Server Config

We have django-web-server, postgres, cassandra, qdrant services running in the server. Here postgres, cassandra, qdrant are the database thats needed for this application.

Note:
It also includes ml-service in the development environment. This is a simulated ml-service, since every time we can't have access or test in the prod environment. This ml-service have two endpoints as in real prod where every time it gives the same response.
Important : Always use `Makefile` commands. You can have a look at it for more clarity. Any command attached with the `-dev` or `-prod` is specific to that environment.

  • For dev use the docker-compose.yaml commands when you are working on dev environment. So start the server with docker compose. Starting individual components produce some errors.

  • For prod env, I recommend you to use docker-swarm.yaml over the docker-compose.prod.yaml

Example .env file

The below code block is the example how the .env file looks like. This should be in the root directory of server $ROOT/server. This should be present for the server to run (both in dev & prod environments)

POSTGRES_DB=megacad
POSTGRES_USER=megacad
POSTGRES_PASSWORD=postgres
POSTGRES_HOST=postgres
CASSANDRA_HOST=cassandra
CASSANDRA_USER=cassandra
CASSANDRA_PASSWORD=cassandra
CASSANDRA_CLUSTER_NAME=cassandra
CASSANDRA_ENDPOINT_SNITCH=GossipingPropertyFileSnitch
CASSANDRA_KEYSPACE=megacad
QDRANT_COLLECTION_NAME=megacad
QDRANT_HOST=qdrant
DJANGO_SECRET_KEY=django
DJANGO_ADMIN_USERNAME=admin
DJANGO_ADMIN_EMAIL_ID=admin
DJANGO_ADMIN_PASSWORD=admin
ML_SERVICE_HOST=ml_service
ML_SERVICE_PORT=5000
ROOT_PATH=
WEB_ADDRESS=megacad.com
DJANGO_HOST=django
LLM_ENDPOINT=/v1/llm
EMBEDDING_ENDPOINT=/v1/embed

Build

Since the server is running on a ppc64le architecture, many of the python packages wont directly get installed. So the docker files for the each service which don't support the ppc64le environments are build from the source. This is not hte case for the dev environment. dev environment is build keeping mac-os and linux-os in the mind. So it can handle x64 builds.

Qdrant-DB is running as x64 container on top of the QEMU virtualizer, even in the production environment. Anyone could try to improve the qemu overhead by try writing a dockerfile code to build the qdrant-db form the source that works on the ppc64le architecture