Add conversation history summarization for long chats

When a conversation exceeds 20 messages (10 exchanges), the oldest
messages are summarized into a compact 3-5 sentence paragraph using the
intent model, and only the most recent 6 messages are passed verbatim.
The summary is injected into the system prompt so the model retains
context without the full token cost. For short conversations the check
is O(1) and returns immediately. The status indicator shows
"Summarizing conversation history..." when the LLM call is needed.

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

src/fabledassistant/services/generation_task.py

@@ -16,7 +16,7 @@ from fabledassistant.config import Config
 from fabledassistant.models import async_session
 from fabledassistant.models.conversation import Message
 from fabledassistant.services.generation_buffer import GenerationBuffer, GenerationState
-from fabledassistant.services.llm import ChatChunk, build_context, generate_completion, stream_chat, stream_chat_with_tools
+from fabledassistant.services.llm import ChatChunk, build_context, generate_completion, stream_chat, stream_chat_with_tools, summarize_history_for_context
 from fabledassistant.services.chat import update_conversation_title
 from fabledassistant.services.intent import IntentResult, classify_intent
 from fabledassistant.services.logging import log_generation
@@ -172,43 +172,55 @@ async def run_generation(
     MAX_TOOL_ROUNDS = 5
     msg_id = buf.assistant_message_id
 
-    # Phase 1: launch all independent work in parallel so nothing waits on anything
-    # unnecessarily. build_context (note search + system prompt) and the intent LLM
-    # call are the two slow legs — run them concurrently.
     buf.append_event("status", {"status": "Building context..."})
 
-    context_task = asyncio.create_task(build_context(
-        user_id, history, context_note_id, user_content, exclude_note_ids=exclude_note_ids
-    ))
-    tools_task = asyncio.create_task(get_tools_for_user(user_id))
-    intent_model_task = asyncio.create_task(get_setting(user_id, "intent_model", ""))
-
-    # Tools + intent-model setting are fast DB calls — get them first so intent
-    # can start immediately while build_context is still running.
-    tools, intent_model_setting = await asyncio.gather(tools_task, intent_model_task)
+    # Phase 1: Quick DB calls — resolve tools list and intent model in parallel.
+    tools, intent_model_setting = await asyncio.gather(
+        get_tools_for_user(user_id),
+        get_setting(user_id, "intent_model", ""),
+    )
     intent_model = intent_model_setting or Config.OLLAMA_INTENT_MODEL or model
     logger.info(
         "Starting generation for conv %d: model=%s, intent_model=%s, tools=%d",
         conv_id, model, intent_model, len(tools),
     )
 
-    # Start intent classification in parallel with remaining build_context work.
+    # Phase 2: Summarize long conversation history if needed.
+    # For short conversations (<= threshold) this returns immediately with no LLM call.
+    # For long conversations it spends ~500ms to compress old exchanges, which
+    # more than pays for itself by reducing the prefill token count every turn.
+    history_to_use = history
+    history_summary: str | None = None
+    if len(history) > 20:  # matches _HISTORY_SUMMARY_THRESHOLD in llm.py
+        buf.append_event("status", {"status": "Summarizing conversation history..."})
+        history_to_use, history_summary = await summarize_history_for_context(history, intent_model)
+
+    # Phase 3: Build context and classify intent in parallel — the two slow legs.
     pre_intent: IntentResult = IntentResult()
     intent_timing_ms: int | None = None
     if tools:
         intent_history = [
-            m for m in history
+            m for m in history_to_use
             if m.get("role") in ("user", "assistant") and m.get("content")
         ][-6:]
         buf.append_event("status", {"status": "Analyzing your request..."})
         t_intent = time.monotonic()
+        context_task = asyncio.create_task(build_context(
+            user_id, history_to_use, context_note_id, user_content,
+            exclude_note_ids=exclude_note_ids,
+            history_summary=history_summary,
+        ))
         intent_task = asyncio.create_task(
             classify_intent(user_content, tools, intent_model, history=intent_history)
         )
         (messages, context_meta), pre_intent = await asyncio.gather(context_task, intent_task)
         intent_timing_ms = int((time.monotonic() - t_intent) * 1000)
     else:
-        messages, context_meta = await context_task
+        messages, context_meta = await build_context(
+            user_id, history_to_use, context_note_id, user_content,
+            exclude_note_ids=exclude_note_ids,
+            history_summary=history_summary,
+        )
 
     # Emit context event
     buf.append_event("context", {"context": context_meta})