Application tracing configuration
Configure these server-side values in the deployment environment. Never use aNEXT_PUBLIC_ prefix.
The production endpoint is:
Safe startup behavior
Tracing becomes active only when the feature flag and a valid HTTPS endpoint are present. Missing configuration leaves the Quantura API operational with tracing disabled. Invalid or unavailable observability infrastructure must not block normal API requests. Exports are bounded and flushed after the response using Vercel’swaitUntil lifecycle. Only server-defined route templates, HTTP methods/statuses, elapsed time, and generated request/trace IDs are recorded. Raw URLs, query strings, headers, bodies, cookies, IP addresses, account identifiers, and raw error messages are deliberately excluded. Automatic outbound HTTP tracing is not enabled because provider URLs may contain sensitive identifiers. Unmatched paths use the fixed label unmatched.
Query credentials and documentation secrets
The repository’s encrypted GitHub Actions secrets containGITLAB_OBSERVABILITY_API_KEY, SIGNOZ_API_KEY, and MINTLIFY_API_KEY. They are not browser variables and are not needed by the public website build. The first two are service-account query keys used with SIGNOZ-API-KEY; they are not GitLab source-control tokens or SigNoz ingestion keys.
The GitLab query API is https://140928869.gitlab-o11y.com/api/v1. The separate SigNoz instance is https://suited-macaque.us2.signoz.cloud. Configuring the latter’s query key does not send application telemetry to SigNoz Cloud; that would require its distinct ingestion configuration.
GitLab CI telemetry
GitLab CI/CD telemetry is auto-wired for this project withGITLAB_OBSERVABILITY_EXPORT=traces,metrics,logs. The repository uses an explicit .gitlab-ci.yml and no longer relies on the deprecated Herokuish Auto Test stage. Application traces and CI pipeline telemetry are independent signals.
MCP access
The Observability MCP endpoint is:SIGNOZ-API-KEY header. Store that query credential in the client’s secure secret store; never commit it to an MCP configuration file.
Verification
After deployment:- Call
/api/healthand a representative authenticated API endpoint. - Open GitLab Observability Services.
- Select
quantura-apiand verify traces carrygitlab.project.id,gitlab.project.name,service.version, anddeployment.environment.name. - Confirm authorization headers and request bodies are not captured as span attributes.
- Query
http.server.request.countandhttp.server.request.durationmetrics, and theHTTP request completedlogs for the same service/version. An HTTP success from ingestion alone does not establish searchable retention: verify through the query API/dashboard as well.
observability-verification.yml GitHub Action requests production API health and runs scripts/verify_gitlab_telemetry.py. It fails unless all three service-scoped queries return positive data within the last 30 minutes, and prints counts/status only—not individual customer records.
The SSR, newsletter, Python compatibility APIs, model workers, and Vercel platform/build logs require separate instrumentation or a correctly authenticated collector/drain before claiming complete platform-wide coverage. CI telemetry configuration must be checked in GitLab; an application deployment does not run a GitLab pipeline by itself.
GitLab Observability is currently documented by GitLab as an experimental feature, so Quantura treats export failures as non-fatal and retains Vercel runtime logs as the operational fallback.