Provide a separate manual-mapping helper tool for validation (not the evaluation platform)

[filiperochalopes]

User story

As a validator, I want an easy way to access local WHO ICD API tools for manual mapping during validation without using the same platform that will be evaluated, so the validation workflow does not influence evaluation results.

Use case

Manually map 200–250 terms using decision-support helpers (e.g. WHO ICD API) while recording results in a shared spreadsheet.

Requirements

• Confirm evaluation uses a cloud-hosted WHO ICD API (Docker container) as the API source.
• Provide a manual-mapping helper that is separate from the evaluation platform (do not use the platform planned for evaluation).
• Support an approach similar to the previous V1 Lab page (individual tools manually are OK).
• Enable validators to access the APIs/tools separately to help create new mappings for validation sets.
• Keep manual validation output in a shared spreadsheet accessible to all validators.

Acceptance criteria

• Given a validator needs decision support, they can use a standalone helper (separate from the evaluation platform) to query the local WHO ICD API.
• Given a mapping is produced manually, the result can be recorded in a shared spreadsheet for the validation set.
• The helper does not depend on or reuse the platform that will be evaluated.

[filiperochalopes]

What tools should be available on this page (the AI Agent or just those provided by WHO)? Who will select the sample of the 200 clinical concepts (will it be a subset of the CIEL)? Should I build the page with the test dataset to be shown to IMO? Can they access the other pages of the CIEL Lab V1, or should I create a separate environment?

I imagine that the best approach would be to capture new examples outside of CIEL to see how CIEL performs with Bridge without perfect matches.