docker-compose.yaml

include:
  - path:
    - docker-compose-vectordb.yaml
    - docker-compose-nim-ms.yaml

services:
  chain-server:
    container_name: chain-server
    image: chain-server:${TAG:-latest}
    build:
      context: ./
      dockerfile: src/chain_server/Dockerfile
      args:
        APP_PATH: 'app_chain'
    volumes:
      - ./prompt.yaml:/prompt.yaml
    command: --port 8081 --host 0.0.0.0
    environment:
      APP_PATH: 'app_chain'
      APP_VECTORSTORE_URL: "http://milvus:19530"
      APP_VECTORSTORE_NAME: "milvus"
      APP_LLM_MODELNAME: ${APP_LLM_MODELNAME:-"meta/llama3-70b-instruct"}
      APP_LLM_MODELENGINE: nvidia-ai-endpoints
      APP_LLM_SERVERURL: ${APP_LLM_SERVERURL:-""}
      APP_EMBEDDINGS_MODELNAME: ${APP_EMBEDDINGS_MODELNAME:-nvidia/nv-embedqa-e5-v5}
      APP_EMBEDDINGS_MODELENGINE: ${APP_EMBEDDINGS_MODELENGINE:-nvidia-ai-endpoints}
      APP_EMBEDDINGS_SERVERURL: ${APP_EMBEDDINGS_SERVERURL:-""}
      APP_TEXTSPLITTER_MODELNAME: Snowflake/snowflake-arctic-embed-l
      APP_TEXTSPLITTER_CHUNKSIZE: 506
      APP_TEXTSPLITTER_CHUNKOVERLAP: 200
      APP_RANKING_MODELNAME: ${APP_RANKING_MODELNAME:-"nvidia/nv-rerankqa-mistral-4b-v3"} # Leave it blank to avoid using ranking
      APP_RANKING_MODELENGINE: ${APP_RANKING_MODELENGINE:-nvidia-ai-endpoints}
      APP_RANKING_SERVERURL: ${APP_RANKING_SERVERURL:-""}
      NVIDIA_API_KEY: ${NVIDIA_API_KEY}
      APP_RETRIEVER_TOPK: 4
      APP_RETRIEVER_SCORETHRESHOLD: 0.25
      POSTGRES_PASSWORD: ${POSTGRES_PASSWORD:-password}
      POSTGRES_USER: ${POSTGRES_USER:-postgres}
      POSTGRES_DB: ${POSTGRES_DB:-api}
      COLLECTION_NAME: ${COLLECTION_NAME:-ifu_rag}
      OTEL_EXPORTER_OTLP_ENDPOINT: http://otel-collector:4317
      OTEL_EXPORTER_OTLP_PROTOCOL: grpc
      LOGLEVEL: ${LOGLEVEL:-INFO}
    ports:
    - "8081:8081"
    expose:
    - "8081"
    shm_size: 5gb
    depends_on:
      nemollm-embedding:
        condition: service_healthy
        required: false
      nemollm-inference:
        condition: service_healthy
        required: false
      ranking-ms:
        condition: service_healthy
        required: false


  rag-playground:
    container_name: rag-playground
    image: rag-playground:${TAG:-latest}
    build:
      context: src/rag_playground/
      dockerfile: Dockerfile
      args:
        PLAYGROUND_MODE: ${PLAYGROUND_MODE:-speech}
    command: --port 8090
    environment:
      APP_SERVERURL: http://chain-server
      APP_SERVERPORT: 8081
      APP_MODELNAME: ${APP_LLM_MODELNAME:-"meta/llama3-70b-instruct"}
      OTEL_EXPORTER_OTLP_ENDPOINT: http://otel-collector:4317
      OTEL_EXPORTER_OTLP_PROTOCOL: grpc
      ENABLE_TRACING: false
      # if locally hosting Riva:
      #RIVA_API_URI: <riva-ip-address>:50051
      #TTS_SAMPLE_RATE: 48000
      # if using Riva API Endpoint on NVIDIA API Catalog
      RIVA_API_URI: grpc.nvcf.nvidia.com:443
      NVIDIA_API_KEY: ${NVIDIA_API_KEY}
      RIVA_ASR_FUNCTION_ID: 1598d209-5e27-4d3c-8079-4751568b1081
      RIVA_TTS_FUNCTION_ID: 0149dedb-2be8-4195-b9a0-e57e0e14f972
      TTS_SAMPLE_RATE: 48000

    ports:
    - "8090:8090"
    expose:
    - "8090"
    depends_on:
    - chain-server

networks:
  default:
    name: nvidia-rag