docker-compose.yml

services:

  llm: &llm
    image: ollama/ollama:latest
    profiles: ["linux"]
    networks:
      - net
    volumes:
      - $PWD/ollama:/root/.ollama
      - $PWD/ollama_entrypoint.sh:/entrypoint.sh
    entrypoint: ["/usr/bin/bash", "/entrypoint.sh"]
    environment:
      - LLM=${LLM-llama3.1}
    tty: true

  llm-gpu-nvidia:
    <<: *llm
    profiles: ["linux-gpu-nvidia"]
    environment:
      - LLM=${LLM-llama3.1}
    deploy:
      resources:
        reservations:
          devices:
            - driver: nvidia
              count: all
              capabilities: [gpu]
    devices:
      - "/dev/nvidiactl:/dev/nvidiactl"
      - "/dev/nvidia-uvm:/dev/nvidia-uvm"
      - "/dev/nvidia-uvm-tools:/dev/nvidia-uvm-tools"
      - "/dev/nvidia-modeset:/dev/nvidia-modeset"
      - "/dev/nvidia0:/dev/nvidia0"
    volumes:
      - $PWD/ollama:/root/.ollama
      - $PWD/ollama_entrypoint.sh:/entrypoint.sh
            
  llm-gpu-amd:
    <<: *llm
    profiles: ["linux-gpu-amd"]
    image: ollama/ollama:rocm
    environment:
      - HSA_OVERRIDE_GFX_VERSION=11.0.2
      - LLM=${LLM-llama3.1}
      - OLLAMA_DEBUG=1
    devices:
      - "/dev/kfd:/dev/kfd"
      - "/dev/dri:/dev/dri"
    volumes:
      - $PWD/ollama:/root/.ollama
      - $PWD/ollama_entrypoint.sh:/entrypoint.sh

  database:
    user: "${UID}:${GID}"
    image: neo4j:${NEO4J_VERSION-5.25.1}
    ports:
      - 7687:7687
      - 7474:7474
    volumes:
      - $PWD/data:/data
    environment:
      - NEO4J_AUTH=${NEO4J_USERNAME-neo4j}/${NEO4J_PASSWORD-password}
      - NEO4J_PLUGINS=["apoc"]
      - NEO4J_db_tx__log_rotation_retention__policy=false
      - NEO4J_dbms_security_procedures_unrestricted=apoc.*
    healthcheck:
        test: ["CMD-SHELL", "wget --no-verbose --tries=1 --spider localhost:7474 || exit 1"]
        interval: 15s
        timeout: 30s
        retries: 10
    depends_on:
      database-seeding:
        condition: service_completed_successfully
    networks:
      - net

  database-seeding:
    user: "${UID}:${GID}"
    image: neo4j/neo4j-admin:${NEO4J_VERSION-5.21.2}
    volumes:
      - $PWD/data:/data
      - $PWD/backups:/backups
      - $PWD/seed-databases.sh:/seed-databases.sh
    environment:
      - NEO4J_ADDRESS=neo4j://database
      - NEO4J_USERNAME=${NEO4J_USERNAME-neo4j}
      - NEO4J_PASSWORD=${NEO4J_PASSWORD-password}
    command: /seed-databases.sh
    networks:
      - net

  loader: &loader
    build:
      context: ./build-context/
      dockerfile: Dockerfile
    volumes:
      - $PWD/embedding_model:/embedding_model
    environment: &environment
      NEO4J_URI: ${NEO4J_URI-neo4j://database:7687}
      NEO4J_PASSWORD: ${NEO4J_PASSWORD-password}
      NEO4J_USERNAME: ${NEO4J_USERNAME-neo4j}
      NEO4J_DATABASE: ${NEO4J_DATABASE-neo4j}
      OPENAI_API_KEY: ${OPENAI_API_KEY-}
      GOOGLE_API_KEY: ${GOOGLE_API_KEY-}
      OLLAMA_BASE_URL: ${OLLAMA_BASE_URL-http://host.docker.internal:11434}
      LLM: ${LLM-llama2}
      EMBEDDING_MODEL: ${EMBEDDING_MODEL-sentence_transformer}
      LANGCHAIN_ENDPOINT: ${LANGCHAIN_ENDPOINT-"https://api.smith.langchain.com"}
      LANGCHAIN_TRACING_V2: ${LANGCHAIN_TRACING_V2-false}
      LANGCHAIN_PROJECT: ${LANGCHAIN_PROJECT-}
      LANGCHAIN_API_KEY: ${LANGCHAIN_API_KEY-}
      AWS_ACCESS_KEY_ID: ${AWS_ACCESS_KEY_ID-}
      AWS_SECRET_ACCESS_KEY: ${AWS_SECRET_ACCESS_KEY-}
      AWS_DEFAULT_REGION: ${AWS_DEFAULT_REGION-}
      STREAMLIT_FILE: loader.py
      PREPARATION_QUERY: |
        MATCH (r:Regesta) 
        CALL {
          WITH r
          SET r.concatText = apoc.text.join([x in [r.summary, r.commentary, r.archivalHistory] where x is not null], ' ')
        } IN CONCURRENT TRANSACTIONS
      LABEL: Regesta
      PROPERTY_IDENTIFIER: uid
      PROPERTY_TEXT: concatText
      PROPERTY_EMBEDDING: embedding
      PROMPT_CONTEXT: den Regesta Imperii
      RETRIEVAL_QUERY: |
        WITH node, score, 
          COLLECT { MATCH (node)-[:EXTERNAL_SOURCE|REFERENCES]->(ref) RETURN apoc.text.join([x IN [ref.shortTitle, ref.url] WHERE x IS NOT NULL], ' ') } AS refs, 
          COLLECT { MATCH (node)-[:PLACE_OF_ISSUE]->(p:Place) RETURN apoc.text.join([x IN [p.latLong, p.normalizedGerman, p.wikidataId] WHERE x IS NOT NULL], ' ') } AS placesOfIssue,
          COLLECT { MATCH (node)<-[:APPEARS_IN]->(ref) RETURN apoc.text.join([x IN [ref.type, ref.label] WHERE x IS NOT NULL], ' ') } AS appearsIn
        RETURN '### Text:\n' + node.concatText + '\n' +
               '### Orte:\n' + apoc.text.join(placesOfIssue, '\n') +
               '### Referenzen:\n' + apoc.text.join(refs, '\n') + apoc.text.join(appearsIn, '\n') as text,           
            score,
            { source:  coalesce(node.url, 'n/a') } AS metadata  
        ORDER BY score ASC
    ports:
      - 8602:8080
    networks:
    - net
    depends_on:
      database:
        condition: service_healthy
      database-seeding:
        condition: service_completed_successfully
    develop:
      watch:
      - action: rebuild
        path: .
        ignore:
        - pdf_bot.py
        - front-end/

  bot:
    <<: *loader
    environment: 
      <<: *environment
      STREAMLIT_FILE: bot.py
    ports:
      - 8601:8080

  api:
    <<: *loader
    # environment:
    #   <<: *environment
    # volumes:
    #   - $PWD/embedding_model:/embedding_model
    entrypoint:  [ "uvicorn", "api:app", "--host", "0.0.0.0", "--port", "8080" ]
    healthcheck:
      test: ["CMD-SHELL", "curl --fail http://localhost:8080"]
      interval: 5s
      timeout: 3s
      retries: 5
    ports:
      - 8604:8080

networks:
  net: