OpenSRE is an open-source CI/CD reliability agent that investigates failed pipeline runs, identifies root causes, and pushes fixes to pull request branches for review. Designed for self-hosted deployment with user-controlled LLM providers—including Anthropic, OpenAI, Ollama, Google Gemini, and others—it operates via a CLI offering both an interactive REPL shell and a headless mode for automation, plus a Python API for embedding in internal tooling. Upon a CI failure, OpenSRE automatically fetches failing check logs, recent commits, deployments, and correlated observability data from integrated platforms such as Grafana, Datadog, Honeycomb, and CloudWatch. It masks sensitive identifiers before external LLM calls, reasons across connected systems using a tool-calling loop to distinguish CI-caused from code-caused failures, and returns evidence-linked responses. Remediation fixes are pushed to the PR branch, and summaries can be posted to Slack, PagerDuty, or Telegram. The agent integrates with over 60 services spanning infrastructure (Kubernetes, AWS, Azure, ArgoCD), databases (PostgreSQL, MongoDB, ClickHouse, Snowflake), data platforms (Airflow, Kafka, Temporal), and dev tools (GitHub, GitLab, Jenkins). Deployment options include a managed cloud gateway, an AMI-based systemd gateway, and containerized self-hosting on Railway, ECS, or Vercel. The project, currently in public alpha, is built on a custom agent architecture documented in AGENTS.md, uses semantic end-to-end test catalogs for validation, and is licensed under Apache 2.0 with opt-out telemetry via Sentry.
Read original
github-trending/python