Generate conversation titles via LLM instead of truncating first line

- _maybe_set_title sets a quick truncated fallback immediately
- Then fires a background thread to ask the LLM for a 5-8 word summary
- Background thread doesn't block the streaming response
- Title appears in sidebar on first chunk, then upgrades when LLM responds

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

cheddahbot/agent.py

@@ -5,6 +5,7 @@ from __future__ import annotations
 import base64
 import json
 import logging
+import threading
 import uuid
 from collections.abc import Generator
 from pathlib import Path
@@ -294,19 +295,55 @@ class Agent:
                 self._memory.auto_flush(conv_id)
 
     def _maybe_set_title(self, conv_id: str, user_input: str):
-        """Set conversation title from first user message if still 'New Chat'."""
+        """Set conversation title from first user message if still 'New Chat'.
+
+        Sets a quick truncated fallback immediately, then fires a background
+        thread to generate a proper 5-8 word LLM summary.
+        """
         try:
             current_title = self.db.get_conversation_title(conv_id)
             if current_title and current_title != "New Chat":
                 return
-            title = user_input.split("\n", 1)[0].strip()
+            # Immediate fallback: first line, truncated
-            if len(title) > 50:
+            fallback = user_input.split("\n", 1)[0].strip()
-                title = title[:47] + "..."
+            if len(fallback) > 50:
-            if title:
+                fallback = fallback[:47] + "..."
-                self.db.update_conversation_title(conv_id, title)
+            if fallback:
+                self.db.update_conversation_title(conv_id, fallback)
+            # Fire background LLM call to generate a better title
+            threading.Thread(
+                target=self._generate_title,
+                args=(conv_id, user_input),
+                daemon=True,
+            ).start()
         except Exception as e:
             log.warning("Failed to set conversation title: %s", e)
 
+    def _generate_title(self, conv_id: str, user_input: str):
+        """Background: ask the LLM for a 5-8 word conversation title."""
+        try:
+            prompt = user_input[:500]  # cap input to keep it cheap
+            messages = [
+                {
+                    "role": "system",
+                    "content": (
+                        "Generate a short 5-8 word title summarizing this conversation opener. "
+                        "Reply with ONLY the title — no quotes, no punctuation at the end, no explanation."
+                    ),
+                },
+                {"role": "user", "content": prompt},
+            ]
+            parts = []
+            for chunk in self.llm.chat(messages, tools=None, stream=False):
+                if chunk.get("type") == "text":
+                    parts.append(chunk["content"])
+            title = "".join(parts).strip().strip('"').strip("'")
+            if title and len(title) <= 60:
+                self.db.update_conversation_title(conv_id, title)
+                log.debug("Generated conversation title: %s", title)
+        except Exception as e:
+            log.warning("Background title generation failed: %s", e)
+
     def respond_to_prompt(self, prompt: str) -> str:
         """Non-streaming response for scheduled tasks / internal use."""
         result_parts = []