AI Assistant which answers questions to a given PDF document. Has to be run on a local Server, with litestar run (Terminal command which has be run in the same directory as app.py).

Package dependencies can be found in the conda_env.yaml.

No working UI at the moment. But backend is fully functional and can be reached via curl command, or similar tools.

Local Server can be reached at http://127.0.0.1:8000.

• Documents can be uploaded via: curl -X POST "http://localhost:8000/upload/pdf" -F "file=@[PATH/TO/FILE/HERE]"
• Questions can be asked via: curl -X POST "http://localhost:8000/ask" -H "Content-Type: application/json" -d '{"question": "[PUT QUESTION HERE]"}'

Paths given for the upload can be relative.

Important: This repo does not contain the model! I used the gguf of Phi-3-mini-4k-instruct-q4. This is a very small model with a low context length, does not work on documents longer than approximately 4-6 pages. If you want to use this app you should get either the same or a compatible model. This is put into the model/ directory at the same level as the source directory is.

At the End the directory structure looks like this:

├── ai-assistant
    ├── data    # Has a sample CV for testing inside
    ├── model   # Create this and put the model here
    ├── source  # Here the code resides

• Working UI Interface.