent=2))
+    return 0
+if __name__ == "__main__":
+    sys.exit(main())