Environment Variables
RedLite uses several environment variables for configuration. This document lists all environment variables used throughout the project and when they are needed.
Core RedLite Variables
REDLITE_DATA_DIR
Purpose: Specifies the directory where RedLite stores its data files, including run metadata and cached results.
Default: ~/.cache/redlite (user's home cache directory)
When needed: Optional. Only set this if you want to change the default data storage location.
PORT
Purpose: Specifies the port number for the RedLite Web UI.
Default: 8000
When needed: Optional. Only needed when running the RedLite Web UI and you want to use a different port than the default.
API Keys for AI Models
OPENAI_API_KEY
Purpose: API key for accessing OpenAI models (GPT-3.5, GPT-4, etc.)
When needed: Required when using OpenAIModel for text generation tasks.
GEMINI_API_KEY
Purpose: API key for accessing Gogole Gemini models
When needed: Required when using GeminiModel for text generation tasks.
HF_TOKEN
Purpose: Hugging Face authentication token for accessing models and datasets from the Hugging Face Hub.
When needed: Required when using HFModel with private models or models that require authentication. Also needed for higher rate limits.
NGC_API_KEY
Purpose: NVIDIA NGC (NVIDIA GPU Cloud) API key for accessing NVIDIA's LLaMA NIM (NVIDIA Inference Microservice) models.
When needed: Required when using NVIDIA's LLaMA NIM service for text generation.
AWS Credentials
AWS_ACCESS_KEY_ID
Purpose: AWS access key ID for authenticating with AWS Bedrock services.
When needed: Required when using AwsBedrockModel for accessing AWS Bedrock text generation models.
AWS_SECRET_ACCESS_KEY
Purpose: AWS secret access key for authenticating with AWS Bedrock services.
When needed: Required when using AwsBedrockModel for accessing AWS Bedrock text generation models.
Zeno Integration
ZENO_API_KEY
Purpose: API key for uploading evaluation results to Zeno platform for visualization and analysis.
When needed: Required when using the Zeno upload functionality (redlite upload command or redlite.zeno.upload() function).