Skip to content
beskid Beskid

Beskid

Jump to a Beskid service

Beskid

Jump to a Beskid service

Operate Nexus

Store Nexus tokens in the deployment secret manager. Do not print, commit, or expose NEXUS_MCP_AUTH_TOKEN, webhook secrets, or provider keys.

FieldVerified value
PurposeInteractive repository graph, indexed catalogue, and optional MCP access.
AudienceReaders inspect public graphs. Authorized maintainers manage repositories and MCP connections.
Public boundaryThe public hostname must pass through the configured proxy trust boundary. Do not publish port 8452 directly.
Local boundaryThe Nexus container exposes internal port 8452 to the Compose edge network; production Compose publishes no host port. Data lives below GITNEXUS_HOME.
AuthenticationCaddy and Authentik forward-auth protect every public route. MCP can require an additional bearer token.
Persistent statenexus-data persists /data/gitnexus, including indexes and generated code documentation.
Container imageghcr.io/cyber-nomad-collective/beskid-nexus.
Health checkInternal GET /api/health on container port 8452.
Deployment ownerThe Nexus repository builds its service image. The platform lane supplies the runtime boundary.
Secret sourceDeployment secrets include optional MCP, webhook, and OpenRouter values. Keep them in the configured lane store.
MonitoringHealth, index completion, and graph-load results show separate failure boundaries.
RecoveryRestore nexus-data, verify the proxy boundary, and re-index only after the stored catalogue is safe.

Confirm the repository graph task and the configured GITNEXUS_HOME. Confirm that the forward-auth proxy protects the public hostname.

  1. Inspect the Nexus container’s internal /api/health healthcheck on port 8452 against the service contract.
  2. Open the proxy-protected hostname.
  3. Verify the repository catalogue.
  4. Run one authorized catalogue analysis.
  5. Wait for its terminal result.

The internal /api/health check succeeds. The selected repository graph loads after Caddy forward-auth and Authentik sign-in, and separate application permissions control management functions.

If health fails, inspect the image and nexus-data mount. If only sign-in fails, restore the proxy header contract and redeploy. If an index is corrupt, restore its data or remove only that verified index, then run a controlled re-index.

For signed-in graph reading, see Explore Nexus. For protected operations, review service containers and volumes.