Ghost in the Machine: DEADLOOP, a Local-First Reliability Monitor for AI Agents

说明与用途

DEADLOOP 是一个本地优先的 AI 代理可靠性监控器，检测重复工具调用和循环，生成基于证据的事件，并允许人类在不改变代理行为的情况下进行干预。

问题

AI agents often get trapped in repetitive tool calls, unproductive loops, or circular workflows, wasting tokens and time.

结果

DEADLOOP detects and reports repetitive agent behavior locally, enabling humans to intervene without external API calls, preserving data privacy and reducing token waste.

工作流

• 1. Agent emits tool-call events to DEADLOOP via DeadLoopClient.send_event.
• 2. DEADLOOP ingests events, stores them in SQLite/PostgreSQL.
• 3. Detector logic evaluates traces for loop patterns and session budgets.
• 4. If a finding is detected, DEADLOOP generates an incident with event IDs, metrics, confidence, and uncertainty.
• 5. Human reviews incident, can acknowledge, dismiss, request pause, or approve pause.
• 6. If pause approved, agent execution is halted without claiming control.

证据

DEADLOOP ingests its tool-call trace and answers: Is this repetition, or legitimate retrying?

来源

The core is fully deterministic; an optional open-weight model only explains the findings. It never changes a finding or executes an intervention.

来源

It runs offline. Detection, risk scoring, ingestion, and incident management are deterministic and local. No API key is required for the core product to function.

来源

The sample simulator generates real execution events through the same ingestion, persistence, detection, and incident paths as production.

来源