Local Runtime Api Scenarios

Step 2: Start API server

Use either:

• python scripts/run_api.py
• uvicorn app.api.main:app --host 127.0.0.1 --port 8000

Health check:

• curl.exe http://127.0.0.1:8000/docs

Step 3: Run scenario requests

Use the payloads and commands in scenarios.md.

Always include:

• workflow_type: "invoice_governance"
• request.input_payload
• request.actor
• request.budget

Step 4: Validate outputs and artifacts

For each scenario:

• capture run.run_id
• check run.status, run.current_state, run.pending_tool_id, run.steps
• call GET /workflows/{run_id}/audit

Also inspect:

• artifacts/audit.jsonl
• artifacts/runs.jsonl

Endpoint Map

• POST /workflows/execute
• GET /workflows/{run_id}
• POST /workflows/{run_id}/approval
• POST /workflows/{run_id}/resume
• GET /workflows/{run_id}/audit

Result Expectations by Scenario

• happy path -> COMPLETED
• deny tool by role -> BLOCKED
• require approval -> AWAITING_APPROVAL then COMPLETED after approval + resume
• approval rejected -> BLOCKED
• hard budget exceeded -> BLOCKED
• malformed input -> FAILED

Failure Handling

• Never silently ignore failed requests.
• If an API call returns non-2xx, report request body and response body.
• If observed status differs from expected, mark scenario failed and include run.steps[-1].error when present.