Jsonl based evaluation flow

[yangm2]

What type of PR is this? (check all applicable)

• Refactor
• Feature
• Bug Fix
• Optimization
• Documentation Update
• Infrastructure
• Maintenance

Description

This PR starts to move away from the original csv file with the test scenarios and uses the JSONL format with a scenario-schema that we can validate.

The scenario-schema is important because we reference the underlying structure of the scenarios (e.g. facts, city) in the evaluation flow (e.g. instructions to the LLM-as-a-Judge).

This flow also makes it easier to leverage the LangSmith web interface to make changes to the dataset. This is important for the non-coder members of the team who are responsible for the legal content of the dataset.

Unrelated - put the PR Code Review criteria into CLAUDE.md.

Related Tickets & Documents

• Related Issue #
• Closes #

QA Instructions, Screenshots, Recordings

Please replace this line with instructions on how to test your changes, a note on the devices and browsers this has been tested on, as well as any relevant images for UI changes.

Added/updated tests?

• Yes
• No, and this is why: not yet
• I need help with writing tests

Documentation

• If this PR changes the system architecture, Architecture.md has been updated

[optional] Are there any post deployment tasks we need to perform?

[yangm2]

@TruMichael-jpg & @dan-moncada - just a head's up that this is something I'm working on to enable the LangSmith web interface flow. Checkout the updates to EVALUATION.md to get an idea about the direction the flow is going.

@leekahung - head's up that I'm moving the evaluation stuff into a separate directory