Nimbus is a self-hosted CI platform built around Firecracker microVMs, org-scoped storage, and end-to-end observability for AI evaluation workloads. It replaces GitHub-hosted runners while keeping execution on infrastructure you control.

• Getting Started
• Configuration Reference
• Operations Guide
• Firecracker Security Hardening
• ClickHouse Schema
• Runbook

• Control Plane: Handles GitHub webhooks (HMAC + timestamp validation), manages DB-backed job leases and rate limits, and coordinates agent registration.
• Host Agent: Polls for assignments, provisions Firecracker microVMs, enforces capability restrictions, and persists in-flight state.
• Cache Proxy: Org-scoped artifact cache with optional S3 backend, eviction policies, and protected metrics endpoint.
• Logging Pipeline: Authenticated ClickHouse ingestion with org/repo filters on queries.
• Docker Layer Cache: OCI-compatible registry that enforces org-prefixed repositories and metadata ownership.
• Web Dashboard: React/Vite SPA for monitoring jobs, agents, logs, and system health.

• Lease fencing with fence tokens prevents duplicate job execution.
• Delivery ID tracking plus timestamp tolerance block webhook replays.
• Org-level access controls across cache, logging, and Docker registry assets.
• Metrics endpoints require a bearer token or loopback access by default.

1. Install dependencies and bootstrap the environment (uv venv, uv pip install -e .).
2. Configure the required environment variables and secrets (see Configuration).
3. Follow the detailed setup in Getting Started to launch services and run the test suite.

Workflows can target Nimbus runners by setting runs-on: nimbus. The control plane verifies workflow_job signatures (X-Hub-Signature-256 plus X-Hub-Signature-Timestamp), enforces per-org rate limits, and dispatches jobs to agents via leased assignments.

Nimbus publishes curated container images, such as nimbus/ai-eval-runner (Node.js, eval2otel, ollama client) for evaluation workloads. See Getting Started for example usage.

• Day-two procedures, monitoring, and ClickHouse schema live in the Operations Guide.
• Firecracker jailer, seccomp, and capability dropping guidance is documented in Firecracker Security Hardening.

• Complete: Multi-tenant isolation, lease fencing, webhook replay protection, distributed rate limiting, metrics endpoint authentication, tenant analytics dashboard.
• In Progress: Rootfs attestation, storage quotas, performance tuning.
• Planned: Snapshot boot support, browser automation for dashboard E2E tests.

Nimbus is ready for pilot deployments; major readiness items are summarized in the Operations Guide. Contributions improving security, observability, and distributed test coverage are welcome.

• Performance optimization
• Additional eval-specific runners