🤖 Provider API update - togetherai@8.0.0

[vercel-ai-sdk]

Breaking changes

• GET /evaluation/{id}/status
  • 200 response schema narrowed: results is no longer nullable.
  • 200 response schema narrowed: removed the { error: string } variant from results oneOf. Clients should rely on the top-level status field (e.g., error, user_error) to detect errors.

New features

• GET /evaluation

  • Added new endpoint for retrieving evaluation jobs (replacement for the deprecated /evaluations route).

• POST /evaluation

  • Added new endpoint to create a new evaluation job.

• GET /evaluation/model-list

  • Added new endpoint to retrieve the list of models available for evaluation.

• POST /evaluation/{id}/update

  • Added new endpoint to update an existing evaluation job.

• GET /evaluations

  • Added optional query parameter: userId (admin-only; empty string returns all jobs).

• GET /evaluations/model-list

  • Added optional query parameter: model_source (default: all).

Fixes

• GET /evaluation/{id}

  • Clarified success and error response descriptions.
  • Standardized tag to evaluation.

• GET /evaluations

  • Clarified success and error response descriptions.
  • Documented explicit 500 error response schema.
  • Marked endpoint as deprecated in the summary; recommends using /evaluation.
  • Standardized tag to evaluation.

• GET /evaluations/model-list

  • Added description for model names in the response.
  • Documented 400 error response; refined 500 error description.
  • Marked endpoint as deprecated in the 200 response description; recommends using /evaluation/model-list.
  • Standardized tag to evaluation.

https://github.com/gr2m/ai-provider-monitor/releases/tag/togetherai%408.0.0