DocumentationSelf-Hosted V1 preview
scenariox.ai

Self-Hosted V1 preview. Check the release limitations before deployment.

EMERGENCY / ASSESSMENT

Emergency: Workspace or Evaluation Failure

Use this file when candidate workspaces cannot start, submissions do not progress, evaluations stall, or reports are unavailable while the rest of the platform is reachable.

Preserve assessment integrity

  • Do not create synthetic telemetry, evidence, scores, or reports.
  • Do not edit session, workspace, evaluation, evidence, scoring, or audit records.
  • Do not mount the Docker socket directly into workspace-orchestrator.
  • Record the assessment, candidate, session, workspace, evaluation, report, and trace identifiers that are available.

Collect platform state

./scenariox version
./scenariox status
./scenariox health
./scenariox logs

Workspace failure

Capture each boundary separately:

./scenariox logs session
./scenariox logs workspace-orchestrator
./scenariox logs docker-socket-proxy

Record whether the failure occurred during allocation, image pull, network creation, container start, IDE readiness, submission, or cleanup. Do not bypass docker-socket-proxy to test a workaround.

If logs show a transient worker failure and Docker authority is healthy:

./scenariox restart workspace-orchestrator
./scenariox health
./scenariox logs workspace-orchestrator

Evaluation or report failure

Capture the processing services:

./scenariox logs evaluation-service
./scenariox logs reporting-service
./scenariox logs reporting-api

Record the last visible session state and whether an evaluation or report identifier exists. Preserve the original submitted workspace and event history.

If logs show a transient process failure rather than missing or inconsistent evidence:

./scenariox restart evaluation-service reporting-service reporting-api
./scenariox health
./scenariox logs evaluation-service

Escalate

Escalate with the UTC timeline, release identity, identifiers, health output, redacted boundary logs, and the last successful candidate or recruiter action.

Do not mark an assessment complete, invent a score, or manually retry database state. Deterministic evaluation requires the original evidence snapshot and supported processing path.