[hartmark]

Summary

I've added backend support for OpenAI-compatible batching.

https://developers.openai.com/api/reference/resources/batches/methods/create

Validation

• npm run lint
• npm run test:unit
• npm run test:coverage
• Coverage is still >= 60% for statements, lines, functions, and branches
• SonarQube PR analysis is green or any remaining issues are explicitly documented below

Tests Added Or Updated

I've added .http test to test the actual endpoints for the API. I've also added integration and unit-tests.

Coverage Notes

I've added testcases for all affected code and new code

Reviewer Notes

• This adds just the endpoint, perhaps a opt-in option or option to opt-out?
• Files are saved in database
• There is no user-facing UI for the batches, I'm thinking of adding a simple grid to list batches and their statuses, and a grid for files.
• My goal is to support Mistral AI batching, so next step is to have a check if all models refer to a mistral target it should call the Mistral Batch API.
  https://docs.mistral.ai/api/endpoint/batch

[gemini-code-assist]

Code Review

This pull request implements a comprehensive batch processing system and file management API, including database migrations, background polling services, and REST endpoints for batch and file operations. It also adds support for persistent API keys via environment variables. Feedback focuses on optimizing database performance by throttling batch updates, improving memory efficiency when handling large file uploads, and fixing potential storage leaks in the cleanup logic. Additionally, it is recommended to handle potential unhandled promise rejections in the asynchronous batch item processing to ensure system stability.