Skip to content

OpenAI.md

Latest commit

 

History

History
86 lines (65 loc) · 5.63 KB

OpenAI.md

File metadata and controls

86 lines (65 loc) · 5.63 KB

OpenAIProvider Configuration Guide

The OpenaiProvider lets you connect your chatbot to OpenAI’s LLM, either directly via the public API or through a proxy. Note that configurations can vary greatly between providers; this guide covers the available configurations for the OpenaiProvider. You may find other providers’ guides here.

ℹ️ Info:
This configuration guide assumes you have completed the setup for the LlmConnector plugin according to the guide here.

1. Import Provider

import { OpenaiProvider } from "@rcb-plugins/llm-connector";

2. Basic Instantiation

The OpenaiProvider supports two modes: "direct" or "proxy". Here’s a minimal example for "direct" mode:

const openai = new OpenaiProvider({
  mode: "direct",                     // "direct" or "proxy"
  model: "gpt-4",                     // required
  apiKey: process.env.OPENAI_API_KEY, // required in "direct" mode
  responseFormat: "stream",           // "stream" (default) or "json"
});

⚠️ Warning: You are strongly discouraged from using mode: "direct" in production, which risk exposing your API keys on the client side. The "direct" mode is meant to simplify testing, but is typically unsafe in production. You should instead be looking to use "proxy" mode with API key secured server-side. You may refer to this lightweight LLM proxy project if you would like a ready-to-go example.

3. Configuration Options

Option Type Required Default Description
mode "direct" | "proxy" ✅ always — Selects direct or proxy mode.
model string ✅ always — The OpenAI model to call.
apiKey string ✅ if mode: "direct" — Your OpenAI API key (only with mode: "direct").
baseUrl string ✅ if mode: "proxy" https://api.openai.com/v1/chat/completions Override the base URL (must be provided in proxy mode).
method string ❌ "POST" HTTP method for the request.
headers Record<string, string> ❌ {} Additional HTTP headers.
body Record<string, string> ❌ {} Additional HTTP body.
systemMessage string ❌ null Prepends a system prompt to every conversation.
responseFormat "stream" | "json" ❌ "stream" Determines whether to use stream endpoint from the provider or fetch a full JSON output.
messageParser (msgs: Message[]) => CustomMessage[] ❌ null Custom parser converting React ChatBotify Message[] into desired message format for the provider.
debug boolean ❌ false Enables debug logging for the provider.

4. Advanced Example

const openai = new OpenaiProvider({
  mode: "proxy",
  model: "gpt-4.1-nano",
  baseUrl: "https://my-proxy.example.com/chat/completions",
  method: "POST",
  headers: {
    "X-Custom-Header": "value",
  },
  body: {
    temperature: "0.7",
    max_tokens: "500",
  },
  systemMessage: "You are a helpful assistant that translates English to French.",
  responseFormat: "stream",
  messageParser: (msgs) => msgs.map((m) => ({ role: m.sender.toLowerCase(), content: String(m.content) })),
});

5. How It Works Under the Hood

  1. Constructor: sets defaults such as endpoint.
  2. sendMessages():
    • Fetches endpoint with JSON body.
      • Streams via SSE if responseFormat === "stream".
      • Otherwise returns full text from JSON.

Check out other providers here.