.env.example

# AI PDF Search Engine Configuration

# ========================================
# REQUIRED API KEYS
# ========================================

# OpenAI API Key - Get from https://platform.openai.com/api-keys
OPENAI_API_KEY=your_openai_api_key_here

# Pinecone API Key - Get from https://app.pinecone.io/
PINECONE_API_KEY=your_pinecone_api_key_here
PINECONE_INDEX_NAME=pdf-search-engine

# Supabase Configuration - Get from https://supabase.com/
SUPABASE_URL=your_supabase_project_url
SUPABASE_KEY=your_supabase_anon_key

# ========================================
# EMBEDDING CONFIGURATION
# ========================================

# OpenAI Embedding Model Settings
EMBEDDING_MODEL=text-embedding-3-small
EMBEDDING_DIMENSION=512

# Document Processing Settings
CHUNK_SIZE=1000
CHUNK_OVERLAP=200

# ========================================
# RAG OPTIMIZATION SETTINGS
# ========================================

# Document Relevance Threshold (0.0-1.0)
# Higher values = more strict document matching
# Lower values = more permissive document matching
SIMILARITY_THRESHOLD=0.7

# Maximum documents to retrieve per query
# More documents = better context but slower responses
MAX_DOCUMENTS_PER_QUERY=5

# LLM Response Settings
# Temperature: 0 = deterministic, 1 = creative
LLM_TEMPERATURE=0

# Maximum tokens in response (affects response length)
MAX_TOKENS=1000

# ========================================
# CHATBOT INTERFACE SETTINGS
# ========================================

# Maximum chat messages to keep in memory
# Higher values = better conversation context but more memory usage
MAX_CHAT_HISTORY=20

# Whether to display source documents in the UI
ENABLE_SOURCE_DISPLAY=true

# ========================================
# OPTIONAL: PINECONE ENVIRONMENT
# ========================================

# Pinecone Environment (usually auto-detected)
PINECONE_ENVIRONMENT=gcp-starter