.env.example

# Vector Graph RAG Configuration
# Copy this file to .env and fill in your actual values
# The .env file is gitignored and will not be committed

# ============================================================================
# API Server Configuration
# ============================================================================
# API server port (default: 8000)
# VGRAG_API_PORT=8000

# ============================================================================
# OpenAI Configuration (Required)
# ============================================================================
# Your OpenAI API key for LLM and embeddings
OPENAI_API_KEY=your_openai_api_key_here

# Optional: Custom OpenAI API base URL (for compatible APIs)
# OPENAI_BASE_URL=https://api.openai.com/v1

# ============================================================================
# Milvus Configuration
# ============================================================================
# Milvus connection URI
# For Milvus Lite (local file): ./vector_graph_rag.db
# For Milvus server: http://localhost:19530
# For Zilliz Cloud: https://your-endpoint.zillizcloud.com:443
MILVUS_URI=./vector_graph_rag.db

# Optional: Milvus authentication token (for Zilliz Cloud)
# MILVUS_TOKEN=your_milvus_token_here

# Optional: Milvus database name
# MILVUS_DB=default

# ============================================================================
# Model Configuration (Optional - uses VGRAG_ prefix)
# ============================================================================
# LLM model for triplet extraction and reranking
# VGRAG_LLM_MODEL=gpt-4o-mini

# Embedding model (HuggingFace model name or OpenAI model name)
# VGRAG_EMBEDDING_MODEL=facebook/contriever

# Embedding dimension (768 for contriever, 1536 for text-embedding-3-small)
# VGRAG_EMBEDDING_DIMENSION=768

# ============================================================================
# Milvus Index Configuration (Optional)
# ============================================================================
# Milvus index type (AUTOINDEX, FLAT, IVF_FLAT, HNSW, etc.)
# VGRAG_MILVUS_INDEX_TYPE=AUTOINDEX

# Milvus consistency level (Strong, Bounded, Session, Eventually)
# VGRAG_MILVUS_CONSISTENCY_LEVEL=Bounded

# ============================================================================
# Retrieval Configuration (Optional)
# ============================================================================
# Number of top entities to retrieve
# VGRAG_ENTITY_TOP_K=10

# Number of top relations to retrieve
# VGRAG_RELATION_TOP_K=10

# Entity similarity threshold (keep if score > threshold)
# VGRAG_ENTITY_SIMILARITY_THRESHOLD=0.9

# Relation similarity threshold (keep if score > threshold, -1 keeps all)
# VGRAG_RELATION_SIMILARITY_THRESHOLD=-1.0

# Degree of subgraph expansion (1 or 2 recommended)
# VGRAG_EXPANSION_DEGREE=1

# Number of final passages to return
# VGRAG_FINAL_TOP_K=3

# ============================================================================
# LLM Settings (Optional)
# ============================================================================
# Temperature for LLM generation
# VGRAG_LLM_TEMPERATURE=0.0

# Maximum retries for LLM API calls
# VGRAG_LLM_MAX_RETRIES=3

# ============================================================================
# Processing Settings (Optional)
# ============================================================================
# Batch size for embedding and insertion
# VGRAG_BATCH_SIZE=100

# ============================================================================
# Collection Configuration (Optional)
# ============================================================================
# Prefix for collection names (useful for multiple datasets)
# VGRAG_COLLECTION_PREFIX=ds_2wikimultihopqa

# Collection names (usually don't need to change)
# VGRAG_ENTITY_COLLECTION=vgrag_entities
# VGRAG_RELATION_COLLECTION=vgrag_relations
# VGRAG_PASSAGE_COLLECTION=vgrag_passages