DigitalOcean 发布 Signals 与 Simulations 工具检测静默失败的 Agent 会话
DigitalOcean 推出新工具 Signals 和 Simulations，用于标记静默失败的 Agent 会话，并在发布前针对修复进行重放验证。Signals 用低成本规则监控所有实时会话，识别用户重复提问、要求转人工、工具循环重试等人类困扰信号；随后 Simulations 用假想用户在修复后的 Agent 上重放这些会话，检查整段对话是否正常。
Agent failures often pass every health check. A tool call hits the wrong database region, returns 200, and the user gets a slower, worse answer. Nothing alerts.
DigitalOcean’s new tools flag agent sessions that fail silently, then replay them against a fix before release.
Because, the clue behind an agent bug is often sitting in something nobody can search.
Signals watches every live session with cheap rules, not an AI model, that look for human trouble signs such as a user repeating themselves, asking for a human, or a tool being retried in a loop, then passes those sessions to Simulations, which replays them with a fake user against your fixed agent to check the whole conversation works before the fix ships.
Until now teams had to pay an LLM to grade every conversation, read transcripts by hand, or invent test cases from imagination, while here the real failures from production automatically become the test set.
来源：rohanpaul_ai · x.com