feature/fast-embeddings

[gecBurton]

Context

As a User I want the model to be loaded on start up and not on every embedding, so that the embeddings are done quickly

Changes proposed in this pull request

Previously we were using the fastapi Depends to handle dependency injection for the ml model, but this loads the model on every function call.

We could not switch to fastapi's lifetime because this is incompatible with faststream.

By reverting to pure faststream we can use its own lifetime which does work but this requires that we switch away from the http health check to a container healthcheck

Guidance to review

Relevant links

depends on https://github.com/i-dot-ai/i-ai-core-infrastructure/pull/260

Things to check

• I have added any new ENV vars in all deployed environments
• I have tested any code added or changed
• I have run integration tests https://github.com/i-dot-ai/redbox-copilot/actions/runs/9130721587