Feat/group photo processor worker

[maya-ots]

summary

implements a JetStream worker that processes group photos , detects all faces, matches them against enrolled users, and creates PhotoFace and PhotoApproval records for each match.

technical implementation

• added PHOTO_GROUP_PROCESS subject to NatsSubjects.
• created db/generated/photo_faces.py querier for creating PhotoFace records.
• created db/generated/photo_approvals.py querier for creating PhotoApproval records.
• added list_users_with_embedding() query to db/generated/user.py.
• created app/worker/photo_processor/schema/event.py Pydantic model for the JetStream event payload.
• created app/worker/photo_processor/main.py worker that:
  • fetches photo from MinIO.
  • detects all faces in the photo and computes embeddings.
  • compares each embedding against all enrolled users using cosine similarity.
  • creates a PhotoFace record per detected face.
  • creates a PhotoApproval record with decision=pending for each matched user.

testing

• Added unit tests for PhotoGroupProcessWorker :
  • handles no faces detected (exits early).
  • creates PhotoFace and PhotoApproval for matched users.
  • does not create PhotoApproval for unmatched users.

[wailbentafat]

test have to be in test folder if it doesnt exist create it and for the images dont do tis image_path = r"C:\Users\Lenovo\OneDrive\Desktop\MultiAI\multAI-back\app\images\image1.png" add iamge to the test file and test with it

[wailbentafat]

noo this is wrong logique here for each face_embedding search about the user tht have the face get his id and then use in the phoot_face record creation and dont creaet bbox=none fix it

[wailbentafat]

this logiue must be in the db not in the service level so we can ensure durabilite

[wailbentafat]

refacotr this to use pgvector features

[wailbentafat]

and here just make sure tht decsision is enum in the service level