datawhalechina/hello-agents · error · AgentException

RiskAssessmentAgent 执行失败: {str(e)}

Error message

RiskAssessmentAgent 执行失败: {str(e)}

What it means

This is the catch-all wrapper thrown by RiskAssessmentAgent.run's except block. Any exception inside run() — including the empty-indicator AgentException, LLM failures inside _assess_risk, or prompt formatting errors — is caught, agent state is set to 'error', and a new AgentException is raised with the prefixed message. The original exception text is embedded via str(e) but the traceback chain is lost because it is re-raised as a new exception without 'from e'.

Source

Thrown at Co-creation-projects/Shawnxyxy-HealthRecordAgent/backend/agents/risk_assess.py:26

class RiskAssessmentAgent(BaseAgent):
    def __init__(self, task_id=None, llm=None):
        super().__init__(name="RiskAssessment", task_id=task_id, llm=llm)

    async def run(self, input_data: Dict[str, Any]) -> Dict[str, Any]:
        try:
            indicator_results = input_data["indicator_results"]
            if not indicator_results:
                raise AgentException("缺少健康指标分析结果")
            self.set_state("running")

            result = await self._assess_risk(indicator_results)

            self.set_state("completed")
            return result
        except Exception as e:
            self.set_state("error")
            raise AgentException(f"RiskAssessmentAgent 执行失败: {str(e)}")

    async def _assess_risk(self, indicator_results: Dict[str, Any]) -> Dict[str, Any]:
        risk_prompt = f"""
你是一名专业的健康风险评估专家。

以下是某用户的健康指标分析结果（已由其他智能体完成分析）：
{indicator_results}

请你完成以下任务：
1. 综合判断用户的整体健康风险等级（low / medium / high）
2. 列出主要风险因素（不超过 5 条）
3. 推测可能存在的潜在健康风险或疾病方向
4. 给出你评估的置信度（0~1 之间的小数）

请以 JSON 格式返回，例如：
{{
  "overall_risk_level": "medium",
  "risk_factors": ["高胆固醇", "睡眠不足"],

View on GitHub (pinned to 606a07d341)

Solutions

  1. Read the suffix of the message after 'RiskAssessmentAgent 执行失败: ' — that is str(e) of the original exception and identifies the real failure.
  2. If it is '缺少健康指标分析结果', fix the upstream input (see the empty-indicator error).
  3. If it mentions network/timeout/auth, fix the LLM configuration (api_key, base_url, model name in config).
  4. Improve the code: use 'raise AgentException(...) from e' and catch AgentException separately so precondition errors are not double-wrapped.

Example fix

# before
except Exception as e:
    self.set_state("error")
    raise AgentException(f"RiskAssessmentAgent 执行失败: {str(e)}")

# after
except AgentException:
    self.set_state("error")
    raise
except Exception as e:
    self.set_state("error")
    raise AgentException(f"RiskAssessmentAgent 执行失败: {str(e)}") from e
Defensive patterns

Strategy: try-catch

Try / catch

try:
    result = await risk_agent.run(input_data)
except AgentException as e:
    msg = str(e).removeprefix("RiskAssessmentAgent 执行失败: ")
    log.error("risk agent failed: %s", msg, exc_info=True)
    # classify by inner message; do not blind-retry LLM/auth failures
    raise

Prevention

When it happens

Trigger: Any failure inside run(): empty indicator_results (error 140), exceptions from self._assess_risk (LLM timeout, API auth failure, malformed JSON in the model response), or AttributeError from unexpected input shapes.

Common situations: LLM API key/quota problems surfacing through the agent; upstream output shape drift; retry loops that see only the wrapper message and can't classify the root cause.

Related errors

AI-assisted analysis of datawhalechina/hello-agents@606a07d341 (2026-08-14). Data as JSON: /api/errors/c43be3955a7bea16. Report an issue: GitHub.