BF Streaming API

The BetterForms Streaming API allows real time streaming of responses from LLM services like OpenAI and Google Gemini to be initiated.

Overview

Creating a call to this service will create a post to the LLM and stream results back to the appropriate channel. This service is in beta release as of the time of this document.

Avail in bf-staging only

//create endpoint portal.myapp.com//stream/create

Method: POST

Key	Type	Description
apiKey	string	Unique API key for BF (find in app settings)
channels	array	Array of BF Messaging communication channels
actionName	string	Name of the action to be performed
payload	object	Object containing parameters for the action, this contains params that are passed on to the streaming service
payload.provider	string	Name of the AI provider to be used (e.g., "openai", "gemini"). Defaults to "openai" if not specified.
payload.apiKey	string	Unique API key for the selected AI provider
payload.stream	bool	Indicator if streaming is enabled. If false, result is returned directly in the POST response.
payload.messages	array	Array of messages to be processed. Structure may vary slightly based on the provider.
payload.model	string	Model to be used for processing (e.g., "gpt-4-turbo-preview" for OpenAI, "gemini-pro" for Gemini)
payload.generationConfig	object	(Optional - Gemini specific) Object containing generation parameters for the Gemini provider.

Note on OpenAI Parameters: Parameters like functions, max_tokens, seed, and temperature previously documented for OpenAI are not explicitly passed to the OpenAI SDK in the current service implementation. They might be subject to the OpenAI SDK's default behaviors or could be added in future updates.

{
  "apiKey": "BFAPI_xxxxxxxx-xxxxxx-xxxxxx",
  "channels": [
    "anonymous"
  ],
  "actionName": "assistantReceiveResultsStream",
  "payload": {
    "provider": "openai", // or "gemini"
    "apiKey": "sk-xxxxxxx-xxxxxxxx-xxxxxxx", // Your OpenAI or Gemini API Key
    "stream": true,
    "messages": [
      {
        "content": "You are a helpful assistant.",
        "role": "system" // For Gemini, 'system' role might need to be adapted or handled as part of 'history'
      },
      {
        "content": "Tell me a fun fact about space.",
        "role": "user"
      }
    ],
    "model": "gpt-4-turbo-preview", // or a Gemini model like "gemini-pro"
    // "generationConfig": { // Example for Gemini
    //   "temperature": 0.7,
    //   "maxOutputTokens": 2048
    // }
  }
}