How We Built an Autonomous AI Agent That Audited 187 Subprojects and Got PRs Merged by Google Cloud Engineers

Описание и применение

Автономный агент ApiPatch автоматически обнаружил и исправил устаревшие вызовы API в 187 подпроектов репозитория awesome-llm-apps, создав чистые, разделённые PR, которые были приняты мейнтейнерами и инженерами Google Cloud.

Задача

Fast‑moving dependencies in AI frameworks break existing code, causing import errors, runtime failures, and manual churn to fix deprecated API calls.

Результат

ApiPatch audited 187 subprojects, generated 3 PRs (including a merged PR #1148), and demonstrated that automated, AST‑validated migrations can be accepted by large open‑source maintainers and Google Cloud engineers.

Рабочий процесс

• Scan repository to discover subproject roots via requirements.txt or pyproject.toml
• Analyze call sites against known deprecation patterns for major frameworks
• Validate proposed changes with AST integrity and syntax checks
• Create isolated Git branches per subproject and open surgical Pull Requests
• Maintain requirements alignment and import order
• Review and merge PRs by maintainers

Доказательства

ApiPatch is an autonomous AI agent engineered specifically to detect and repair deprecated API calls across codebases, verify them against syntax and AST integrity shields, and automatically generate clean, partitioned pull requests.

Источник

We ran ApiPatch on awesome-llm-apps using subproject-level isolation. Here are the patches it generated: 1. Modernizing RouteLLM Config (PR #1148 — MERGED 🎉)

Источник