The user uploaded content (images/documents) are now directly
uploaded to S3, and a pre-signed URL is given to the LLM.
The code is a bit complicated because there are behaviors
from docs, other from drive, and in the end some parts are
not used.
This will need some cleanup.
Improvements:
- Allow to upload PDF without vectorized storage (to let
LLM manage them on the fly)
- Allow to configure whether a model can support images/PDF
or not
This is the first step to allow several models to be available to
the users.
This is not a very good implementation, as it aims to keep the
ability to use django settings for configuration:
- makes it easy to deploy only one model
- allow changes like before in tests
Next step would be to remove the django settings and update
all tests etc.
This is the first commit which provides all the first stack for a
working chat.
This is a first implementation with:
- Vercel SDK for the frontend part
- OpenAI Agent SDK for the backend
The stack can use a local LLM with docker ot a remote one.
This implementation is more a draft, but it provides the project
structure.
All tests are working even if we lack a lot of them.
charles