Implement document title generation and UI improvements

- Add title generation functionality using selected LLM
- Generate title automatically during summarization process
- Display generated title in UI between diarization and summary sections
- Use generated title for naming exported transcript and summary files
- Separate LLM selection from summary section for better UX
- Update frontend to handle title events from API
- Add CSS styling for title display
- Update export models and services to include title information
- Sanitize titles for safe filename usage

Features:
- Automatic title generation from transcript content
- Title displayed prominently in results
- Exported files named with generated titles
- Improved UI organization with separated LLM controls

frontend/app.js

@@ -5,6 +5,7 @@ const state = {
   diarizedUtterances: null,
   diarizationStats: null,
   summary: '',
+  title: '',
   audioUrl: null,
   sourcePath: null,
   uploadedFile: null,
@@ -34,6 +35,7 @@ const elements = {
   transcriptTemplate: document.getElementById('utterance-template'),
   utteranceCount: document.getElementById('utterance-count'),
   summaryOutput: document.getElementById('summary-output'),
+  titleOutput: document.getElementById('title-output'),
   diarizationPanel: document.getElementById('diarization-summary'),
   diarizationMetrics: document.getElementById('diarization-metrics'),
   speakerBreakdown: document.getElementById('speaker-breakdown'),
@@ -492,11 +494,14 @@ async function handleSummaryGeneration() {
   state.summarizing = true;
   setStatus('Generating summary...', 'info');
   elements.summaryOutput.textContent = '';
+  elements.titleOutput.textContent = '';
+  state.title = '';
 
   const payload = {
     transcript: state.utterances.map((u) => u.text).join('\n'),
     llm_model: elements.llmSelect.value,
     prompt: elements.promptInput.value || 'Summarize the transcript below.',
+    generate_title: true,
   };
 
   try {
@@ -521,7 +526,10 @@ async function handleSummaryGeneration() {
       for (const line of lines) {
         if (!line.trim()) continue;
         const event = JSON.parse(line);
-        if (event.type === 'partial' && event.content) {
+        if (event.type === 'title' && event.content) {
+          state.title = event.content;
+          elements.titleOutput.textContent = event.content;
+        } else if (event.type === 'partial' && event.content) {
           elements.summaryOutput.innerHTML = renderMarkdown(event.content);
         }
       }
@@ -542,6 +550,7 @@ async function handleExportTranscript() {
     format: elements.transcriptFormat.value,
     include_timestamps: elements.includeTimestamps.checked,
     utterances: state.utterances,
+    title: state.title || null,
   };
   await downloadFile('/api/export/transcript', payload, 'transcript');
 }
@@ -552,6 +561,7 @@ async function handleExportSummary() {
     format: elements.summaryFormat.value,
     summary: elements.summaryOutput.textContent,
     metadata: {},
+    title: state.title || null,
   };
   await downloadFile('/api/export/summary', payload, 'summary');
 }

frontend/index.html

@@ -61,10 +61,13 @@
         </section>
 
         <section class="panel">
-          <h2>Summarization</h2>
+          <h2>Language Model</h2>
           <label for="llm-select">LLM Model</label>
           <select id="llm-select"></select>
+        </section>
 
+        <section class="panel">
+          <h2>Summarization</h2>
           <label for="prompt-input">Custom Prompt</label>
           <textarea id="prompt-input" rows="4">Summarize the transcript below.</textarea>
         </section>
@@ -145,6 +148,11 @@
             <div id="speaker-breakdown"></div>
           </section>
 
+          <section class="panel">
+            <h2>Document Title</h2>
+            <div id="title-output" class="title-display"></div>
+          </section>
+
           <section class="panel">
             <h2>Summary</h2>
             <div id="summary-output" class="summary"></div>

frontend/styles.css

@@ -472,6 +472,25 @@ button:hover {
   color: #93c5fd;
 }
 
+.title-display {
+  min-height: 40px;
+  background: rgba(15, 23, 42, 0.5);
+  border-radius: 12px;
+  padding: 1rem;
+  border: 1px solid rgba(148, 163, 184, 0.15);
+  font-size: 1.25rem;
+  font-weight: 600;
+  color: #f1f5f9;
+  text-align: center;
+}
+
+.title-display:empty::before {
+  content: "No title generated";
+  color: #64748b;
+  font-weight: 400;
+  font-style: italic;
+}
+
 .export-grid {
   display: grid;
   grid-template-columns: repeat(auto-fit, minmax(200px, 1fr));

src/server/models/export.py

@@ -16,9 +16,11 @@ class TranscriptExportRequest(BaseModel):
     format: str
     utterances: List[UtterancePayload]
     include_timestamps: bool = True
+    title: Optional[str] = None
 
 
 class SummaryExportRequest(BaseModel):
     format: str
     summary: str
     metadata: Optional[Dict[str, str]] = None
+    title: Optional[str] = None

src/server/models/summarization.py

@@ -7,3 +7,4 @@ class SummaryRequest(BaseModel):
     transcript: str = Field(..., min_length=1)
     llm_model: str
     prompt: str = Field("Summarize the transcript below.")
+    generate_title: bool = Field(default=True)

src/server/services/export_service.py

@@ -2,6 +2,7 @@ from __future__ import annotations
 
 from datetime import datetime
 from typing import Tuple
+import re
 
 from src.export_utils import (
     SUBTITLE_FORMATS,
@@ -13,6 +14,18 @@ from src.export_utils import (
 from ..models.export import SummaryExportRequest, TranscriptExportRequest
 
 
+def _sanitize_filename(title: str) -> str:
+    """Sanitize title for use in filename"""
+    if not title:
+        return ""
+    # Remove or replace invalid filename characters
+    sanitized = re.sub(r'[<>:"/\\|?*]', '', title)
+    # Replace spaces and other characters with underscores
+    sanitized = re.sub(r'[^\w\-_\.]', '_', sanitized)
+    # Limit length
+    return sanitized[:50] if sanitized else ""
+
+
 def _build_utterance_tuples(payload: TranscriptExportRequest):
     utterances = [(u.start, u.end, u.text) for u in payload.utterances]
     has_speakers = any(u.speaker is not None for u in payload.utterances)
@@ -45,7 +58,8 @@ def generate_transcript_export(payload: TranscriptExportRequest) -> Tuple[str, s
         raise ValueError(f"Unsupported transcript export format: {payload.format}")
 
     timestamp = datetime.now().strftime("%Y%m%d_%H%M%S")
-    filename = f"transcript_{timestamp}{fmt['extension']}"
+    title_part = f"_{_sanitize_filename(payload.title)}" if payload.title else ""
+    filename = f"transcript{title_part}_{timestamp}{fmt['extension']}"
     return content, filename, fmt["mime_type"]
 
 
@@ -56,5 +70,6 @@ def generate_summary_export(payload: SummaryExportRequest) -> Tuple[str, str, st
     fmt = SUMMARY_FORMATS[payload.format]
     content = fmt["function"](payload.summary, payload.metadata)
     timestamp = datetime.now().strftime("%Y%m%d_%H%M%S")
-    filename = f"summary_{timestamp}{fmt['extension']}"
+    title_part = f"_{_sanitize_filename(payload.title)}" if payload.title else ""
+    filename = f"summary{title_part}_{timestamp}{fmt['extension']}"
     return content, filename, fmt["mime_type"]

src/server/services/summarization_service.py

@@ -4,13 +4,20 @@ from typing import Dict, Iterable
 
 from fastapi import HTTPException
 
-from src.summarization import summarize_transcript
+from src.summarization import summarize_transcript, generate_title
 
 from ..models.summarization import SummaryRequest
 
 
 def iter_summary_events(payload: SummaryRequest) -> Iterable[Dict[str, str]]:
     try:
+        # Generate title if requested
+        title = None
+        if payload.generate_title:
+            title = generate_title(payload.transcript, payload.llm_model)
+            yield {"type": "title", "content": title}
+
+        # Generate summary
         generator = summarize_transcript(
             transcript=payload.transcript,
             selected_gguf_model=payload.llm_model,

src/summarization.py

@@ -204,6 +204,62 @@ def summarize_transcript_langchain(transcript: str, selected_gguf_model: str, pr
         yield f"[Error during summarization: {str(e)}]"
 
 
+def create_title_prompt() -> PromptTemplate:
+    """Prompt for generating a document title"""
+    template = """Generate a concise, descriptive title for this transcript. The title should capture the main topic or theme.
+
+Transcript:
+{text}
+
+Title:"""
+    return PromptTemplate(template=template, input_variables=["text"])
+
+
+def generate_title(transcript: str, selected_gguf_model: str) -> str:
+    """
+    Generate a title for the transcript using the selected LLM.
+    Returns a concise title that captures the main topic.
+    """
+    if not transcript or not transcript.strip():
+        return "Untitled Document"
+
+    try:
+        # Get the LLM
+        llm = get_llm(selected_gguf_model)
+        title_prompt = create_title_prompt()
+
+        # Use first 2000 tokens for title generation to avoid excessive context
+        tokens = llm.tokenize(transcript.encode('utf-8'))
+        if len(tokens) > 2000:
+            # Truncate to first 2000 tokens and decode back to text
+            truncated_tokens = tokens[:2000]
+            truncated_text = llm.detokenize(truncated_tokens).decode('utf-8')
+        else:
+            truncated_text = transcript
+
+        # Format the prompt
+        formatted_prompt = title_prompt.format(text=truncated_text)
+
+        # Generate title
+        response = llm.create_chat_completion(
+            messages=[
+                {"role": "system", "content": "You are an expert at creating concise, descriptive titles for documents and transcripts."},
+                {"role": "user", "content": formatted_prompt}
+            ],
+            stream=False,
+            max_tokens=50,  # Limit title length
+        )
+
+        title = response['choices'][0]['message']['content'].strip()
+        # Clean up the title (remove quotes, extra whitespace)
+        title = title.strip('"\'').strip()
+        return title if title else "Untitled Document"
+
+    except Exception as e:
+        print(f"Error generating title: {e}")
+        return "Untitled Document"
+
+
 # Alias pour maintenir la compatibilité
 summarize_transcript = summarize_transcript_langchain