Quick-capture notes: enrich content via main model

Add _process_note() — a second LLM pass using the main model that
transforms raw capture text into a well-formed note with a genuine
summary title and formatted body. Replaces the previous behaviour of
using the captured text verbatim as both title and body.

The processing prompt instructs the model to:
- Generate a 3-8 word summary title (never a verbatim copy)
- Format the body appropriately: bullet lists for items, clean prose
  for stream-of-thought, organised paragraphs for raw notes/fragments
- Preserve all original information without inventing new facts

The enrichment pass runs for both the intent-classified create_note
path and the fallback path. On LLM/parse failure it degrades safely
to the old verbatim behaviour.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

src/fabledassistant/routes/quick_capture.py

@@ -5,13 +5,16 @@ appropriate item (note, task, calendar event, todo) in a single synchronous
 request.  No SSE, no conversation ID, no streaming.
 """
 
+import json
 import logging
+import re
 
 from quart import Blueprint, jsonify, request
 
 from fabledassistant.auth import get_current_user_id, login_required
 from fabledassistant.config import Config
 from fabledassistant.services.intent import classify_capture_intent
+from fabledassistant.services.llm import generate_completion
 from fabledassistant.services.tools import execute_tool, get_tools_for_user
 
 logger = logging.getLogger(__name__)
@@ -22,6 +25,48 @@ quick_capture_bp = Blueprint("quick_capture", __name__, url_prefix="/api/quick-c
 # (delete_*) and read-only queries — worst-case fallback is a plain note.
 _CAPTURE_TOOL_NAMES = {"create_note", "create_task", "create_event", "update_note", "research_topic"}
 
+_NOTE_PROCESS_PROMPT = """\
+You are a note-taking assistant. The user has sent a quick-capture snippet. \
+Transform it into a well-formed note.
+
+Respond with ONLY a JSON object — no other text, no code fences:
+{{"title": "short descriptive title", "body": "note content in markdown"}}
+
+Rules:
+- title: 3–8 words, a genuine summary — do NOT copy the input verbatim
+- body: process the input thoughtfully:
+  - Lists of items → formatted bullet list
+  - A stream-of-thought or observation → clean prose, lightly organised
+  - Raw notes or fragments → organised paragraphs with a brief intro line
+  - URLs → include the URL and a one-sentence description of what it points to
+  - Preserve ALL information from the original; do not invent new facts
+  - Use markdown formatting (##, -, **, etc.) where it aids readability
+  - Keep it concise — do not pad with filler"""
+
+
+async def _process_note(text: str, model: str) -> tuple[str, str]:
+    """Use the main model to transform raw capture text into a title + body.
+
+    Returns (title, body). Falls back to (truncated text, full text) on any failure.
+    """
+    messages = [
+        {"role": "system", "content": _NOTE_PROCESS_PROMPT},
+        {"role": "user", "content": text},
+    ]
+    try:
+        raw = await generate_completion(messages, model, max_tokens=1024, num_ctx=4096)
+        raw = raw.strip()
+        raw = re.sub(r"^```(?:json)?\s*", "", raw)
+        raw = re.sub(r"\s*```$", "", raw).strip()
+        parsed = json.loads(raw)
+        title = str(parsed.get("title", "")).strip() or text[:60]
+        body = str(parsed.get("body", "")).strip() or text
+        return title, body
+    except Exception:
+        logger.warning("Note processing LLM call failed, using raw text", exc_info=True)
+        fallback_title = text if len(text) <= 80 else text[:77] + "..."
+        return fallback_title, text
+
 
 @quick_capture_bp.route("", methods=["POST"])
 @login_required
@@ -66,6 +111,14 @@ async def quick_capture_route():
                 logger.exception("Quick-capture research failed for topic: %s", topic)
                 return jsonify({"error": f"Research failed: {exc}"}), 500
 
+        # For notes, run a second LLM pass to generate a proper title and
+        # well-formed body rather than using the raw capture text verbatim.
+        if intent.tool_name == "create_note":
+            model = Config.OLLAMA_MODEL
+            title, body = await _process_note(text, model)
+            intent.arguments["title"] = title
+            intent.arguments["body"] = body
+
         result = await execute_tool(uid, intent.tool_name, intent.arguments)
         if result.get("success"):
             item_type = result.get("type", "note")
@@ -86,10 +139,10 @@ async def quick_capture_route():
         )
         # Fall through to plain-note fallback
 
-    # Fallback: create a plain note with the raw text as the body.
-    # Always preserve the full text as the body; derive a short title.
-    fallback_title = text if len(text) <= 80 else text[:77] + "..."
-    fallback_body = text
+    # Fallback: classify_capture_intent returned no-tool (e.g. LLM parse failure).
+    # Still process the text through the note enrichment pass.
+    model = Config.OLLAMA_MODEL
+    fallback_title, fallback_body = await _process_note(text, model)
 
     result = await execute_tool(
         uid, "create_note", {"title": fallback_title, "body": fallback_body}