MongoDB schema & indexes for messages and conversations

[butuz0]

1. Configured MongoDB and Mongo-Express services.
2. Added mongodb to settings.py DATABASES, configured DATABASE_ROUTERS to avoid relations and migrations conflicts when using two databases. Configuring project to use MongoDB.
3. Created conversations with MongoDB documents Conversation and Message with default_auto_field = "django_mongodb_backend.fields.ObjectIdAutoField". Creating MongoDB documents.
4. Created indexes for Conversation and Message documents. Collections and Indexes are created automatically by running mongodb migrations. Adding MongoDB Indexes.
5. Configured pytest to use settings_test.py. Added cleanup fixture for MongoDB tests.

[mehalyna]

The PR exposes MongoDB (port 27017) and mongo-express (port 8081) on the host with no clear authentication or network restrictions. This can leak data or allow remote modification of the development database when docker-compose is started on any host.
Restrict exposure in docker-compose for non-local development (remove host ports or bind to localhost), enable authentication, and document the risk. Example change:

mongodb:
  # remove "27017:27017" or change to "127.0.0.1:27017:27017" for local-only bind
  ports:
    - "127.0.0.1:27017:27017"

mongo-express:
  ports:
    - "127.0.0.1:8081:8081"
  environment:
    - ME_CONFIG_MONGODB_ADMINUSERNAME=${MONGO_INITDB_ROOT_USERNAME}
    - ME_CONFIG_MONGODB_ADMINPASSWORD=${MONGO_INITDB_ROOT_PASSWORD}

Also add docs warning and ensure credentials are set via .env in CI/CD.

[butuz0]

Authentication is enabled by setting mongodb and mongo-exopress credentials in backend/.env, which containers load through env_file.
Originally i avoided setting those values directly in docker-compose in the way you suggested: ME_CONFIG_MONGODB_ADMINUSERNAME=${MONGO_INITDB_ROOT_USERNAME}, etc, because in that case I would also need a top-level .env file next to the docker-compose file to supply those variables.
I have changed ports bindings and added docs backend/docs/services/mongodb.md.
I apologize if i have misunderstood you somewhere.

[mehalyna]

The fixture is declared autouse=True, scope="session", so the MongoDB test DB is dropped only before and after the whole test session. Tests in tests/test_mongodb.py assume an empty DB per test (they assert count()==1). Running tests together will accumulate documents across tests and cause flakiness.
Change fixture to function-scoped or explicitly clean collections between tests. Example:

@pytest.fixture(autouse=True, scope="function")
def clean_test_mongo():
    ...
    client.drop_database(db_name)
    yield
    client.drop_database(db_name)

Or drop collections at the start of each test to preserve session-level setup time if needed.

[butuz0]

changed scope to "function"

[mehalyna]

The script runs migrate --database=mongodb immediately after PostgreSQL migrations without ensuring the MongoDB service is ready. This can cause startup failures in environments where MongoDB takes longer to start.
Add a wait-for-mongo step (a management command or loop checking connection) before running the mongodb migrate, or rely on docker-compose depends_on with healthchecks. Example snippet:

# wait_for_mongo (pseudo)
python manage.py wait_for_mongo --timeout=60 || exit 1
python manage.py migrate --database=mongodb --no-input || exit 1

Or implement a simple loop using mongo --eval "db.adminCommand('ping')".

[butuz0]

added wait_for_mongodb command