Repository navigation
Validate tutorial: conversational agent #131
Description
Goal: validate that docs/tutorial-conversational-agent.md works 100% from a clean environment as part of Sprint 5 adoption readiness.
Definition of done:
- Execute the tutorial end to end exactly as documented.
- Confirm all commands work without hidden assumptions.
- Confirm expected AgentOps outputs are produced, including results/report artifacts where applicable.
- Confirm CI/CD, Azure resources, and Watchdog steps work when the tutorial includes them.
- Capture evidence and update the tutorial if any step is unclear, broken, or incomplete.
Activity
Dongbumlee commentedon May 14, 2026
Parking this issue.
Audit during validation revealed the 'conversational agent' tutorial promises features AgentOps does not implement today:
extra_fields: [history]config keytests/unit/test_agentops_config.py:155)historyJSONL column forwarded to agentgrep -rn extra_fields src/agentops/returns 0 matches)orchestrator.py:543iterates rows independently; each is a fresh invocationrequest_field/response_fieldoverrides, single-turn Q&A scoringWhat remains is identical to the basic Foundry agent / HTTP agent tutorials — the dataset shape is the differentiator, not the invocation semantics. Validating the tutorial as a renamed copy of #127/#129 adds little value; reframing it requires a product call on whether AgentOps should support multi-turn (
history) and per-row passthrough (extra_fields) at all.Resuming after the product direction on conversation-state is decided. Branch
fix/issue-131-conversational-agent-validationhad no commits and has been discarded clean. Audit notes live in the session plan and can be reconstituted in 5 minutes when this comes back.