fix(agent): add wall-clock timeout for streaming LLM requests

nanobot/agent/runner.py

@@ -806,11 +806,17 @@ class AgentRunner:
         else:
             coro = spec.runtime.provider.chat_with_retry(**kwargs)
 
-        # Streaming requests already have provider-level idle timeouts
-        # (NANOBOT_STREAM_IDLE_TIMEOUT_S). Do not also apply the outer wall-clock
-        # LLM timeout here, or healthy long reasoning streams can be killed just
-        # because total elapsed time exceeded NANOBOT_LLM_TIMEOUT_S.
-        outer_timeout_s = None if (wants_streaming or wants_progress_streaming) else timeout_s
+        # Streaming requests also have provider-level idle timeouts
+        # (NANOBOT_STREAM_IDLE_TIMEOUT_S), but a stream that keeps producing
+        # very slow deltas can still run forever. Use a more generous wall-clock
+        # timeout for streaming while preserving NANOBOT_LLM_TIMEOUT_S=0 as an
+        # opt-out for all LLM wall-clock timeouts.
+        is_streaming_request = wants_streaming or wants_progress_streaming
+        outer_timeout_s = (
+            max(300.0, timeout_s * 2)
+            if is_streaming_request and timeout_s is not None
+            else timeout_s
+        )
         try:
             response = (
                 await coro if outer_timeout_s is None

tests/agent/test_runner_core.py

@@ -189,7 +189,7 @@ async def test_runner_times_out_hung_llm_request():
 
 
 @pytest.mark.asyncio
-async def test_runner_does_not_apply_outer_wall_timeout_to_streaming_requests():
+async def test_runner_applies_outer_wall_timeout_to_streaming_requests():
     from nanobot.agent.hook import AgentHook, AgentHookContext
     from nanobot.agent.runner import AgentRunner
 
@@ -216,8 +216,13 @@ async def test_runner_does_not_apply_outer_wall_timeout_to_streaming_requests():
             streamed.append(delta)
 
     runner = AgentRunner()
-    wait_for = AsyncMock(side_effect=AssertionError("streaming path must not use wait_for"))
-    with patch("nanobot.agent.runner.asyncio.wait_for", wait_for):
+    wait_for_calls: list[float] = []
+
+    async def fake_wait_for(coro, *, timeout):
+        wait_for_calls.append(timeout)
+        return await coro
+
+    with patch("nanobot.agent.runner.asyncio.wait_for", fake_wait_for):
         result = await runner.run(make_run_spec(provider,
             initial_messages=[{"role": "user", "content": "think for a while"}],
             tools=tools,
@@ -232,7 +237,47 @@ async def test_runner_does_not_apply_outer_wall_timeout_to_streaming_requests():
     assert result.final_content == "still alive"
     assert streamed == ["still ", "alive"]
     provider.chat_with_retry.assert_not_awaited()
-    wait_for.assert_not_awaited()
+    assert wait_for_calls == [300.0]
+
+
+@pytest.mark.asyncio
+async def test_runner_times_out_never_ending_streaming_request():
+    from nanobot.agent.hook import AgentHook
+    from nanobot.agent.runner import AgentRunner
+
+    provider = MagicMock(spec=LLMProvider)
+
+    async def chat_stream_with_retry(*, on_content_delta, **kwargs):
+        await asyncio.sleep(3600)
+
+    provider.chat_stream_with_retry = chat_stream_with_retry
+    provider.chat_with_retry = AsyncMock()
+    tools = MagicMock()
+    tools.get_definitions.return_value = []
+
+    class StreamingHook(AgentHook):
+        def wants_streaming(self) -> bool:
+            return True
+
+    async def fake_wait_for(coro, *, timeout):
+        coro.close()
+        raise asyncio.TimeoutError
+
+    runner = AgentRunner()
+    with patch("nanobot.agent.runner.asyncio.wait_for", fake_wait_for):
+        result = await runner.run(make_run_spec(provider,
+            initial_messages=[{"role": "user", "content": "think forever"}],
+            tools=tools,
+            model="test-model",
+            max_iterations=1,
+            max_tool_result_chars=_MAX_TOOL_RESULT_CHARS,
+            hook=StreamingHook(),
+            llm_timeout_s=200,
+        ))
+
+    assert result.stop_reason == "error"
+    assert result.final_content == "Error calling LLM: timed out after 400s"
+    provider.chat_with_retry.assert_not_awaited()
 
 
 @pytest.mark.asyncio