Local Model Integration Guide

April 2, 2023 ยท View on GitHub

Bibliothecarius initialization provides the chat model from OpenAI for dialogue, but you can also introduce your own model. Please refer to the following instructions.

1.Implement Local Interface

Bibliothecarius initiates a request:

  • POST

  • Content-Type: application/json

  • Parameters:

    {
	"input":"user's question after assembly",
  // Historical record
  "chatContextList":[
    {
      // Sorting
      "sort": 0,
      "user": "user's question",
      "assistant": "AI output"
    }
  ],
  // User configuration parameters
  "params":{}
}

Your local service needs to return a result:

string

2. Integration

You can create the interface using API or a visual interface.

Creating an Interface Using API

Please refer to the API documentation

Request interface POST:/external/model

Parameters:

{
    "chatAddress": "chat interface address",
    "name": "name (unique and cannot be duplicated with system built-ins)",
    "checkParametersAddress": "parameter checking interface (if empty, parameters will not be checked)",
    "remark": "description",
  	// Maximum allowed input request
    "inputMaxToken": 53
}

Creating an Interface Using a Visual Interface

#TODO Front-end implementation is pending

3.Example