fix: append trailing user message so conversation summary calls don't 400

Root cause: get_messages() returns history ending with the just-saved
assistant reply, but the Anthropic Messages API rejects any request whose
last message is not role=user - this failed on every single chat turn in
production, silently swallowed by the existing try/except.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01V57jSQPqwkGG8BuAXg59X5

Claude outputs/main.py

@@ -1086,6 +1086,7 @@ async def update_conversation_summary(conversation_id: int):
     try:
         history = await get_messages(conversation_id)
         claude_messages = [{"role": m["role"], "content": m["content"]} for m in history]
+        claude_messages.append({"role": "user", "content": "Fasse das bisherige Gespraech zusammen."})
         completion = await asyncio.to_thread(
             claude_client.messages.create,
             model=CLAUDE_MODEL,

Claude outputs/tests/test_memory.py

@@ -237,6 +237,34 @@ async def test_update_conversation_summary_upserts_summary_and_embedding():
     mock_upsert.assert_called_once_with(3, "Nutzer fragte nach seinem Hund Bruno.", [0.1, 0.2])
 
 
+@pytest.mark.asyncio
+async def test_update_conversation_summary_ends_messages_with_user_role():
+    """Regression: history always ends with the just-saved assistant reply,
+    but the Anthropic Messages API rejects a request whose last message is
+    not role=user (no prefill support) - a trailing user instruction must
+    be appended, or this call 400s on every real conversation."""
+    history = [
+        {"role": "user", "content": "Wie geht es meinem Hund?", "tokens_used": None, "created_at": None},
+        {"role": "assistant", "content": "Bruno geht es gut.", "tokens_used": 5, "created_at": None},
+    ]
+    completion = MagicMock()
+    text_block = MagicMock()
+    text_block.type = "text"
+    text_block.text = "Zusammenfassung."
+    completion.content = [text_block]
+
+    main.claude_client = MagicMock()
+    main.claude_client.messages.create.return_value = completion
+
+    with patch.object(main, "get_messages", new=AsyncMock(return_value=history)), patch.object(
+        main, "get_embedding", new=AsyncMock(return_value=[0.1])
+    ), patch.object(main, "upsert_conversation_summary", new=AsyncMock()):
+        await main.update_conversation_summary(conversation_id=3)
+
+    _, kwargs = main.claude_client.messages.create.call_args
+    assert kwargs["messages"][-1]["role"] == "user"
+
+
 @pytest.mark.asyncio
 async def test_update_conversation_summary_swallows_exceptions():
     with patch.object(main, "get_messages", new=AsyncMock(side_effect=RuntimeError("DB down"))):