refactor(hotel_receptionist): split scenarios into scoped and open-ended files - #7065
Merged
Conversation
…ded files Pure split of scenarios.yaml, decided in the 2026-08-27 eng sync: scenarios graded deterministically on the final DB diff (userdata.expected_state) move to scenarios_scoped.yaml (57); scenarios graded on agent_expectations alone — disputes, policy/info Q&A, guardrail probes, judgment calls — move to scenarios_open_ended.yaml (43). Scenario bodies are byte-identical; only the file headers and suite names are new.
![devin-ai-integration[bot]](https://avatars.githubusercontent.com/in/811515?s=60&v=4){kind=small}
theomonnom
approved these changes
Aug 31, 2026
…specific names scenarios_scoped.yaml -> scenarios_tool_flows.yaml scenarios_open_ended.yaml -> scenarios_guardrails_and_faq.yaml Per PR review: names now describe the content (concrete tool-driven guest tasks vs guardrail probes, disputes, and policy/info Q&A) rather than the grading mechanism.
…narios_tool_accuracy.yaml
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
What
Splits
examples/hotel_receptionist/scenarios.yaml(100 scenarios) into two files, as agreed in the 2026-08-27 eng sync:scenarios_tool_flows.yaml(57) — scoped tool-based scenarios. Every scenario drives a concrete tool flow to a deterministic end state and is graded on the final DB diff (userdata.expected_state) in addition toagent_expectations. These are the ones representative of real customer flows and usable to evaluate simulation quality.scenarios_guardrails_and_faq.yaml(43) — open-ended/adversarial scenarios: disputes, complaints, policy/info Q&A, guardrail probes (prompt extraction, privacy pressure, harmful requests), and judgment calls. Graded onagent_expectationsonly; useful for guardrail testing, not sim-quality evaluation.Split criterion
Presence of a deterministic
userdata.expected_state— the objective encoding of "scoped tool-based": the scenario's outcome is asserted as DB state written through the agent's tools. Everything NL-judge-only goes to the guardrails file. If any individual scenario should live on the other side, moving it is a pure cut/paste.Guarantees
The split was done textually (comments and formatting preserved) and validated by script:
expected_state; no guardrails/faq scenario doesOnly the file headers and the two suite
namefields are new. Nothing in the repo referencesscenarios.yamlby path, so no code changes are needed.