跳到主要内容
版本:v0.8.3

Synthorai

Synthorai serves models from Anthropic, OpenAI, Google, DeepSeek, Qwen, Moonshot and Z.ai behind a single OpenAI-compatible endpoint and API key.​

This section describes how to use the Synthorai provider with DB-GPT.​

  1. Sign up at Synthorai and generate an API key.
  2. Set the environment variable SYNTHORAI_API_KEY with your key.
  3. Use the configs/dbgpt-proxy-synthorai.toml configuration when starting DB-GPT.

Model ids are bare rather than vendor-prefixed — claude-opus-5, gpt-5.6-sol, deepseek-v4-pro, glm-5.2 — so set LLM_MODEL_NAME to the id exactly as the catalog lists it.

You can look up models at https://synthorai.io/models/​

Or you can use docker/base/Dockerfile to run DB-GPT with Synthorai:​

# Expose the port for the web server, if you want to run it directly from the Dockerfile
EXPOSE 5670

# Just uncomment the following line in the `Dockerfile` to use Synthorai:
CMD ["dbgpt", "start", "webserver", "--config", "configs/dbgpt-proxy-synthorai.toml"]

Pass the key in at run time rather than baking it into the image with ENV — a key set at build time stays in the image layers and travels with anyone who pulls it:

docker run -it --rm -e SYNTHORAI_API_KEY="your-key" -p 5670:5670 dbgpt:latest