How ashu7531/autosre works
AutoSRE is an incident-response agent with a dashboard and an automatic Sentry-webhook entry point. The main workflow selects or receives an incident, runs a metrics validation check, gathers crash issues, commits, and dependency health through MCP, retrieves similar incidents, and asks Gemini to produce an RCA streamed to the dashboard. The README describes broader collectors and production integrations; this graph includes only integrations demonstrated in the sampled code. The memory implementation uses PostgreSQL/pgvector, despite the README’s demo-limitations section describing an in-memory design.
Main entry points: api/index.ts HTTP and SSE API · agent/index.ts Investigation agent · mcp-server/index.ts MCP tool server · App.tsx Incident dashboard
Components 11
Dashboard
- Incident dashboard
frontend/src/App.tsx
API orchestration
- HTTP and SSE API
src/api/index.ts - Scenario data
src/scenarios/active.json
Agent reasoning
- Investigation agent
src/agent/index.ts
Evidence and memory
- MCP tool server
src/mcp-server/index.ts - Incident retrieval
src/agent/memory.ts - Incident memory
Other components
- Incident operator
- Gemini AI
- Sentry
- GitHub
Connections 17
- Incident operator → Incident dashboard: selects incident
frontend/src/App.tsx - Incident dashboard → HTTP and SSE API: requests scenarios
frontend/src/App.tsx - HTTP and SSE API → Scenario data: loads or selects
src/api/index.ts - Incident dashboard → HTTP and SSE API: triggers investigation
frontend/src/App.tsx - Sentry → HTTP and SSE API: sends webhook
README.md - HTTP and SSE API → Investigation agent: runs investigation
src/api/index.ts - Investigation agent → MCP tool server: calls tools
src/agent/index.ts - MCP tool server → Scenario data: reads scenario
src/mcp-server/index.ts - MCP tool server → Sentry: fetches issues
src/mcp-server/index.ts - MCP tool server → GitHub: fetches commits
src/mcp-server/index.ts - Investigation agent → Incident retrieval: searches incidents
src/agent/index.ts - Incident retrieval → Incident memory: queries vectors
src/agent/memory.ts - Investigation agent → Gemini AI: requests RCA
src/agent/index.ts - Incident retrieval → Gemini AI: supplies context
- Investigation agent → HTTP and SSE API: emits events
src/api/index.ts - HTTP and SSE API → Incident dashboard: streams updates
frontend/src/App.tsx - HTTP and SSE API → Incident memory: reads history
src/api/index.ts