fix(chat): stabilize system prompt, sequence memory extraction, and send stable session id to preserve KV cache (#3360)

* fix(chat): stabilize system prompt, sequence memory extraction, send stable session id to preserve KV cache

Fixes #2927. As diagnosed in the issue, three things in Odysseus's request
pattern actively destroyed local backends' (llama.cpp / LM Studio) KV-cache
continuity, forcing a full prompt re-evaluation (15-30s+) on every turn:

1. Dynamic content folded into the system prompt every turn. Both the chat
   preface (ChatProcessor.build_context_preface) and the agent system prompt
   (_build_system_prompt) injected current_datetime_prompt() — text that
   changes every minute — directly into system-role messages, which llm_core
   then concatenates into the single system message sent as the cached
   prefix. Any byte difference there invalidates the entire cache. Moved this
   to a new current_datetime_context_message() helper that returns a
   standalone user-role message, inserted near the end of the array (right
   before the latest user turn) instead of mixed into the system prompt. The
   static system prefix (preset prompt + safety policy + agent base prompt)
   now stays byte-identical across turns of the same session.

2. Memory/skill extraction side-requests competed with the main completion.
   run_post_response_tasks fired extract_and_store / maybe_extract_skill via
   asyncio.create_task — fire-and-forget coroutines that could overlap the
   next turn's main request and steal llama.cpp's limited processing slots,
   evicting the cached checkpoint. They're now queued through a new
   _run_extraction_jobs_sequentially helper that waits for the session's
   stream to go idle and runs the jobs strictly one at a time.

3. No stable session identifier was sent to local backends, so llama.cpp
   assigned a new processing slot via LRU every turn ("session_id=<empty>
   server-selected (LCP/LRU)"), losing slot affinity. Added
   _apply_local_cache_affinity() in llm_core, which sets session_id and
   cache_prompt: true on outgoing payloads — gated to self-hosted
   OpenAI-compatible endpoints only (never api.openai.com or other cloud
   providers, which reject unrecognized request fields with a 400). Threaded
   session_id through stream_llm / llm_call_async / stream_agent_loop from
   the existing Odysseus session id.

Tests in tests/test_kv_cache_invalidation_2927.py exercise the real payload-
assembly and scheduling code paths: byte-identical system prefix across two
turns of the same session (with a regression check that genuinely changed
instructions DO still change it), the dynamic time block landing as a
user-role message, extraction jobs waiting for the stream to go idle and
running sequentially, and the outgoing payload carrying a stable session_id
(same across turns of one session, different across sessions) only for
self-hosted endpoints. Updated tests/test_user_time.py for the new message
placement.

* fix(tests): accept owner= kwarg in normalize_model_id monkeypatch

The upstream normalize_model_id signature now takes an owner= keyword
argument, and chat_helpers.py passes owner=getattr(sess, "owner", None)
at the call site. Update the test stub lambda to **kwargs so it handles
the new argument without breaking, and update chat_helpers.py to forward
the owner parameter consistently.

---------

Co-authored-by: Alexandre Teixeira <111787685+alteixeira20@users.noreply.github.com>

src/agent_loop.py

@@ -890,9 +890,20 @@ def _build_system_prompt(
 
     # Current date/time for every agent request. This is user-local when the
     # browser provided timezone headers, with a server-local fallback.
+    #
+    # IMPORTANT: this is intentionally NOT prepended into agent_prompt (the
+    # system message) anymore. Its text changes every minute, and local
+    # OpenAI-compatible backends (llama.cpp / LM Studio) key their KV-cache
+    # prefix off the system message byte-for-byte — mixing ever-changing
+    # timestamp text into the (already large, tool-laden) agent system prompt
+    # would invalidate the cached prefix on every single request, forcing a
+    # full prompt re-evaluation each turn (issue #2927). It's built here as a
+    # standalone *user*-role message and inserted near the end of the array,
+    # right alongside _doc_message / _skills_message, below.
+    _datetime_message = None
     try:
-        from src.user_time import current_datetime_prompt
-        agent_prompt = current_datetime_prompt() + agent_prompt
+        from src.user_time import current_datetime_context_message
+        _datetime_message = current_datetime_context_message()
     except Exception:
         pass
 
@@ -1229,6 +1240,9 @@ def _build_system_prompt(
         last_user_idx += 1  # the document message is now at last_user_idx
     if _skills_message:
         merged.insert(last_user_idx, _skills_message)
+        last_user_idx += 1
+    if _datetime_message:
+        merged.insert(last_user_idx, _datetime_message)
 
     return merged, mcp_schemas
 
@@ -2158,6 +2172,7 @@ async def stream_agent_loop(
             prompt_type=prompt_type if round_num == 1 else None,
             tools=all_tool_schemas if all_tool_schemas else None,
             timeout=agent_stream_timeout,
+            session_id=session_id,
         ):
             if time.time() > _round_deadline:
                 logger.warning(f"[agent] round {round_num} stream exceeded wall-clock deadline; cutting off")