Skip to content

Configuration Reference

Complete reference for all configuration options

Environment Variables

All configuration is managed through environment variables loaded from .env file.

Core Settings

ENVIRONMENT

  • Type: str
  • Default: development
  • Options: development, staging, production
  • Description: Application environment mode
ENVIRONMENT=production

LOG_LEVEL

  • Type: str
  • Default: INFO
  • Options: DEBUG, INFO, WARNING, ERROR, CRITICAL
  • Description: Logging verbosity level
LOG_LEVEL=INFO

API_RATE_LIMIT

  • Type: str
  • Default: 30/minute
  • Format: {count}/{time_unit}
  • Description: Rate limiting for API endpoints
API_RATE_LIMIT=100/minute

LLM Configuration

LLM_PROVIDER

  • Type: str
  • Default: openai
  • Options: openai, azure, bedrock, anthropic
  • Description: LLM service provider
LLM_PROVIDER=openai

LLM_MODEL

  • Type: str
  • Default: gpt-5-mini
  • Options: Provider-specific model names
  • Description: LLM model to use for generation

OpenAI Models: 

LLM_MODEL=gpt-5-mini              # Recommended: Fast, cheap
LLM_MODEL=gpt-5                   # Higher quality
LLM_MODEL=gpt-4o                  # Best quality, expensive

Azure OpenAI Models: 

LLM_MODEL=gpt-5-mini              # Deployment name

AWS Bedrock Models: 

LLM_MODEL=anthropic.claude-3-sonnet-20240229-v1:0
LLM_MODEL=anthropic.claude-3-haiku-20240307-v1:0
LLM_MODEL=amazon.titan-text-express-v1

LLM_TEMPERATURE

  • Type: float
  • Default: 0.2
  • Range: 0.0 - 2.0
  • Description: Response randomness (lower = more deterministic)
LLM_TEMPERATURE=0.2

LLM_MAX_TOKENS

  • Type: int
  • Default: 4000
  • Description: Maximum tokens in LLM response
LLM_MAX_TOKENS=4000

OpenAI Settings

OPENAI_API_KEY

  • Type: str
  • Required: Yes (if LLM_PROVIDER=openai)
  • Description: OpenAI API key
OPENAI_API_KEY=sk-proj-...

OPENAI_ORG_ID

  • Type: str
  • Optional: Yes
  • Description: OpenAI organization ID
OPENAI_ORG_ID=org-...

Azure OpenAI Settings

AZURE_OPENAI_ENDPOINT

  • Type: str
  • Required: Yes (if LLM_PROVIDER=azure)
  • Format: https://{resource-name}.openai.azure.com
  • Description: Azure OpenAI endpoint URL
AZURE_OPENAI_ENDPOINT=https://my-resource.openai.azure.com

AZURE_OPENAI_API_KEY

  • Type: str
  • Required: Yes (if LLM_PROVIDER=azure)
  • Description: Azure OpenAI API key
AZURE_OPENAI_API_KEY=abc123...

AZURE_OPENAI_DEPLOYMENT

  • Type: str
  • Required: Yes (if LLM_PROVIDER=azure)
  • Description: Deployment name for the model
AZURE_OPENAI_DEPLOYMENT=gpt-5-mini

AZURE_OPENAI_API_VERSION

  • Type: str
  • Default: 2023-05-15
  • Description: Azure OpenAI API version
AZURE_OPENAI_API_VERSION=2023-05-15

AWS Bedrock Settings

AWS_REGION

  • Type: str
  • Default: us-east-1
  • Description: AWS region for Bedrock
AWS_REGION=us-east-1

AWS_ACCESS_KEY_ID

  • Type: str
  • Required: Yes (if LLM_PROVIDER=bedrock)
  • Description: AWS access key ID
AWS_ACCESS_KEY_ID=AKIA...

AWS_SECRET_ACCESS_KEY

  • Type: str
  • Required: Yes (if LLM_PROVIDER=bedrock)
  • Description: AWS secret access key
AWS_SECRET_ACCESS_KEY=...

Anthropic Settings

ANTHROPIC_API_KEY

  • Type: str
  • Required: Yes (if LLM_PROVIDER=anthropic)
  • Description: Anthropic API key
ANTHROPIC_API_KEY=sk-ant-...

Vector Store Configuration

VECTOR_STORE_TYPE

  • Type: str
  • Default: faiss
  • Options: faiss, qdrant
  • Description: Vector store backend
VECTOR_STORE_TYPE=qdrant

QDRANT_URL

  • Type: str
  • Default: http://localhost:6333
  • Required: Yes (if VECTOR_STORE_TYPE=qdrant)
  • Description: Qdrant server URL
QDRANT_URL=http://qdrant:6333

QDRANT_API_KEY

  • Type: str
  • Optional: Yes
  • Description: Qdrant API key (if authentication enabled)
QDRANT_API_KEY=...

QDRANT_COLLECTION_NAME

  • Type: str
  • Default: greengovrag
  • Description: Qdrant collection name
QDRANT_COLLECTION_NAME=greengovrag

FAISS_INDEX_PATH

  • Type: str
  • Default: data/vectors/faiss_index
  • Description: Path to FAISS index file
FAISS_INDEX_PATH=data/vectors/faiss_index

Embedding Configuration

EMBEDDING_MODEL

  • Type: str
  • Default: sentence-transformers/all-MiniLM-L6-v2
  • Description: Embedding model name
EMBEDDING_MODEL=sentence-transformers/all-MiniLM-L6-v2
EMBEDDING_MODEL=sentence-transformers/all-mpnet-base-v2

EMBEDDING_DIMENSIONS

  • Type: int
  • Default: 384
  • Description: Embedding vector dimensions
EMBEDDING_DIMENSIONS=384

Database Configuration

DATABASE_URL

  • Type: str
  • Required: Yes
  • Format: postgresql://{user}:{password}@{host}:{port}/{database}
  • Description: PostgreSQL connection string
DATABASE_URL=postgresql://greengovrag:password@localhost:5432/greengovrag

DATABASE_POOL_SIZE

  • Type: int
  • Default: 20
  • Description: Database connection pool size
DATABASE_POOL_SIZE=20

DATABASE_MAX_OVERFLOW

  • Type: int
  • Default: 10
  • Description: Max overflow connections beyond pool size
DATABASE_MAX_OVERFLOW=10

Cloud Storage Configuration

CLOUD_PROVIDER

  • Type: str
  • Default: local
  • Options: local, aws, azure
  • Description: Cloud storage provider
CLOUD_PROVIDER=aws

AWS_S3_BUCKET

  • Type: str
  • Required: Yes (if CLOUD_PROVIDER=aws)
  • Description: S3 bucket name for document storage
AWS_S3_BUCKET=greengovrag-documents

AZURE_STORAGE_CONNECTION_STRING

  • Type: str
  • Required: Yes (if CLOUD_PROVIDER=azure)
  • Description: Azure Blob Storage connection string
AZURE_STORAGE_CONNECTION_STRING=DefaultEndpointsProtocol=https;...

AZURE_STORAGE_CONTAINER

  • Type: str
  • Default: greengovrag
  • Description: Azure Blob Storage container name
AZURE_STORAGE_CONTAINER=greengovrag

Caching Configuration

CACHE_ENABLED

  • Type: bool
  • Default: true
  • Description: Enable query result caching
CACHE_ENABLED=true

CACHE_TTL_SECONDS

  • Type: int
  • Default: 3600
  • Description: Cache time-to-live in seconds
CACHE_TTL_SECONDS=3600

REDIS_URL

  • Type: str
  • Optional: Yes
  • Format: redis://{host}:{port}/{db}
  • Description: Redis connection string for caching
REDIS_URL=redis://localhost:6379/0

ETL Pipeline Configuration

CHUNK_SIZE

  • Type: int
  • Default: 500
  • Description: Text chunk size in tokens
CHUNK_SIZE=500

CHUNK_OVERLAP

  • Type: int
  • Default: 100
  • Description: Overlap between chunks in tokens
CHUNK_OVERLAP=100

CHUNK_BATCH_SIZE

  • Type: int
  • Default: 100
  • Description: Number of chunks to process in parallel
CHUNK_BATCH_SIZE=100

DOCUMENTS_CONFIG_PATH

  • Type: str
  • Default: configs/documents_config.yml
  • Description: Path to document sources configuration
DOCUMENTS_CONFIG_PATH=configs/documents_config.yml

RAW_DATA_DIR

  • Type: str
  • Default: data/raw
  • Description: Directory for downloaded documents
RAW_DATA_DIR=data/raw

PROCESSED_DATA_DIR

  • Type: str
  • Default: data/processed
  • Description: Directory for processed chunks
PROCESSED_DATA_DIR=data/processed

RAG Configuration

TOP_K_RESULTS

  • Type: int
  • Default: 5
  • Description: Number of documents to retrieve
TOP_K_RESULTS=5

MIN_RELEVANCE_SCORE

  • Type: float
  • Default: 0.3
  • Range: 0.0 - 1.0
  • Description: Minimum relevance score to include document
MIN_RELEVANCE_SCORE=0.3
  • Type: bool
  • Default: true
  • Description: Enable BM25 + vector hybrid search
ENABLE_HYBRID_SEARCH=true

BM25_WEIGHT

  • Type: float
  • Default: 0.3
  • Range: 0.0 - 1.0
  • Description: Weight for BM25 score in hybrid search
BM25_WEIGHT=0.3

CORS Configuration

CORS_ORIGINS

  • Type: list[str]
  • Default: ["http://localhost:3000"]
  • Format: Comma-separated URLs
  • Description: Allowed CORS origins
CORS_ORIGINS=http://localhost:3000,https://app.greengovrag.com

CORS_ALLOW_CREDENTIALS

  • Type: bool
  • Default: true
  • Description: Allow credentials in CORS requests
CORS_ALLOW_CREDENTIALS=true

Configuration Profiles

Development Profile

# .env.development
ENVIRONMENT=development
LOG_LEVEL=DEBUG
VECTOR_STORE_TYPE=faiss
DATABASE_URL=postgresql://greengovrag:greengovrag@localhost:5432/greengovrag
CLOUD_PROVIDER=local
LLM_PROVIDER=openai
LLM_MODEL=gpt-5-mini
CACHE_ENABLED=false

Production Profile

# .env.production
ENVIRONMENT=production
LOG_LEVEL=INFO
VECTOR_STORE_TYPE=qdrant
QDRANT_URL=http://qdrant:6333
DATABASE_URL=postgresql://greengovrag:${DB_PASSWORD}@rds-endpoint:5432/greengovrag
CLOUD_PROVIDER=aws
AWS_S3_BUCKET=greengovrag-prod-documents
LLM_PROVIDER=azure
LLM_MODEL=gpt-5-mini
CACHE_ENABLED=true
CACHE_TTL_SECONDS=3600

Configuration Loading

Configuration is loaded in this order (later sources override earlier):

  1. Default values in backend/green_gov_rag/config.py
  2. Environment variables from .env file
  3. System environment variables (highest priority)

Example: 

from green_gov_rag.config import settings

# Access configuration
print(settings.llm_provider)  # 'openai'
print(settings.llm_model)      # 'gpt-5-mini'

Validation

Configuration is validated on startup using Pydantic:

from pydantic import BaseSettings, validator

class Settings(BaseSettings):
    llm_provider: str = "openai"

    @validator("llm_provider")
    def validate_provider(cls, v):
        allowed = ["openai", "azure", "bedrock", "anthropic"]
        if v not in allowed:
            raise ValueError(f"LLM provider must be one of {allowed}")
        return v

Validation errors will prevent the application from starting.

Secrets Management

Local Development

Use .env file (never commit to Git):

# .env
OPENAI_API_KEY=sk-...
DATABASE_URL=postgresql://...

Production (AWS)

Use AWS Systems Manager Parameter Store:

aws ssm put-parameter \
  --name "/greengovrag/prod/openai-api-key" \
  --value "sk-..." \
  --type "SecureString"

Production (Azure)

Use Azure Key Vault:

az keyvault secret set \
  --vault-name greengovrag-kv \
  --name openai-api-key \
  --value "sk-..."

Last Updated: 2025-11-22