Merge origin/main into feat/session-replay-file-cap-invariants

Preserve main's timestamp/tool-context replay semantics while keeping the PR's session history and file-cap budgets.

Made-with: Cursor

tests/agent/test_auto_compact.py

@@ -178,7 +178,11 @@ class TestAgentLoopTTLParam:
             content="hello",
         )
         await loop._process_message(msg)
-        session.get_history.assert_called_once_with(max_messages=7, max_tokens=333)
+        session.get_history.assert_called_once_with(
+            max_messages=7,
+            max_tokens=333,
+            include_timestamps=True,
+        )
 
     @pytest.mark.asyncio
     async def test_session_file_cap_archives_and_trims_old_messages(self, tmp_path):

tests/agent/test_context_prompt_cache.py

@@ -188,6 +188,17 @@ def test_identity_has_no_behavioral_instructions(tmp_path) -> None:
     assert "Execution Rules" not in identity
 
 
+def test_system_prompt_does_not_warn_about_message_time_markers(tmp_path) -> None:
+    """Parroting is prevented by not annotating assistant turns in history;
+    no prompt-level warning about ``[Message Time: ...]`` is needed."""
+    workspace = _make_workspace(tmp_path)
+    builder = ContextBuilder(workspace)
+
+    prompt = builder.build_system_prompt()
+
+    assert "Message Time" not in prompt
+
+
 def test_default_soul_template_contains_execution_rules() -> None:
     """Default SOUL.md template must contain execution rules with act/plan layering."""
     soul = (pkg_files("nanobot") / "templates" / "SOUL.md").read_text(encoding="utf-8")

tests/agent/test_loop_save_turn.py

@@ -535,7 +535,14 @@ async def test_system_subagent_followup_is_persisted_before_prompt_assembly(tmp_
     )
 
     non_system = [m for m in seen["initial_messages"] if m.get("role") != "system"]
-    assert [m["content"] for m in non_system[:2]] == ["question", "working"]
+    assert "question" in non_system[0]["content"]
+    assert "working" in non_system[1]["content"]
+    # User turns carry the timestamp prefix so the model can reason about
+    # relative time. Assistant turns do NOT, otherwise the model treats those
+    # past replies as in-context examples and starts its own outputs with
+    # ``[Message Time: ...]`` (which then leaks back to the user).
+    assert "[Message Time:" in non_system[0]["content"]
+    assert "[Message Time:" not in non_system[1]["content"]
     assert non_system[2]["content"].count("subagent result") == 1
     assert "Current Time:" in non_system[2]["content"]
 
@@ -657,3 +664,63 @@ def test_subagent_followup_skips_empty_content() -> None:
 
     assert loop._persist_subagent_followup(session, msg) is False
     assert session.messages == []
+
+
+def test_set_tool_context_passes_thread_session_key_to_spawn(tmp_path: Path) -> None:
+    loop = _make_full_loop(tmp_path)
+
+    loop._set_tool_context(
+        "slack",
+        "C123",
+        metadata={"slack": {"thread_ts": "1700.42", "channel_type": "channel"}},
+        session_key="slack:C123:1700.42",
+    )
+
+    spawn_tool = loop.tools.get("spawn")
+    assert spawn_tool is not None
+    assert spawn_tool._session_key.get() == "slack:C123:1700.42"
+
+
+@pytest.mark.asyncio
+async def test_system_subagent_followup_uses_thread_session_and_slack_metadata(tmp_path: Path) -> None:
+    loop = _make_full_loop(tmp_path)
+    loop.consolidator.maybe_consolidate_by_tokens = AsyncMock(return_value=False)  # type: ignore[method-assign]
+
+    thread_session = loop.sessions.get_or_create("slack:C123:1700.42")
+    thread_session.add_message("user", "thread question")
+    loop.sessions.save(thread_session)
+
+    seen: dict[str, list[dict]] = {}
+
+    async def fake_run_agent_loop(initial_messages, **_kwargs):
+        seen["initial_messages"] = initial_messages
+        return (
+            "done",
+            [],
+            [*initial_messages, {"role": "assistant", "content": "done"}],
+            "stop",
+            False,
+        )
+
+    loop._run_agent_loop = fake_run_agent_loop  # type: ignore[method-assign]
+
+    outbound = await loop._process_message(
+        InboundMessage(
+            channel="system",
+            sender_id="subagent",
+            chat_id="slack:C123",
+            content="subagent result",
+            session_key_override="slack:C123:1700.42",
+            metadata={"subagent_task_id": "sub-1"},
+        )
+    )
+
+    assert outbound is not None
+    assert outbound.channel == "slack"
+    assert outbound.chat_id == "C123"
+    assert outbound.metadata == {"slack": {"thread_ts": "1700.42"}}
+    assert "thread question" in seen["initial_messages"][1]["content"]
+
+    loop.sessions.invalidate("slack:C123:1700.42")
+    persisted = loop.sessions.get_or_create("slack:C123:1700.42")
+    assert any(m.get("subagent_task_id") == "sub-1" for m in persisted.messages)

tests/agent/test_loop_tool_context.py

@@ -0,0 +1,90 @@
+from pathlib import Path
+from unittest.mock import MagicMock
+
+import pytest
+
+from nanobot.agent.loop import AgentLoop
+from nanobot.bus.queue import MessageBus
+from nanobot.providers.base import LLMResponse, ToolCallRequest
+
+
+class _ContextRecordingTool:
+    name = "cron"
+    concurrency_safe = False
+
+    def __init__(self) -> None:
+        self.contexts: list[dict] = []
+
+    def set_context(
+        self,
+        channel: str,
+        chat_id: str,
+        metadata: dict | None = None,
+        session_key: str | None = None,
+    ) -> None:
+        self.contexts.append({
+            "channel": channel,
+            "chat_id": chat_id,
+            "metadata": metadata,
+            "session_key": session_key,
+        })
+
+    async def execute(self, **_kwargs) -> str:
+        return "created"
+
+
+class _Tools:
+    def __init__(self, tool: _ContextRecordingTool) -> None:
+        self.tool = tool
+
+    def get(self, name: str):
+        return self.tool if name == "cron" else None
+
+    def get_definitions(self) -> list:
+        return []
+
+    def prepare_call(self, name: str, arguments: dict):
+        return (self.tool, arguments, None) if name == "cron" else (None, arguments, None)
+
+
+@pytest.mark.asyncio
+async def test_loop_hook_preserves_metadata_when_resetting_tool_context(tmp_path: Path) -> None:
+    provider = MagicMock()
+    calls = {"n": 0}
+
+    async def chat_with_retry(**_kwargs):
+        calls["n"] += 1
+        if calls["n"] == 1:
+            return LLMResponse(
+                content=None,
+                tool_calls=[ToolCallRequest(id="call_1", name="cron", arguments={"action": "add"})],
+            )
+        return LLMResponse(content="done", tool_calls=[])
+
+    provider.chat_with_retry = chat_with_retry
+    provider.get_default_model.return_value = "test-model"
+
+    loop = AgentLoop(
+        bus=MessageBus(),
+        provider=provider,
+        workspace=tmp_path,
+        model="test-model",
+    )
+    cron = _ContextRecordingTool()
+    loop.tools = _Tools(cron)
+
+    metadata = {"slack": {"thread_ts": "111.222", "channel_type": "channel"}}
+    await loop._run_agent_loop(
+        [],
+        channel="slack",
+        chat_id="C123",
+        metadata=metadata,
+        session_key="slack:C123:111.222",
+    )
+
+    assert cron.contexts[-1] == {
+        "channel": "slack",
+        "chat_id": "C123",
+        "metadata": metadata,
+        "session_key": "slack:C123:111.222",
+    }

tests/agent/test_runner.py

@@ -1060,11 +1060,10 @@ async def test_next_turn_after_llm_error_keeps_turn_boundary(tmp_path):
 
     request_messages = provider.chat_with_retry.await_args_list[1].kwargs["messages"]
     non_system = [message for message in request_messages if message.get("role") != "system"]
-    assert non_system[0] == {"role": "user", "content": "first question"}
-    assert non_system[1] == {
-        "role": "assistant",
-        "content": _PERSISTED_MODEL_ERROR_PLACEHOLDER,
-    }
+    assert non_system[0]["role"] == "user"
+    assert "first question" in non_system[0]["content"]
+    assert non_system[1]["role"] == "assistant"
+    assert _PERSISTED_MODEL_ERROR_PLACEHOLDER in non_system[1]["content"]
     assert non_system[2]["role"] == "user"
     assert "second question" in non_system[2]["content"]
 

tests/agent/test_session_manager_history.py

@@ -194,6 +194,87 @@ def test_get_history_preserves_reasoning_content():
     ]
 
 
+def test_get_history_annotates_user_turns_but_not_assistant_turns():
+    """Only user turns carry the timestamp prefix.
+
+    Annotating assistant turns trains the model (via in-context examples) to
+    start its own replies with ``[Message Time: ...]``. User-side stamps are
+    enough to pin adjacent assistant replies for relative-time reasoning.
+    """
+    session = Session(key="test:timestamps")
+    session.messages.append({
+        "role": "user",
+        "content": "10 点提醒是昨天发生的",
+        "timestamp": "2026-04-26T22:00:00",
+    })
+    session.messages.append({
+        "role": "assistant",
+        "content": "记下来了",
+        "timestamp": "2026-04-26T22:00:05",
+    })
+
+    history = session.get_history(max_messages=500, include_timestamps=True)
+
+    assert history == [
+        {
+            "role": "user",
+            "content": "[Message Time: 2026-04-26T22:00:00]\n10 点提醒是昨天发生的",
+        },
+        {
+            "role": "assistant",
+            "content": "记下来了",
+        },
+    ]
+
+
+def test_get_history_annotates_proactive_assistant_deliveries_with_timestamps():
+    """Cron / heartbeat assistant pushes still carry a timestamp prefix.
+
+    These proactive deliveries can sit hours away from the next user reply,
+    so the model needs to know when they fired. They are rare enough that
+    they don't act as in-context demonstrations encouraging the model to
+    prefix its own normal replies with ``[Message Time: ...]``.
+    """
+    session = Session(key="test:proactive-timestamps")
+    session.messages.append({
+        "role": "assistant",
+        "content": "记得喝水",
+        "timestamp": "2026-04-26T15:00:00",
+        "_channel_delivery": True,
+    })
+    session.messages.append({
+        "role": "user",
+        "content": "好",
+        "timestamp": "2026-04-26T18:00:00",
+    })
+
+    history = session.get_history(max_messages=500, include_timestamps=True)
+
+    assert history == [
+        {
+            "role": "assistant",
+            "content": "[Message Time: 2026-04-26T15:00:00]\n记得喝水",
+        },
+        {
+            "role": "user",
+            "content": "[Message Time: 2026-04-26T18:00:00]\n好",
+        },
+    ]
+
+
+def test_get_history_does_not_annotate_tool_results_with_timestamps():
+    session = Session(key="test:tool-timestamps")
+    session.messages.append({"role": "user", "content": "run tool"})
+    session.messages.extend(_tool_turn("ts", 0))
+    session.messages[-1]["timestamp"] = "2026-04-26T22:00:10"
+
+    history = session.get_history(max_messages=500, include_timestamps=True)
+
+    tool_result = history[-1]
+    assert tool_result["role"] == "tool"
+    assert tool_result["content"] == "ok"
+
+
 # --- Window cuts mid-group: assistant present but some tool results orphaned ---
 
 def test_window_cuts_mid_tool_group():