First assurance run
End-to-end path from context template to History using the reference SUT.
This walkthrough is a tutorial path for the Studio loop (see also the docs home map): generate a suite in Setup Wizard, export it, exercise the reference chatbot, score in Assurance Lab, and read results in History.
Overview
- Create a context template
- Generate and approve dimensions
- Select agents and generate drafts
- Review drafts and Save an approved suite
- Export JSONL
- Exercise the recipe-chatbot (or your own SUT)
- Assure via Studio or Gateway
- Open History
Steps
1. Generate an approved suite
In Studio, open Setup Wizard, or follow Generate: Context Template → optional PII → field enhance → Generate Dimensions → hub (agents + dimensions) → draft review → Save (new suite or merge). Suites are APPROVED at Save.
2. Export
Open Exports, select the suite, choose JSONL or Simulator JSONL. See Export formats.
3. Exercise the system under test
Use the recipe-chatbot reference SUT:
curl -s http://localhost:9000/api/chat \
-H 'Content-Type: application/json' \
-d '{"message":"Suggest a vegetarian pasta for two"}'Prefer enabling Langfuse on the chatbot so traces are named recipe-chatbot-session.
4. Assure
Pick one:
- Studio paste / JSONL — paste the conversation into Assurance (see Assure)
- Langfuse bridge — SDK Langfuse bridge
- Eval-only API — supply
execution_resultsto QAVE Studio workflows (see SDK overview)
5. Read History
Open History to inspect scores, evidence, and filters.
Next
- Wire the Assurance SDK for repeatable runs
- Add SME review in Workbench
- See Coming soon for one-step drive-and-score status