Execution Trees, Not More Logs: A Better Debugging Model for AI Agents

Описание и применение

Деревья выполнения позволяют явно показать структуру и причинные связи в работе AI‑агентов, выявляя вложенность, откаты, повторные попытки и параллелизм, и дают возможность создавать детерминированные проверки.

Задача

Flat logs cannot reveal causal relationships, ownership, fallbacks, retries, or concurrency in AI agent executions.

Результат

Execution trees provide explicit causal structure, expose failures and recoveries, reveal retries and concurrency, and enable deterministic checks, improving debugging efficiency.

Рабочий процесс

• 1. Instrument agent code with AgentInspect wrappers (inspectRun, step, step.tool, step.llm).
• 2. Run the agent to generate a trace stored in a directory.
• 3. View the execution tree locally with npx agent-inspect view.
• 4. Identify suspicious shapes (nested, fallback, retry, parallel).
• 5. Encode deterministic checks using npx agent-inspect check or TraceContract API.
• 6. Run checks as CI gates to enforce invariants.

Доказательства

A flat log can tell you that five things happened. It often cannot tell you which operation caused the next one, which failure triggered a fallback, or whether three tool calls were children of one planning step or unrelated work.

Источник

AgentInspect provides wrappers for a run and for named steps.

Источник

The tree does not replace raw event data. It is a projection of that data for the question developers usually ask first: *What path did this run take?*

Источник

npx agent-inspect view travel-planner
--dir .agent-inspect
--summary

Источник

npx agent-inspect check travel-planner
--dir .agent-inspect
--preset trajectory
--required-tool search-flights
--fail-on-observation failed

Источник