https://github.com/prapeller/Async_API_sprint_2

• provides films search possibility by name/genre and other film properties

https://github.com/prapeller/Auth_sprint_2

• provides user session control by authentication/authorization

https://github.com/prapeller/ugc_sprint_2

• provides possibility for users collect and control their activity
• provides possibility for analytics team analise users activity, make business conclusions, and improve online cinema based on these conclusions

https://github.com/prapeller/notifications_sprint_1

• provides possibility for users to receive/not receive (switch off/on) notifications (email / telegram / interface)
• provides possibility for cinema to send notifications:

https://github.com/prapeller/graduate_work

• provides possibility for users to get the recommended films based on their ugc

notes:

• weight - importance of vector's part. if vector consists of 3 describable parts (film genres names/film description/film type) then weight of genres (3) is more than weight of film description (2)
• film vector - list of dots-floats, vectorized in n-dimension space based on describable parts (genres names (weight 3) / description (weight 2) / type (weight 1))
• mongo_ugc - mongo db from 'ugc service' with user likes (film_uuid list)
• postgres_search - postgresql db from 'search service' with films film_uuid/genre(varchar)/description(text)/type(varchar)

containers:

• vector_recommendations (milvus)
  • stores film vectors (film_uuid: [point, point]) at vector in-memory vector db
• api_recommendations
  • fastapi based http interface to
    • get/set film vectors to vector db
    • recommend films for current user
• etl_postgres_to_vector_recommendations
  • constantly (on any postgres_search.film table update (triggered on comparison of film.updated_at field timestamp and lastly index update timestamp in 'last_etl_update' file)) extracting updated film_data, transforming with gensim(word2vec model), and loading to vector_recommendations using fastapi post route
• redis_recommendations
  • stores in redis cache recent frontend requests results

data flow:

• etl_postgres_to_vector_recommendations loads updated films vectors to vector_db
• api_recommendations (after user request)
  • checks cache_recommendations first
  • if cache misses, request goes to mongo_api for getting list of current_user bookmarked film_uuids, and request vector_recommendations to get a film_uuids of closest neighbors (by vectors value)
  • film_uuids are to set in cache and sent back to the frontend

• make api-redis-build-loc

• cd api_recommendations python3.11 -m venv venv && source venv/bin/activate && pip install -r requirements/local.txt

• export DEBUG=True && export DOCKER=False && python main.py

• swagger docs can be found at 127.0.0.1:8085/docs

• make build-loc

• swagger can be found at 127.0.0.1:85/docs