| name | run-certification |
| description | Run the full certifier pipeline (Phase 0+1+2+3) locally against a Langfuse trace using scripts/run_certification.py, then summarize the .tmp/ output tree. Use for the certifier inner dev loop without starting the FastAPI service. |
run-certification
Exercise the certifier pipeline end-to-end against one trace, off the API path.
Prerequisites
.env filled in with LANGFUSE_* and AZURE_OPENAI_* (the script reads them directly).
- Python deps installed. PDF rendering needs
reportlab (pip install --user reportlab).
Steps
- Run the pipeline. Prefer the trace-ID form, which auto-resolves agent/experiment/run IDs:
./scripts/run_certification.py --trace-id <LANGFUSE_TRACE_UUID>
Or pass IDs directly:
./scripts/run_certification.py \
--agent-id <UUID> --experiment-id <UUID> --run-id <UUID> --agent-name vaya
- Useful flags:
--skip-cert (stop after Phase 0+1), --no-pdf (skip PDF render), --runs-per-fault N (Phase 2 significance), --batch-size N (LLM classification batch), --workspace <dir> (default .tmp/), --debug (retain Phase 3 intermediates).
- Summarize output. Outputs land under
.tmp/{agent_id}/{experiment_id}/. Report the key artifacts:
fault-bucketing/{run_id}/metrics/*_metrics.json
aggregation/aggregation.json
cert-builder/certification.json (final certificate)
cert-builder/certification.pdf (rendered report)
Notes
.tmp/ is gitignored — safe to leave artifacts there.
- To re-render a PDF from an existing
certification.json, use scripts/render_certification_pdf.py.
- To verify the service path instead of the script path, use the
pipeline-smoke-test skill.