Settings reference
Environment variables that configure Chain. All are optional unless stated otherwise. Start from the .env template in your Easy8 package.
Embeddings
Set these when you use your own TEI (Text Embeddings Inference) server.
| Variable |
Default |
Description |
EMBEDDINGS_URL |
(none) |
Base URL of the TEI (Text Embeddings Inference) server used for vector generation. |
TEI_API_KEY |
(none) |
Optional API key for the TEI server. When set, it is sent as a Bearer token in the Authorization header of every request to the TEI server. |
Supported embeddings model
For self-hosted deployments the supported embeddings model is BAAI/bge-m3, served through TEI on NVIDIA GPU infrastructure. CPU-only deployment is not supported. For the GPU memory requirement, see Embeddings hardware recommendations.
LLM providers
These variables apply to cloud-provider scenarios. Self-hosted on-premises LLM deployments use LLM_URL in the scenario guides.
| Variable |
Default |
Description |
OPENAI_API_KEY |
(none) |
Enables OpenAI LLMs and embeddings models. |
AZURE_OPENAI_API_KEY |
(none) |
Enables Azure OpenAI LLMs and embeddings. Requires AZURE_OPENAI_ENDPOINT. |
AZURE_OPENAI_ENDPOINT |
(none) |
Azure OpenAI service endpoint URL. |
OPENAI_EMBEDDINGS_API_KEY |
(none) |
Separate OpenAI API key used only for embeddings, for deployments that use non-OpenAI LLMs but OpenAI embeddings. |
MISTRAL_API_KEY |
(none) |
Enables MistralAI hosted models. |
LLM_URL |
(none) |
Base URL of the self-hosted TGI or vLLM server. |
VLLM_API_KEY |
(none) |
Optional API key for vLLM. When set, it is sent as a Bearer token in the Authorization header. |
VLLM_MODEL_ALIASES |
(none) |
Comma-separated SERVED_MODEL_ID=CHAIN_LLM_ID mappings for deployment-specific vLLM served model ids. |
DEFAULT_LLM_ID |
(none) |
Override the default LLM model ID. |
Infrastructure
| Variable |
Default |
Description |
CHROMADB_URL |
http://chroma:8000 |
URL of the ChromaDB vector database. |
CELERY_BROKER_URL |
redis://redis:6379/0 |
Redis URL used as Celery message broker. |
CELERY_RESULT_BACKEND |
redis://redis:6379/0 |
Redis URL used for Celery task results. |
CHAIN_DATA_PATH |
/data |
Directory for persistent data (configurations, temporary files). |
Security
| Variable |
Default |
Description |
EASY_JWT_SECRET |
(none) |
JWT secret for authenticating HTTP endpoints. When unset, JWT authentication is disabled. |
EASY_AI_CORS_ALLOWED_ORIGINS |
(none) |
Comma-separated list of allowed CORS origins. In debug mode, all origins are allowed. |
CORS_DISABLED |
false |
Set to true to disable CORS middleware entirely. |
REQUESTS_CA_BUNDLE |
(none) |
Path to a custom CA bundle used by Python requests clients. |
SSL_CERT_FILE |
(none) |
Path to a custom CA bundle used by Python SSL, httpx, OpenAI SDK, and LangChain clients. |
CURL_CA_BUNDLE |
(none) |
Path to a custom CA bundle used by curl-compatible tooling and some HTTP client integrations. |
Observability
| Variable |
Default |
Description |
LANGCHAIN_TRACING_V2 |
true (when API key is set) |
Enable or disable LangSmith tracing. |
LANGCHAIN_API_KEY |
(none) |
LangSmith API key. Setting this enables tracing. |
LANGCHAIN_PROJECT |
clients |
LangSmith project name for trace grouping. |
LANGCHAIN_ENDPOINT |
https://api.smith.langchain.com |
LangSmith API endpoint. |
Debug
| Variable |
Default |
Description |
CHAIN_DEBUG_MODE |
false |
Enable debug mode (true or 1). |