feat: inject remembered facts and similar summaries into chat system prompt

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

Claude outputs/main.py

@@ -731,8 +731,14 @@ async def execute_tool(name: str, tool_input: dict) -> str:
 MAX_TOOL_ROUNDS = 5
 
 
-async def run_chat_completion(claude_messages: list):
-    system_prompt = f"{CLAUDE_SYSTEM_PROMPT}\n\n{CALENDAR_ASSISTANT_INSTRUCTIONS}"
+async def run_chat_completion(claude_messages: list, conversation_id: int):
+    latest_user_message = claude_messages[-1]["content"]
+    memory_context = await build_memory_context(latest_user_message, conversation_id)
+
+    system_prompt = f"{CLAUDE_SYSTEM_PROMPT}\n\n{CALENDAR_ASSISTANT_INSTRUCTIONS}\n\n{MEMORY_ASSISTANT_INSTRUCTIONS}"
+    if memory_context:
+        system_prompt = f"{system_prompt}\n\n{memory_context}"
+
     messages = list(claude_messages)
     total_input = 0
     total_output = 0
@@ -743,7 +749,7 @@ async def run_chat_completion(claude_messages: list):
             model=CLAUDE_MODEL,
             max_tokens=1024,
             system=system_prompt,
-            tools=CALENDAR_TOOLS,
+            tools=CALENDAR_TOOLS + MEMORY_TOOLS,
             messages=messages,
         )
         total_input += completion.usage.input_tokens
@@ -900,7 +906,7 @@ async def chat(request: ChatRequest):
         claude_messages = [{"role": m["role"], "content": m["content"]} for m in history]
         claude_messages.append({"role": "user", "content": request.message})
 
-        response_text, output_tokens, tokens_used = await run_chat_completion(claude_messages)
+        response_text, output_tokens, tokens_used = await run_chat_completion(claude_messages, conversation_id)
 
         await save_message(conversation_id, DEFAULT_USER_ID, "assistant", response_text, output_tokens)
 

Claude outputs/tests/test_chat_tools.py

@@ -42,7 +42,10 @@ async def test_run_chat_completion_without_tool_use():
     main.claude_client = MagicMock()
     main.claude_client.messages.create.return_value = completion
 
-    text, output_tokens, total_tokens = await main.run_chat_completion([{"role": "user", "content": "Hi"}])
+    with patch.object(main, "build_memory_context", new=AsyncMock(return_value="")):
+        text, output_tokens, total_tokens = await main.run_chat_completion(
+            [{"role": "user", "content": "Hi"}], conversation_id=1
+        )
 
     assert text == "Hallo!"
     assert output_tokens == 5
@@ -64,9 +67,11 @@ async def test_run_chat_completion_executes_tool_and_returns_followup():
     main.claude_client = MagicMock()
     main.claude_client.messages.create.side_effect = [first, second]
 
-    with patch.object(main, "list_upcoming_events", new=AsyncMock(return_value=[])):
+    with patch.object(main, "list_upcoming_events", new=AsyncMock(return_value=[])), patch.object(
+        main, "build_memory_context", new=AsyncMock(return_value="")
+    ):
         text, output_tokens, total_tokens = await main.run_chat_completion(
-            [{"role": "user", "content": "Was steht diese Woche an?"}]
+            [{"role": "user", "content": "Was steht diese Woche an?"}], conversation_id=1
         )
 
     assert text == "Naechste Woche steht nichts an."
@@ -102,9 +107,9 @@ async def test_run_chat_completion_handles_two_sequential_tool_calls():
 
     with patch.object(main, "list_upcoming_events", new=AsyncMock(return_value=[{"uid": "abc"}])), patch.object(
         main, "update_event", new=AsyncMock(return_value={"uid": "abc"})
-    ):
+    ), patch.object(main, "build_memory_context", new=AsyncMock(return_value="")):
         text, output_tokens, total_tokens = await main.run_chat_completion(
-            [{"role": "user", "content": "Verschiebe den Termin X"}]
+            [{"role": "user", "content": "Verschiebe den Termin X"}], conversation_id=1
         )
 
     assert text == "Termin verschoben."
@@ -128,9 +133,11 @@ async def test_run_chat_completion_lists_recent_emails():
     main.claude_client = MagicMock()
     main.claude_client.messages.create.side_effect = [first, second]
 
-    with patch.object(main, "list_recent_emails", new=AsyncMock(return_value=[])):
+    with patch.object(main, "list_recent_emails", new=AsyncMock(return_value=[])), patch.object(
+        main, "build_memory_context", new=AsyncMock(return_value="")
+    ):
         text, output_tokens, total_tokens = await main.run_chat_completion(
-            [{"role": "user", "content": "Was ist neu im Postfach?"}]
+            [{"role": "user", "content": "Was ist neu im Postfach?"}], conversation_id=1
         )
 
     assert text == "Du hast 5 neue Mails."

Claude outputs/tests/test_memory.py

@@ -186,3 +186,29 @@ async def test_build_memory_context_skips_summaries_when_embedding_fails():
         context = await main.build_memory_context("Hallo", conversation_id=1)
 
     assert context == ""
+
+
+@pytest.mark.asyncio
+async def test_run_chat_completion_includes_memory_context_in_system_prompt():
+    completion = MagicMock()
+    completion.stop_reason = "end_turn"
+    text_block = MagicMock()
+    text_block.type = "text"
+    text_block.text = "Bruno geht es gut."
+    completion.content = [text_block]
+    usage = MagicMock()
+    usage.input_tokens = 10
+    usage.output_tokens = 5
+    completion.usage = usage
+
+    main.claude_client = MagicMock()
+    main.claude_client.messages.create.return_value = completion
+
+    with patch.object(
+        main, "build_memory_context", new=AsyncMock(return_value="Bekannte Fakten ueber den Nutzer:\n- Hund heisst Bruno")
+    ):
+        await main.run_chat_completion([{"role": "user", "content": "Wie geht es meinem Hund?"}], conversation_id=1)
+
+    _, kwargs = main.claude_client.messages.create.call_args
+    assert "Hund heisst Bruno" in kwargs["system"]
+    assert kwargs["tools"] == main.CALENDAR_TOOLS + main.MEMORY_TOOLS