gmakstutis/LightRAG
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions
LightRAG/.env.unraid.example
Claude 9578d26b17
Add .env.unraid.example for Dell T140 server configuration 
This template is optimized for:
- Dell T140 (6-core Xeon E-2226G, 32GB RAM)
- Docker deployment on Unraid
- GPT-5-mini LLM
- Voyage-3-large embeddings (2048 dims)
- Jina reranker
- Neo4j graph storage + Postgres (pgvector) for vector/KV/doc storage
- Books, articles, and podcast transcripts use case

Performance settings tuned for ~40-50% CPU utilization during heavy processing.
2025-11-15 09:52:41 +00:00

249 lines
9.1 KiB
Text
Raw Blame History

###########################################################################################
### LightRAG Configuration - Optimized for Books, Articles, and Podcast Transcripts
###
### FILL IN THE FOLLOWING PLACEHOLDERS:
### - YOUR_OPENAI_API_KEY_HERE (for GPT-5-mini)
### - YOUR_VOYAGE_API_KEY_HERE (for voyage-3-large embeddings)
### - YOUR_JINA_API_KEY_HERE (for Jina reranker)
### - Neo4j connection details (URI, username, password, database)
### - Postgres connection details (host, port, user, password, database)
###########################################################################################
###########################
### Server Configuration
###########################
HOST=0.0.0.0
PORT=9621
WEBUI_TITLE='My Graph KB'
WEBUI_DESCRIPTION="Knowledge Graph for Books, Articles & Podcasts"
### Worker Configuration
### Optimized for Dell T140 (6-core Xeon E-2226G, 32GB RAM)
WORKERS=3
TIMEOUT=180
### Directory Configuration
### Uncomment and set if you want custom paths (otherwise defaults to ./inputs and ./rag_storage)
# INPUT_DIR=/path/to/your/inputs
# WORKING_DIR=/path/to/your/working_dir
### Logging
LOG_LEVEL=INFO
# LOG_DIR=/path/to/logs
#####################################
### Authentication (Optional)
#####################################
### Uncomment if you want to secure your API
# LIGHTRAG_API_KEY=your-secure-api-key-here
# AUTH_ACCOUNTS='admin:admin123'
# TOKEN_SECRET=Your-Secret-Key-For-JWT
# TOKEN_EXPIRE_HOURS=48
######################################################################################
### Query Configuration
### Optimized for long-form content (books, articles, transcripts)
######################################################################################
ENABLE_LLM_CACHE=true
### Increase these for better context with long documents
TOP_K=50
CHUNK_TOP_K=25
MAX_ENTITY_TOKENS=8000
MAX_RELATION_TOKENS=10000
MAX_TOTAL_TOKENS=40000
### Use VECTOR method for better semantic matching
KG_CHUNK_PICK_METHOD=VECTOR
#########################################################
### Reranking Configuration - JINA AI
#########################################################
RERANK_BINDING=jina
RERANK_BY_DEFAULT=true
RERANK_MODEL=jina-reranker-v2-base-multilingual
RERANK_BINDING_HOST=https://api.jina.ai/v1/rerank
RERANK_BINDING_API_KEY=YOUR_JINA_API_KEY_HERE
### Keep all chunks for reranking (let Jina do the filtering)
MIN_RERANK_SCORE=0.0
########################################
### Document Processing Configuration
########################################
ENABLE_LLM_CACHE_FOR_EXTRACT=true
SUMMARY_LANGUAGE=English
### Chunk size optimized for long-form content
CHUNK_SIZE=1200
CHUNK_OVERLAP_SIZE=100
### Entity types optimized for mixed content (books, articles, podcasts)
ENTITY_TYPES='["Person", "Organization", "Location", "Event", "Concept", "Method", "Theory", "Technology", "Product", "Research", "Topic"]'
### Summary configuration
FORCE_LLM_SUMMARY_ON_MERGE=8
SUMMARY_MAX_TOKENS=1500
SUMMARY_LENGTH_RECOMMENDED_=800
SUMMARY_CONTEXT_SIZE=15000
### Related chunks per entity/relation
RELATED_CHUNK_NUMBER=5
###############################
### Concurrency Configuration
### Optimized for Dell T140 (6-core Xeon E-2226G, 32GB RAM)
###############################
MAX_ASYNC=6
MAX_PARALLEL_INSERT=2
EMBEDDING_FUNC_MAX_ASYNC=12
EMBEDDING_BATCH_NUM=10
###########################################################################
### LLM Configuration - OpenAI GPT-5-mini
###########################################################################
LLM_BINDING=openai
LLM_MODEL=gpt-5-mini
LLM_BINDING_HOST=https://api.openai.com/v1
LLM_BINDING_API_KEY=YOUR_OPENAI_API_KEY_HERE
LLM_TIMEOUT=180
### GPT-5 configuration
OPENAI_LLM_MAX_COMPLETION_TOKENS=9000
OPENAI_LLM_TEMPERATURE=0.7
#######################################################################################
### Embedding Configuration - Voyage AI (voyage-3-large)
### ⚠️ IMPORTANT: Do NOT change after processing your first file!
#######################################################################################
EMBEDDING_BINDING=openai
EMBEDDING_MODEL=voyage-3-large
EMBEDDING_DIM=2048
EMBEDDING_SEND_DIM=false
EMBEDDING_TOKEN_LIMIT=32000
EMBEDDING_BINDING_HOST=https://api.voyageai.com/v1
EMBEDDING_BINDING_API_KEY=YOUR_VOYAGE_API_KEY_HERE
EMBEDDING_TIMEOUT=60
####################################################################
### Workspace Configuration (Optional)
### Uncomment to isolate data between different LightRAG instances
####################################################################
# WORKSPACE=my_knowledge_base
############################
### Storage Configuration
############################
### Using Neo4j for graph + Postgres for everything else
### Graph Storage
LIGHTRAG_GRAPH_STORAGE=Neo4JStorage
### Vector, KV, and Doc Status Storage (All in Postgres)
LIGHTRAG_VECTOR_STORAGE=PGVectorStorage
LIGHTRAG_KV_STORAGE=PGKVStorage
LIGHTRAG_DOC_STATUS_STORAGE=PGDocStatusStorage
###########################
### PostgreSQL Configuration
### FILL IN YOUR DETAILS:
### - POSTGRES_HOST: Docker container name (e.g., 'postgres') or IP address
### - POSTGRES_PORT: Usually 5432
### - POSTGRES_USER: Your postgres username
### - POSTGRES_PASSWORD: Your postgres password
### - POSTGRES_DATABASE: Database name (e.g., 'lightrag')
###########################
POSTGRES_HOST=YOUR_POSTGRES_HOST_HERE
POSTGRES_PORT=5432
POSTGRES_USER=YOUR_POSTGRES_USER_HERE
POSTGRES_PASSWORD='YOUR_POSTGRES_PASSWORD_HERE'
POSTGRES_DATABASE=YOUR_POSTGRES_DATABASE_HERE
POSTGRES_MAX_CONNECTIONS=20
### PostgreSQL Vector Storage - HNSW index (best for quality)
POSTGRES_VECTOR_INDEX_TYPE=HNSW
POSTGRES_HNSW_M=16
POSTGRES_HNSW_EF=200
### Connection retry configuration (good for Docker networking)
POSTGRES_CONNECTION_RETRIES=3
POSTGRES_CONNECTION_RETRY_BACKOFF=0.5
POSTGRES_CONNECTION_RETRY_BACKOFF_MAX=5.0
POSTGRES_POOL_CLOSE_TIMEOUT=5.0
###########################
### Neo4j Configuration
### FILL IN YOUR DETAILS:
### - NEO4J_URI: Connection string
### Examples:
### bolt://neo4j:7687 (Docker, no SSL)
### neo4j://neo4j:7687 (Docker, no SSL)
### neo4j+s://your-instance.databases.neo4j.io (Cloud with SSL)
### - NEO4J_USERNAME: Usually 'neo4j'
### - NEO4J_PASSWORD: Your neo4j password
### - NEO4J_DATABASE: Database name (usually 'neo4j')
###########################
NEO4J_URI=YOUR_NEO4J_URI_HERE
NEO4J_USERNAME=neo4j
NEO4J_PASSWORD='YOUR_NEO4J_PASSWORD_HERE'
NEO4J_DATABASE=neo4j
### Neo4j connection pool settings (optimized for your Dell T140)
NEO4J_MAX_CONNECTION_POOL_SIZE=75
NEO4J_CONNECTION_TIMEOUT=30
NEO4J_CONNECTION_ACQUISITION_TIMEOUT=30
NEO4J_MAX_TRANSACTION_RETRY_TIME=30
NEO4J_MAX_CONNECTION_LIFETIME=300
NEO4J_LIVENESS_CHECK_TIMEOUT=30
NEO4J_KEEP_ALIVE=true
###########################################################################################
### WORKER CONFIGURATION - OPTIMIZED FOR YOUR DELL T140
###########################################################################################
### Current settings optimized for:
### - CPU: Intel Xeon E-2226G (6 cores @ 3.40GHz)
### - RAM: 32GB (currently 65% utilized, ~11GB free)
### - Current load: 10-15% CPU usage
###
### Settings:
### WORKERS=3 - 3 Gunicorn server processes (good for 6-core CPU)
### MAX_ASYNC=6 - 6 concurrent LLM API requests (balanced for API limits)
### MAX_PARALLEL_INSERT=2 - 2 documents processed in parallel (prevents API overload)
### EMBEDDING_FUNC_MAX_ASYNC=12 - 12 concurrent embedding requests (embeddings are lighter)
###
### These settings utilize ~40-50% of your CPU during heavy processing, leaving plenty
### of headroom for your other Unraid services. API rate limits are the bottleneck, not
### your server hardware.
###
### If you experience OpenAI rate limit errors, reduce MAX_ASYNC to 4.
### If your server load stays very low (<20%), you can increase MAX_ASYNC to 8.
###########################################################################################
###########################################################################################
### QUICK START CHECKLIST
###########################################################################################
### 1. Fill in API keys:
### - YOUR_OPENAI_API_KEY_HERE (for GPT-5-mini)
### - YOUR_VOYAGE_API_KEY_HERE (for voyage-3-large embeddings)
### - YOUR_JINA_API_KEY_HERE (for reranking)
###
### 2. Fill in Postgres details:
### - POSTGRES_HOST (e.g., 'postgres' if Docker container name)
### - POSTGRES_USER
### - POSTGRES_PASSWORD
### - POSTGRES_DATABASE (create a database named 'lightrag' or similar)
###
### 3. Fill in Neo4j details:
### - NEO4J_URI (e.g., bolt://neo4j:7687 or neo4j://neo4j:7687)
### - NEO4J_PASSWORD
### - NEO4J_DATABASE (usually 'neo4j')
###
### 4. Verify your Postgres has pgvector extension enabled:
### Run in psql: CREATE EXTENSION IF NOT EXISTS vector;
###
### 5. Start LightRAG and verify it connects to both databases
###
### 6. (Optional) Adjust WORKERS/MAX_ASYNC based on your server specs (see guide above)
###
### 7. (Optional) Enable authentication by uncommenting LIGHTRAG_API_KEY
###########################################################################################
Reference in a new issue View git blame Copy permalink