Using FastGPT

Connecting FastGPT to Tutujin API takes two steps: tell FastGPT where to send requests and which key to use (the request URL https://open.tutujin.com/v1 plus your Tutujin API key), then enable the models you want in model configuration, with model IDs matching the names in the Tutujin model hub.

Method 1: add a model channel in the admin pages

For versions that manage model channels in the web interface:

1. Sign in as an admin and open the model provider (or model channel) settings.
2. Add a channel with the OpenAI protocol type and fill in:

3. In model configuration, make sure those models are enabled. For a model that is not listed, add a custom model with the model name as its ID, e.g. deepseek-v4.1-flash.
4. Save; the models now appear in your apps' AI settings.

Method 2: environment variables at deploy time

If you deploy with docker compose and your version supports these variables, set the default request URL and key on the FastGPT service:

services:
  fastgpt:
    environment:
      - OPENAI_BASE_URL=https://open.tutujin.com/v1
      - CHAT_API_KEY=sk-your-key

Restart FastGPT, then enable the models you need in model configuration, with IDs matching the model hub names.

Embedding models for knowledge bases

FastGPT knowledge bases need an embedding model. In model configuration, add or enable an embedding (index) model such as text-embedding-3-small; it uses the same channel as your chat models. Once a knowledge base exists, avoid switching embedding models, or you will have to re-embed all of its data.

Testing and billing

Send a message from an app; a normal reply means the setup works. Every call is recorded under Usage in the Tutujin console and billed at the key's group price. A dedicated key for FastGPT with a spending limit is a good idea.

Troubleshooting