CINEIA – AI

1 · Prepare .env Create FlaskAPI/.env with your database credentials:

DB_HOST=postgresql-yannr.alwaysdata.net
DB_NAME=yannr_00
DB_USER=yannr_01
DB_PASSWORD=Projet1234
SECRET_KEY=dev-key

(The key can be any value - it is only used by Flask.)

2 · Install Python dependencies Activate your virtual-env, then install the two requirement files:

pip install -r DNN_TorchFM_TTower/requirements.txt
pip install -r FlaskAPI/requirements.txt

3 · Train (or re-train) the AI models Run from the project root CINEIA/. Note the fully-qualified module paths (DNN_TorchFM_TTower. prefix).

3-A Train / retrain the Two-Tower recall model

python -m DNN_TorchFM_TTower.models.recall.train_two_tower --epochs 3 --batch 128

3-B Train / retrain the DeepFM re-rank model

python -m DNN_TorchFM_TTower.models.ranking.train_ranking --epochs 3

(You may shorten epochs while testing; saved weights go to DNN_TorchFM_TTower/saved_model/.)

4 · Run the Flask API

python FlaskAPI/app.py # default http://127.0.0.1:5000

Visiting / shows 404 – that is expected; the API lives under /api.

5 · Call the Recommendation endpoint Pattern:

GET /api/recommend/<user_id>?top=<N>

Example:

http://localhost:5000/api/recommend/51