Firebase Function - Search

[yevkim]

MIgrate the search code to Firebase Function Explore keeping the vector files in Firebase Storage

[CJBuzz]

Hi I have tried to implement the search API (predicting schemes) on firebase and tested a little using Thunder Client.

Some notes:

• I have tried to make minimal changes compared to the FastAPI backend so query records are still stored in a dictionary, can explore using firestore potentially(?)
• I haven't discovered a way for firebase to handle lifespan events, so initialisation of the Search models are done upon first request.
• The original ml_logic folder (in backend) which contains the code for the implementation of the Search algorithm can't really be accessed by main.py in the functions folder as the virtual environment is located within the functions folder, so I duplicated the ml_logic folder into the functions folder instead. Might be able to use some form of symlink instead, or in production, store the code for the FastAPI app somewhere else and move the functions venv up.

[CJBuzz]

Successfully implemented Search function with firestore but schemes embedding (.npy file) and index (.faiss file) needs to be re-generated.

Models, preprocessors and tokenizers are also initialised as a class variable so they can be initialised from main.py

[yevkim]

Create the session id and return to frontend along with results Singleton Pattern - Check with Traci Loggin sessionid/queries/Results in the Firestore Deploy the function to firebase