OpenAI-compatible connector

This connector exposes the bot through an API compatible with the OpenAI chat completions API. Any client supporting this API, such as Open WebUI, can then talk to the bot.

• Connector type: openai
• Sources and README: connector-open-ai

Configuration

Create an OpenAI connector in Tock Studio, then use its URL as base URL of the OpenAI API in the client, for instance http://<bot-host>/io/<namespace>/<bot>/openai.

The optional Web Security Mode field identifies the users as for the Web connector (see Security modes).

Example with Open WebUI:

docker run -d -p 3000:3000 -e PORT=3000 -e OFFLINE_MODE=True -e OPENAI_API_KEY=NONE \
  -e OPENAI_API_BASE_URL=http://host.docker.internal:8080/io/app/new_assistant/openai \
  -e ENABLE_FORWARD_USER_INFO_HEADERS=true -v open-webui:/app/backend/data \
  --name open-webui --restart always ghcr.io/open-webui/open-webui:main

Streaming

Answers are streamed by default. To also support non-streamed requests, set tock_openai_support_unstreamed to true.

Chat with Tock

×