Troubleshooting
Debug API credentials, platform IDs, driver configs, endpoint reachability, outputs, generated packs, and capability mismatches.
API and platform ID errors
Check API configuration first:
echo "$DETERMINA_API_BASE_URL" test -n "$DETERMINA_API_TOKEN" determina api-health
Authenticated hosted commands require valid project, system, version, run, baseline, or artifact IDs. Placeholder values such as <project-id> must be replaced before running.
System or driver errors
For native local HTTP setup, confirm the service is running and use check-system.
determina check-system --system-type recommender --system-url http://127.0.0.1:8051
For schema-mapped HTTP, provider drivers, MCP, in-process Python, or custom agent sessions, validate the driver config path and required fields.
Missing output
Pass --output-dir when you need a predictable local folder. For hosted runs, download artifacts by artifact ID when the platform returns them.
determina artifacts download --project-id <project-id> --artifact-id <artifact-id> --output-dir ./determina-output
Capability mismatch
Most confusing failures come from using a command in the wrong lane:
- Hosted production commands use platform IDs.
- Local no-key setup commands can validate direct URLs, driver configs, source specs, upload previews, or existing artifact files.
- Installed-package execution commands such as local
audit, localcompare, localrun-swarm, plan execution, generation, runs, and corpus workflows can fail closed unless your environment has the corresponding platform/core capability. - Search supports generated test cases through advanced workflows, but search test case execution parity is not claimed here.
- Agent generated packs are task/environment review artifacts, not generated customer agents.
Check the CLI reference
Use command groups to confirm whether the command belongs to hosted production, local setup, artifacts, or advanced workflows.