datawhalechina/hello-agents · error · TimeoutException
LLM思考超时
Error message
LLM思考超时
What it means
BaseAgent.think wraps self.llm.ainvoke(full_prompt) in asyncio.wait_for(..., timeout=self.timeout); asyncio.TimeoutError is re-raised as a custom TimeoutException('LLM思考超时'). It means the LLM call did not complete within the agent's configured timeout (default set on the agent, often 60s), not that the model returned an error.
Source
Thrown at Co-creation-projects/Shawnxyxy-HealthRecordAgent/backend/agents/base.py:143
self.trace("LLM TTHINKING TIME",
{
"duration_sec": duration,
"prompt_tokens": len(full_prompt),
}
)
response_text = response.content if hasattr(response, 'content') else str(response)
self.trace("LLM RESPONSE", response_text)
self._add_to_history(f"LLM prompt: {prompt}")
self._add_to_history(f"LLM response: {response_text}")
return response_text
except asyncio.TimeoutError:
raise TimeoutException(f"LLM思考超时")
except Exception as e:
raise AgentException(f"LLM思考失败: {str(e)}")
# ========== Tool 机制 ==========
def add_tool(self, tool_name: str, tool_func: Callable, description: str = ""):
"""添加工具"""
self.tools[tool_name] = {
"function": tool_func,
"description": description
}
def get_tools_description(self) -> str:
"""获取工具描述"""
if not self.tools:
return "暂无可用工具"
descriptions = []
for name, tool_info in self.tools.items():
descriptions.append(f"- {name}: {tool_info['description']}")View on GitHub (pinned to 606a07d341)
Solutions
- Raise the agent's timeout (constructor/config) to comfortably exceed worst-case generation time.
- Shrink the prompt: trim history, summarize context, or cap tool output fed back into think().
- Catch TimeoutException at the call site and retry once — transient provider slowness often clears.
- If it persists, check provider status/latency and whether the base_url is reachable quickly.
Example fix
# before
agent = SymptomCheckAgent(timeout=30) # too tight for long prompts
# after
agent = SymptomCheckAgent(timeout=180)
try:
out = await agent.think(prompt)
except TimeoutException:
out = await agent.think(shortened_prompt) # retry with trimmed context Defensive patterns
Strategy: retry
Try / catch
try:
out = await agent.think(prompt)
except TimeoutException:
await asyncio.sleep(2)
out = await agent.think(trim(prompt)) # retry once with a shorter prompt Prevention
- Size self.timeout to p99 generation time for your model, not the happy path.
- Trim history/context before every think() so prompt growth does not silently cross the timeout.
- Instrument think() duration; a rising trend predicts future timeouts.
When it happens
Trigger: Long prompts or slow reasoning models exceeding self.timeout; streaming disabled so the whole completion must land inside the window; provider under heavy load; the event loop being blocked elsewhere so the coroutine never progresses.
Common situations: Large health-record contexts producing multi-thousand-token prompts; tight timeouts copied from quick smoke tests; degraded LLM provider latency at peak hours.
Related errors
AI-assisted analysis of datawhalechina/hello-agents@606a07d341 (2026-08-14).
Data as JSON: /api/errors/954c8980ae8cf1a1.
Report an issue: GitHub.