modules/transcription/llm_client.py

@@ -1,365 +1,53 @@
 """Pluggable LLM client for transcription summaries.
 
-Phase 3: Full implementation with Anthropic, OpenAI, Ollama, OpenRouter, and Google providers.
+Phase 1: Stub implementation — returns TODO string.
+Phase 3: Wire up Anthropic, OpenAI, and Ollama providers.
 """
 
 import os
-import json
+from typing import Optional
-import logging
-from typing import Optional, Dict, Any
-
-import aiohttp
-
-from modules.transcription.prompts import PROMPT_TEMPLATES
-
-logger = logging.getLogger(__name__)
-
-# Default models per provider
-DEFAULT_MODELS: Dict[str, str] = {
-    "anthropic": "claude-sonnet-4-6",
-    "openai": "gpt-4o",
-    "ollama": "llama3",
-    "openrouter": "anthropic/claude-sonnet-4-6",
-    "google": "gemini-2.0-flash",
-}
-
-# Timeout for LLM calls (seconds)
-LLM_TIMEOUT = 120
-
-
-class LLMCallResult:
-    """Result from an LLM call, containing content and token usage."""
-
-    def __init__(self, content: str, input_tokens: Optional[int] = None,
-                 output_tokens: Optional[int] = None, raw_response: Optional[Dict[str, Any]] = None):
-        self.content = content
-        self.input_tokens = input_tokens
-        self.output_tokens = output_tokens
-        self.raw_response = raw_response
-
-    def to_dict(self) -> Dict[str, Any]:
-        return {
-            "content": self.content,
-            "input_tokens": self.input_tokens,
-            "output_tokens": self.output_tokens,
-        }
 
 
 async def call_llm(
     provider: str,
-    model: Optional[str] = None,
+    model: str,
-    api_key: str = "",
+    api_key: str,
-    system_prompt: str = "",
+    system_prompt: str,
-    user_message: str = "",
+    user_message: str,
-) -> LLMCallResult:
+) -> str:
     """Call an LLM to generate a summary.
 
-    Routes to the appropriate provider implementation.
+    Phase 1 stub — returns a TODO string.
+    Phase 3 will implement actual provider routing.
 
     Args:
         provider: 'anthropic', 'openai', 'ollama', 'openrouter', 'google'
-        model: Model name (falls back to provider default if None)
+        model: Model name (e.g. 'claude-sonnet-4-6', 'gpt-4o', 'llama3')
-        api_key: User's API key (from frontend, passed per-request)
+        api_key: User's API key (from localStorage, passed per-request)
         system_prompt: System prompt template (already filled with transcript)
         user_message: User message content
 
     Returns:
-        LLMCallResult with generated summary text and token counts
+        LLM-generated summary text
-
-    Raises:
-        ValueError: If provider is not supported
-        Exception: If the API call fails
     """
-    provider = provider.lower().strip()
+    # Phase 1 stub — TODO: implement in Phase 3
-    if model is None:
+    return f"[TODO: Implement LLM call for provider={provider}, model={model}]"
-        model = DEFAULT_MODELS.get(provider, "")
-
-    dispatch = {
-        "anthropic": call_anthropic,
-        "openai": call_openai,
-        "ollama": call_ollama,
-        "openrouter": call_openrouter,
-        "google": call_google,
-    }
-
-    if provider not in dispatch:
-        raise ValueError(
-            f"Unsupported provider: {provider}. "
-            f"Supported: {', '.join(dispatch.keys())}"
-        )
-
-    logger.info(f"Calling LLM provider={provider} model={model}")
-    result = await dispatch[provider](
-        api_key=api_key, model=model,
-        system_prompt=system_prompt, user_message=user_message,
-    )
-    logger.info(f"LLM call complete: provider={provider} tokens_in={result.input_tokens} tokens_out={result.output_tokens}")
-    return result
 
 
-# ---------------------------------------------------------------------------
+async def call_anthropic(api_key: str, model: str, system_prompt: str, user_message: str) -> str:
-# Provider implementations
+    """Call Anthropic Claude API."""
-# ---------------------------------------------------------------------------
+    # Phase 3 implementation placeholder
-
+    return f"[TODO: Anthropic call — model={model}]"
-async def call_anthropic(
-    api_key: str, model: str, system_prompt: str, user_message: str,
-) -> LLMCallResult:
-    """Call Anthropic Claude API (messages v2)."""
-    url = "https://api.anthropic.com/v1/messages"
-    headers = {
-        "x-api-key": api_key,
-        "anthropic-version": "2023-06-01",
-        "content-type": "application/json",
-    }
-    payload = {
-        "model": model,
-        "max_tokens": 4096,
-        "system": system_prompt,
-        "messages": [{"role": "user", "content": user_message}],
-    }
-
-    async with aiohttp.ClientSession() as session:
-        async with session.post(url, headers=headers, json=payload,
-                                timeout=aiohttp.ClientTimeout(total=LLM_TIMEOUT)) as resp:
-            if resp.status != 200:
-                body = await resp.text()
-                logger.error(f"Anthropic API error ({resp.status}): {body}")
-                raise Exception(f"Anthropic API error {resp.status}: {body}")
-
-            data = await resp.json()
-
-    # Extract content blocks
-    content_parts = []
-    for block in data.get("content", []):
-        if block.get("type") == "text":
-            content_parts.append(block["text"])
-
-    content = "\n".join(content_parts)
-
-    # Token counts from response
-    usage = data.get("usage", {})
-    input_tokens = usage.get("input_tokens") or usage.get("input_tokens")
-    output_tokens = usage.get("output_tokens") or usage.get("output_tokens")
-
-    # Anthropic v2 uses input_tokens/output_tokens; fall back to input_tokens/input_tokens
-    if not input_tokens:
-        input_tokens = usage.get("input_tokens")
-    if not output_tokens:
-        output_tokens = usage.get("output_tokens")
-
-    return LLMCallResult(
-        content=content,
-        input_tokens=input_tokens,
-        output_tokens=output_tokens,
-        raw_response=data,
-    )
 
 
-async def call_openai(
+async def call_openai(api_key: str, model: str, system_prompt: str, user_message: str) -> str:
-    api_key: str, model: str, system_prompt: str, user_message: str,
+    """Call OpenAI API."""
-) -> LLMCallResult:
+    # Phase 3 implementation placeholder
-    """Call OpenAI Chat Completions API."""
+    return f"[TODO: OpenAI call — model={model}]"
-    url = "https://api.openai.com/v1/chat/completions"
-    headers = {
-        "Authorization": f"Bearer {api_key}",
-        "Content-Type": "application/json",
-    }
-    payload = {
-        "model": model,
-        "messages": [
-            {"role": "system", "content": system_prompt},
-            {"role": "user", "content": user_message},
-        ],
-        "max_tokens": 4096,
-    }
-
-    async with aiohttp.ClientSession() as session:
-        async with session.post(url, headers=headers, json=payload,
-                                timeout=aiohttp.ClientTimeout(total=LLM_TIMEOUT)) as resp:
-            if resp.status != 200:
-                body = await resp.text()
-                logger.error(f"OpenAI API error ({resp.status}): {body}")
-                raise Exception(f"OpenAI API error {resp.status}: {body}")
-
-            data = await resp.json()
-
-    choice = data.get("choices", [{}])[0]
-    content = choice.get("message", {}).get("content", "")
-
-    usage = data.get("usage", {})
-    return LLMCallResult(
-        content=content,
-        input_tokens=usage.get("prompt_tokens"),
-        output_tokens=usage.get("completion_tokens"),
-        raw_response=data,
-    )
 
 
-async def call_ollama(
+async def call_ollama(api_key: str, model: str, system_prompt: str, user_message: str) -> str:
-    api_key: str, model: str, system_prompt: str, user_message: str,
+    """Call local Ollama instance."""
-) -> LLMCallResult:
+    # Phase 3 implementation placeholder
-    """Call local Ollama instance (generate endpoint)."""
+    ollama_url = os.getenv("OLLAMA_URL", "https://ollama.kevlarai.com")
-    ollama_url = os.getenv("OLLAMA_URL", "http://localhost:11434")
+    return f"[TODO: Ollama call — url={ollama_url}, model={model}]"
-    url = f"{ollama_url}/api/generate"
-
-    # Ollama uses a single prompt with system instructions prepended
-    full_prompt = f"{system_prompt}\n\n{user_message}"
-
-    payload = {
-        "model": model,
-        "prompt": full_prompt,
-        "stream": False,
-    }
-
-    headers = {"Content-Type": "application/json"}
-    # Ollama may not need an API key; include if set
-    if api_key:
-        headers["Authorization"] = f"Bearer {api_key}"
-
-    async with aiohttp.ClientSession() as session:
-        async with session.post(url, headers=headers, json=payload,
-                                timeout=aiohttp.ClientTimeout(total=LLM_TIMEOUT)) as resp:
-            if resp.status != 200:
-                body = await resp.text()
-                logger.error(f"Ollama API error ({resp.status}): {body}")
-                raise Exception(f"Ollama API error {resp.status}: {body}")
-
-            data = await resp.json()
-
-    content = data.get("response", "")
-    # Ollama reports total_tokens; split into input/output heuristically
-    total = data.get("total_tokens", 0)
-    prompt_tokens = data.get("prompt_eval_count", None)
-    eval_count = data.get("eval_count", None)
-
-    return LLMCallResult(
-        content=content,
-        input_tokens=prompt_tokens,
-        output_tokens=eval_count,
-        raw_response=data,
-    )
-
-
-async def call_openrouter(
-    api_key: str, model: str, system_prompt: str, user_message: str,
-) -> LLMCallResult:
-    """Call OpenRouter API (OpenAI-compatible chat completions)."""
-    url = "https://openrouter.ai/api/v1/chat/completions"
-    headers = {
-        "Authorization": f"Bearer {api_key}",
-        "Content-Type": "application/json",
-        "HTTP-Referer": os.getenv("APP_URL", "https://classroom-copilot.example.com"),
-        "X-Title": "Classroom Copilot",
-    }
-    payload = {
-        "model": model,
-        "messages": [
-            {"role": "system", "content": system_prompt},
-            {"role": "user", "content": user_message},
-        ],
-        "max_tokens": 4096,
-    }
-
-    async with aiohttp.ClientSession() as session:
-        async with session.post(url, headers=headers, json=payload,
-                                timeout=aiohttp.ClientTimeout(total=LLM_TIMEOUT)) as resp:
-            if resp.status != 200:
-                body = await resp.text()
-                logger.error(f"OpenRouter API error ({resp.status}): {body}")
-                raise Exception(f"OpenRouter API error {resp.status}: {body}")
-
-            data = await resp.json()
-
-    choice = data.get("choices", [{}])[0]
-    content = choice.get("message", {}).get("content", "")
-
-    usage = data.get("usage", {})
-    return LLMCallResult(
-        content=content,
-        input_tokens=usage.get("prompt_tokens"),
-        output_tokens=usage.get("completion_tokens"),
-        raw_response=data,
-    )
-
-
-async def call_google(
-    api_key: str, model: str, system_prompt: str, user_message: str,
-) -> LLMCallResult:
-    """Call Google Gemini API (generateContent)."""
-    url = f"https://generativelanguage.googleapis.com/v1beta/models/{model}:generateContent?key={api_key}"
-
-    payload = {
-        "contents": [
-            {
-                "role": "user",
-                "parts": [{"text": user_message}],
-            }
-        ],
-        "system_instruction": {
-            "parts": [{"text": system_prompt}],
-        },
-        "generationConfig": {
-            "maxOutputTokens": 4096,
-        },
-    }
-
-    headers = {"Content-Type": "application/json"}
-
-    async with aiohttp.ClientSession() as session:
-        async with session.post(url, headers=headers, json=payload,
-                                timeout=aiohttp.ClientTimeout(total=LLM_TIMEOUT)) as resp:
-            if resp.status != 200:
-                body = await resp.text()
-                logger.error(f"Google Gemini API error ({resp.status}): {body}")
-                raise Exception(f"Google Gemini API error {resp.status}: {body}")
-
-            data = await resp.json()
-
-    # Extract text from candidates
-    candidates = data.get("candidates", [])
-    if candidates:
-        content_parts = candidates[0].get("content", {}).get("parts", [])
-        content = "\n".join(p.get("text", "") for p in content_parts)
-    else:
-        content = ""
-
-    # Token usage from usage_metadata
-    usage = data.get("usageMetadata", {})
-    return LLMCallResult(
-        content=content,
-        input_tokens=usage.get("promptTokenCount"),
-        output_tokens=usage.get("candidatesTokenCount"),
-        raw_response=data,
-    )
-
-
-# ---------------------------------------------------------------------------
-# Helper: build prompt from template
-# ---------------------------------------------------------------------------
-
-def build_prompt(summary_type: str, transcript: str) -> tuple[str, str]:
-    """Build system + user prompt from template and transcript.
-
-    Args:
-        summary_type: One of 'full_lesson', 'questions_asked', 'teaching_style',
-                      'key_moments', 'segment'
-        transcript: The full (or segment) transcript text
-
-    Returns:
-        (system_prompt, user_message) tuple
-    """
-    template = PROMPT_TEMPLATES.get(summary_type, PROMPT_TEMPLATES["full_lesson"])
-    # The template has {transcript} placeholder — fill it in
-    filled = template.format(transcript=transcript)
-
-    # Split into system and user: everything before "Transcript:" is the system prompt,
-    # everything from "Transcript:" onward is the user message.
-    transcript_marker = "\n\nTranscript:\n"
-    if transcript_marker in filled:
-        system_prompt, user_message = filled.split(transcript_marker, 1)
-        user_message = "Transcript:\n" + user_message
-    else:
-        system_prompt = "You are an expert educational analyst."
-        user_message = filled
-
-    return system_prompt, user_message

routers/database/init/timetables.py

@@ -162,84 +162,13 @@ async def process_worker_timetable(file_content, user_node_data, worker_node_dat
     finally:
         logging.info(f"Closing driver for {worker_node_data['worker_db_name']}")
         driver.close_driver(neo_driver)
-
 @router.get("/current-period")
 async def get_current_period(user_id: str = ""):
-    """Get the current active timetable period for a teacher.
+    # Phase 1: return stub — TODO: implement Neo4j query in Phase 2
-    
+    return {
-    Queries Neo4j for Academic or Registration periods where now() falls
+        "period_id": None,
-    between start_time and end_time for the given teacher's uuid_string.
+        "event_type": None,
-    """
+        "event_label": None,
-    if not user_id:
+        "start_time": None,
-        return {
+        "end_time": None,
-            "period_id": None,
+    }
-            "event_type": None,
-            "event_label": None,
-            "start_time": None,
-            "end_time": None,
-        }
-    
-    # The user_id from Supabase JWT maps to the TeacherNode's uuid_string
-    teacher_uuid = user_id
-    
-    # Try to find the current period across all known databases
-    # First, try to get the user's database name from the UserNode
-    neo_driver = driver.get_driver()
-    if neo_driver is None:
-        logging.error("Failed to connect to Neo4j for current-period query")
-        return {
-            "period_id": None,
-            "event_type": None,
-            "event_label": None,
-            "start_time": None,
-            "end_time": None,
-        }
-    
-    try:
-        # Query for the current period
-        # Look for Teacher nodes with matching uuid_string that have relationships to Academic/Registration periods
-        query = """
-        MATCH (t:Teacher {uuid_string: $teacher_uuid})-[:HAS_PERIOD]->(p:AcademicPeriod|RegistrationPeriod)
-        WHERE p.start_time <= datetime() AND p.end_time >= datetime()
-        RETURN p.uuid_string AS period_id,
-               p.name AS event_label,
-               p.start_time AS start_time,
-               p.end_time AS end_time,
-               CASE WHEN p:AcademicPeriod THEN 'lesson' ELSE 'registration' END AS event_type
-        ORDER BY p.start_time ASC
-        LIMIT 1
-        """
-        
-        with neo_driver.session() as session:
-            result = session.run(query, teacher_uuid=teacher_uuid)
-            record = result.single()
-            
-            if record:
-                logging.info(f"Found current period for teacher {teacher_uuid}: {record['event_label']}")
-                return {
-                    "period_id": record["period_id"],
-                    "event_type": record["event_type"],
-                    "event_label": record["event_label"],
-                    "start_time": str(record["start_time"]) if record["start_time"] else None,
-                    "end_time": str(record["end_time"]) if record["end_time"] else None,
-                }
-            else:
-                logging.info(f"No current period found for teacher {teacher_uuid}")
-                return {
-                    "period_id": None,
-                    "event_type": None,
-                    "event_label": None,
-                    "start_time": None,
-                    "end_time": None,
-                }
-    except Exception as e:
-        logging.error(f"Error querying current period: {str(e)}")
-        return {
-            "period_id": None,
-            "event_type": None,
-            "event_label": None,
-            "start_time": None,
-            "end_time": None,
-        }
-    finally:
-        driver.close_driver(neo_driver)

routers/transcribe/sessions.py

@@ -1,13 +1,8 @@
 """Transcription sessions router — CRUD endpoints for transcription sessions and segments."""
 
 from fastapi import APIRouter, Depends, HTTPException, Query
-from fastapi.responses import FileResponse
 from typing import Optional, List
 from datetime import datetime
-import io
-import json
-import tempfile
-import os
 
 from modules.auth.supabase_bearer import SupabaseBearer
 from modules.transcription.models import (
@@ -21,10 +16,6 @@ from modules.transcription.models import (
     SummaryResponse,
     ExportFormat,
 )
-from modules.transcription.llm_client import call_llm, build_prompt
-import logging
-
-logger = logging.getLogger(__name__)
 
 router = APIRouter()
 
@@ -40,116 +31,6 @@ def get_user_id(credentials=Depends(SupabaseBearer())) -> str:
     return credentials.get("sub", credentials.get("user_id", ""))
 
 
-def seconds_to_srt_timestamp(seconds: float) -> str:
-    """Convert seconds to SRT timestamp format: HH:MM:SS,mmm"""
-    hours = int(seconds // 3600)
-    minutes = int((seconds % 3600) // 60)
-    secs = int(seconds % 60)
-    millis = int((seconds % 1) * 1000)
-    return f"{hours:02d}:{minutes:02d}:{secs:02d},{millis:03d}"
-
-
-def generate_srt(segments: List[dict]) -> str:
-    """Generate SRT (SubRip subtitle) content from segments."""
-    srt_entries = []
-    for idx, seg in enumerate(segments, start=1):
-        start_sec = float(seg.get("start_seconds", 0))
-        end_sec = float(seg.get("end_seconds", 0))
-        text = seg.get("text", "").strip()
-        if not text:
-            continue
-        start_ts = seconds_to_srt_timestamp(start_sec)
-        end_ts = seconds_to_srt_timestamp(end_sec)
-        # Clean text for SRT (no line breaks within a subtitle block)
-        clean_text = text.replace("\n", " ").strip()
-        srt_entries.append(f"{idx}\n{start_ts} --> {end_ts}\n{clean_text}")
-    return "\n\n".join(srt_entries) + "\n" if srt_entries else ""
-
-
-def generate_txt(segments: List[dict]) -> str:
-    """Generate plain text transcript with timestamps from segments."""
-    lines = []
-    for seg in segments:
-        start_sec = float(seg.get("start_seconds", 0))
-        text = seg.get("text", "").strip()
-        if not text:
-            continue
-        ts = seconds_to_srt_timestamp(start_sec)
-        lines.append(f"[{ts}] {text}")
-    return "\n".join(lines) + "\n" if lines else ""
-
-
-def generate_json_export(session: dict, segments: List[dict],
-                         summaries: List[dict],
-                         canvas_events: List[dict]) -> str:
-    """Generate structured JSON export with segments, metadata, and canvas events."""
-    # Build clean segment list (exclude internal DB fields)
-    clean_segments = []
-    for seg in segments:
-        clean_segments.append({
-            "sequence_index": seg.get("sequence_index"),
-            "text": seg.get("text", ""),
-            "start_seconds": float(seg.get("start_seconds", 0)),
-            "end_seconds": float(seg.get("end_seconds", 0)),
-            "is_final": seg.get("is_final", True),
-            "speaker_label": seg.get("speaker_label"),
-            "keyword_matches": seg.get("keyword_matches"),
-        })
-
-    # Build clean summary list
-    clean_summaries = []
-    for s in summaries:
-        clean_summaries.append({
-            "id": s.get("id"),
-            "summary_type": s.get("summary_type"),
-            "content": s.get("content", ""),
-            "llm_provider": s.get("llm_provider"),
-            "llm_model": s.get("llm_model"),
-            "created_at": s.get("created_at"),
-        })
-
-    # Build clean canvas events list
-    clean_events = []
-    for ev in canvas_events:
-        clean_events.append({
-            "id": ev.get("id"),
-            "event_type": ev.get("event_type"),
-            "session_elapsed_seconds": float(ev.get("session_elapsed_seconds", 0)) if ev.get("session_elapsed_seconds") else None,
-            "timestamp": ev.get("timestamp"),
-            "event_payload": ev.get("event_payload", {}),
-        })
-
-    export_data = {
-        "session": {
-            "id": session.get("id"),
-            "title": session.get("title"),
-            "canvas_type": session.get("canvas_type"),
-            "started_at": session.get("started_at"),
-            "ended_at": session.get("ended_at"),
-            "duration_seconds": session.get("duration_seconds"),
-            "timetable_period_id": session.get("timetable_period_id"),
-            "timetable_event_type": session.get("timetable_event_type"),
-            "timetable_event_label": session.get("timetable_event_label"),
-            "auto_tagged": session.get("auto_tagged", False),
-            "llm_provider": session.get("llm_provider"),
-            "llm_model": session.get("llm_model"),
-            "word_count": session.get("word_count", 0),
-            "segment_count": session.get("segment_count", 0),
-        },
-        "segments": clean_segments,
-        "summaries": clean_summaries,
-        "canvas_events": clean_events,
-    }
-
-    return json.dumps(export_data, indent=2, default=str)
-
-
-def sanitize_filename(name: str) -> str:
-    """Remove or replace characters that are unsafe in filenames."""
-    safe = "".join(c if c.isalnum() or c in " _-." else "_" for c in name)
-    return safe[:100] if safe else "export"
-
-
 @router.post("/sessions", response_model=TranscriptionSessionResponse)
 async def create_session(
     session_data: TranscriptionSessionCreate,
@@ -330,86 +211,28 @@ async def generate_summary(
     summary_request: SummaryGenerateRequest,
     user_id: str = Depends(get_user_id),
 ):
-    """Generate a summary for a session using the specified LLM provider.
+    """Generate a summary for a session (Phase 1 stub)."""
-
-    Phase 3: Full implementation — calls the pluggable LLM client with
-    prompt templates from prompts.py. API key is passed per-request and
-    never stored or logged.
-    """
     supabase = get_supabase_client()
     
     # Verify session exists and user owns it
-    session_check = supabase.supabase.table("transcription_sessions").select("*").eq("id", session_id).eq("user_id", user_id).execute()
+    session_check = supabase.supabase.table("transcription_sessions").select("id").eq("id", session_id).eq("user_id", user_id).execute()
     
     if not session_check.data:
         raise HTTPException(status_code=404, detail="Session not found")
     
-    session = session_check.data[0]
+    # Phase 1 stub: TODO implement LLM call in Phase 3
-    
+    content = "[TODO: Generate summary via LLM — provider={}, model={}]".format(
-    # Build transcript from segments (or use segment range)
+        summary_request.provider, summary_request.model
-    segments_query = supabase.supabase.table("transcription_segments").select("*").eq("session_id", session_id).order("sequence_index")
+    )
-    segments_result = segments_query.execute()
-    
-    if not segments_result.data:
-        raise HTTPException(status_code=400, detail="No segments found for this session")
-    
-    # Apply segment range filter if specified
-    segments = segments_result.data
-    if summary_request.segment_range and len(summary_request.segment_range) == 2:
-        start_idx, end_idx = summary_request.segment_range[0], summary_request.segment_range[1]
-        if start_idx is not None and end_idx is not None:
-            segments = segments[start_idx:end_idx]
-        elif start_idx is not None:
-            segments = segments[start_idx:]
-        elif end_idx is not None:
-            segments = segments[:end_idx]
-    
-    # Build full transcript text from segments
-    transcript_parts = [s["text"] for s in segments if s.get("text")]
-    transcript = "\n".join(transcript_parts)
-    
-    if not transcript.strip():
-        raise HTTPException(status_code=400, detail="Transcript is empty — cannot generate summary")
-    
-    # Build prompt from template
-    system_prompt, user_message = build_prompt(summary_request.summary_type, transcript)
-    
-    # Call the LLM client
-    try:
-        llm_result = await call_llm(
-            provider=summary_request.provider,
-            model=summary_request.model,
-            api_key=summary_request.api_key,
-            system_prompt=system_prompt,
-            user_message=user_message,
-        )
-    except Exception as e:
-        logger.error(f"LLM call failed: {e}")
-        raise HTTPException(status_code=502, detail=f"LLM generation failed: {str(e)}")
-    
-    # Determine segment range for storage
-    seg_start = None
-    seg_end = None
-    if summary_request.segment_range and len(summary_request.segment_range) == 2:
-        seg_start = summary_request.segment_range[0]
-        seg_end = summary_request.segment_range[1]
-    
-    # Build the prompt that was used (for audit trail)
-    prompt_used = f"{system_prompt}\n\n{user_message}" if summary_request.summary_type != "segment" else user_message
     
     # Save summary to database
     summary_data = {
         "session_id": session_id,
         "user_id": user_id,
         "summary_type": summary_request.summary_type,
-        "content": llm_result.content,
+        "content": content,
-        "prompt_used": prompt_used,
         "llm_provider": summary_request.provider,
         "llm_model": summary_request.model,
-        "input_tokens": llm_result.input_tokens,
-        "output_tokens": llm_result.output_tokens,
-        "segment_range_start": seg_start,
-        "segment_range_end": seg_end,
     }
     
     result = supabase.supabase.table("transcription_summaries").insert(summary_data).execute()
@@ -445,65 +268,26 @@ async def export_session(
     export_format: ExportFormat,
     user_id: str = Depends(get_user_id),
 ):
-    """Export session as SRT, TXT, or JSON file download.
+    """Export session as SRT, TXT, or JSON (Phase 1 stub)."""
-
-    Phase 3E: Full implementation — generates properly formatted files
-    and returns them as downloadable responses. API keys are never stored
-    or logged during export.
-    """
     supabase = get_supabase_client()
     
     # Verify ownership
-    session_check = supabase.supabase.table("transcription_sessions").select("*").eq("id", session_id).eq("user_id", user_id).execute()
+    session_check = supabase.supabase.table("transcription_sessions").select("id").eq("id", session_id).eq("user_id", user_id).execute()
     
     if not session_check.data:
         raise HTTPException(status_code=404, detail="Session not found")
     
-    session = session_check.data[0]
-    
     # Get segments
     segments_result = supabase.supabase.table("transcription_segments").select("*").eq("session_id", session_id).order("sequence_index").execute()
     segments = segments_result.data
     
-    # Get summaries (for JSON export)
+    if export_format.format == "srt":
-    summaries_result = supabase.supabase.table("transcription_summaries").select("*").eq("session_id", session_id).execute()
+        # Phase 1 stub — implement in Phase 3
-    summaries = summaries_result.data
+        return {"format": "srt", "content": "[TODO: Generate SRT from segments]"}
-    
+    elif export_format.format == "txt":
-    # Get canvas events (for JSON export)
+        text = "\n".join(s["text"] for s in segments)
-    canvas_result = supabase.supabase.table("canvas_events").select("*").eq("session_id", session_id).order("timestamp").execute()
+        return {"format": "txt", "content": text}
-    canvas_events = canvas_result.data
+    elif export_format.format == "json":
-    
+        return {"format": "json", "content": {"segments": segments}}
-    fmt = export_format.format.lower()
-    
-    if fmt == "srt":
-        content = generate_srt(segments)
-        filename = f"{sanitize_filename(session.get('title', session_id))}_{session.get('started_at', 'export')[:10]}.srt"
-        return FileResponse(
-            io.BytesIO(content.encode("utf-8")),
-            media_type="application/x-subrip",
-            filename=filename,
-            headers={"Content-Disposition": f"attachment; filename*=UTF-8''{filename}"},
-        )
-    
-    elif fmt == "txt":
-        content = generate_txt(segments)
-        filename = f"{sanitize_filename(session.get('title', session_id))}_{session.get('started_at', 'export')[:10]}.txt"
-        return FileResponse(
-            io.BytesIO(content.encode("utf-8")),
-            media_type="text/plain",
-            filename=filename,
-            headers={"Content-Disposition": f"attachment; filename*=UTF-8''{filename}"},
-        )
-    
-    elif fmt == "json":
-        content = generate_json_export(session, segments, summaries, canvas_events)
-        filename = f"{sanitize_filename(session.get('title', session_id))}_{session.get('started_at', 'export')[:10]}.json"
-        return FileResponse(
-            io.BytesIO(content.encode("utf-8")),
-            media_type="application/json",
-            filename=filename,
-            headers={"Content-Disposition": f"attachment; filename*=UTF-8''{filename}"},
-        )
-    
     else:
         raise HTTPException(status_code=400, detail=f"Unsupported format: {export_format.format}")