mirror of
https://github.com/NVIDIA/dgx-spark-playbooks.git
synced 2026-04-23 02:23:53 +00:00
99 lines
3.2 KiB
YAML
99 lines
3.2 KiB
YAML
services:
|
|
app:
|
|
build:
|
|
context: ../..
|
|
dockerfile: deploy/app/Dockerfile
|
|
ports:
|
|
- '3001:3000'
|
|
environment:
|
|
- ARANGODB_URL=http://arangodb:8529
|
|
- ARANGODB_DB=txt2kg
|
|
- LANGCHAIN_TRACING_V2=true
|
|
- GRPC_SSL_CIPHER_SUITES=HIGH+ECDSA:HIGH+aRSA
|
|
- NODE_TLS_REJECT_UNAUTHORIZED=0
|
|
- OLLAMA_BASE_URL=http://ollama:11434/v1
|
|
- OLLAMA_MODEL=llama3.1:8b
|
|
- REMOTE_WEBGPU_SERVICE_URL=http://txt2kg-remote-webgpu:8083
|
|
- NVIDIA_API_KEY=${NVIDIA_API_KEY:-}
|
|
# Node.js timeout configurations for large model processing
|
|
- NODE_OPTIONS=--max-http-header-size=80000
|
|
- UV_THREADPOOL_SIZE=128
|
|
- HTTP_TIMEOUT=1800000
|
|
- REQUEST_TIMEOUT=1800000
|
|
networks:
|
|
- default
|
|
- txt2kg-network
|
|
depends_on:
|
|
- arangodb
|
|
- ollama
|
|
arangodb:
|
|
image: arangodb:latest
|
|
ports:
|
|
- '8529:8529'
|
|
environment:
|
|
- ARANGO_NO_AUTH=1
|
|
volumes:
|
|
- arangodb_data:/var/lib/arangodb3
|
|
- arangodb_apps_data:/var/lib/arangodb3-apps
|
|
arangodb-init:
|
|
image: arangodb:latest
|
|
depends_on:
|
|
arangodb:
|
|
condition: service_started
|
|
restart: on-failure
|
|
entrypoint: >
|
|
sh -c "
|
|
echo 'Waiting for ArangoDB to start...' &&
|
|
sleep 10 &&
|
|
echo 'Creating txt2kg database...' &&
|
|
arangosh --server.endpoint tcp://arangodb:8529 --server.authentication false --javascript.execute-string 'try { db._createDatabase(\"txt2kg\"); console.log(\"Database txt2kg created successfully!\"); } catch(e) { if(e.message.includes(\"duplicate\")) { console.log(\"Database txt2kg already exists\"); } else { throw e; } }'
|
|
"
|
|
ollama:
|
|
build:
|
|
context: ../services/ollama
|
|
dockerfile: Dockerfile
|
|
image: ollama-custom:latest
|
|
container_name: ollama-compose
|
|
ports:
|
|
- '11434:11434'
|
|
volumes:
|
|
- ollama_data:/root/.ollama
|
|
environment:
|
|
- OLLAMA_FLASH_ATTENTION=1 # Enable flash attention for better performance
|
|
- OLLAMA_KEEP_ALIVE=30m # Keep models loaded for 30 minutes
|
|
- OLLAMA_CUDA=1 # Enable CUDA acceleration
|
|
- OLLAMA_LLM_LIBRARY=cuda # Use CUDA library for LLM operations
|
|
- OLLAMA_NUM_PARALLEL=1 # Process one request at a time for 70B models
|
|
- OLLAMA_MAX_LOADED_MODELS=1 # Load only one model at a time to avoid VRAM contention
|
|
- OLLAMA_KV_CACHE_TYPE=q8_0 # Reduce KV cache VRAM usage with minimal performance impact
|
|
- OLLAMA_GPU_LAYERS=999 # Use maximum GPU layers
|
|
- OLLAMA_GPU_MEMORY_FRACTION=0.9 # Use 90% of GPU memory
|
|
- CUDA_VISIBLE_DEVICES=0 # Use GPU 0 (change to 'all' for multi-GPU)
|
|
networks:
|
|
- default
|
|
restart: unless-stopped
|
|
deploy:
|
|
resources:
|
|
reservations:
|
|
devices:
|
|
- driver: nvidia
|
|
count: all
|
|
capabilities: [gpu]
|
|
healthcheck:
|
|
test: ["CMD", "curl", "-f", "http://localhost:11434/api/tags"]
|
|
interval: 30s
|
|
timeout: 10s
|
|
retries: 3
|
|
start_period: 60s
|
|
|
|
volumes:
|
|
arangodb_data:
|
|
arangodb_apps_data:
|
|
ollama_data:
|
|
|
|
networks:
|
|
default:
|
|
driver: bridge
|
|
txt2kg-network:
|
|
driver: bridge
|