Version: dev

Environment Variables

Reference for commonly used environment variables in DB-GPT.

tip

Most configuration is done through TOML config files. Environment variables are useful for secrets, Docker deployments, and overriding defaults.

Model configuration

Variable	Description	Example
`OPENAI_API_KEY`	OpenAI API key	`sk-...`
`OPENAI_API_BASE`	OpenAI-compatible API base URL	`https://api.openai.com/v1`
`DEEPSEEK_API_KEY`	DeepSeek API key	`sk-...`
`SILICONFLOW_API_KEY`	SiliconFlow API key	`sk-...`
`DASHSCOPE_API_KEY`	Alibaba Cloud DashScope API key	`sk-...`
`AIMLAPI_API_KEY`	AI/ML API key	—
`LLM_MODEL`	Default LLM model name (cluster mode)	`glm-4-9b-chat`
`MODEL_SERVER`	Model controller address (cluster mode)	`http://127.0.0.1:8000`

Variable	Description	Example
`CUDA_VISIBLE_DEVICES`	Restrict which GPUs are visible	`0,1`
`DEVICE`	Force device type	`cuda`, `cpu`, `mps`

Variable	Description	Example
`UV_INDEX_URL`	PyPI mirror for uv	`https://pypi.tuna.tsinghua.edu.cn/simple`
`HTTP_PROXY`	HTTP proxy for outbound requests	`http://proxy:8080`
`HTTPS_PROXY`	HTTPS proxy for outbound requests	`http://proxy:8080`
`NO_PROXY`	Hosts to bypass proxy	`localhost,127.0.0.1`

DB-GPT supports environment variable substitution in TOML config files:

[[models.llms]]
api_key = "${env:OPENAI_API_KEY}"

[[models.embeddings]]
api_key = "${env:OPENAI_API_KEY:-default-key}"

Syntax:

Pattern	Behavior
`${env:VAR_NAME}`	Read from environment variable (error if missing)
`${env:VAR_NAME:-default}`	Read from environment, use `default` if not set

When running with Docker, pass environment variables with -e:

docker run -it --rm \
  -e SILICONFLOW_API_KEY=your-key \
  -e DBGPT_LOG_LEVEL=DEBUG \
  -p 5670:5670 \
  eosphorosai/dbgpt-openai

Or with Docker Compose in docker-compose.yml:

services:
  webserver:
    environment:
      - SILICONFLOW_API_KEY=${SILICONFLOW_API_KEY}
      - DBGPT_LOG_LEVEL=INFO