open-sse

open-sse

Universal AI proxy library with SSE streaming support for OpenAI, Claude, Gemini and more.

Installation

npm install open-sse

Features

• Multi-provider support (OpenAI, Claude, Gemini, Copilot, Codex, etc.)
• SSE streaming for real-time responses
• Automatic format translation between providers
• Token refresh & OAuth management
• Account fallback handling
• Combo models (route across multiple providers)

Quick Start

Basic Usage

import { handleChatCore, getModelInfoCore } from "open-sse";

async function handleChat(request) {
  const body = await request.json();
  
  // Get model info (auto-detect provider)
  const modelInfo = await getModelInfoCore(body.model);
  
  // Provider credentials
  const credentials = {
    provider: modelInfo.provider,
    accessToken: "your-token"
  };
  
  // Handle chat with auto translation & streaming
  return await handleChatCore(body, credentials, console);
}

With Token Refresh

import { handleChatCore, isTokenExpiringSoon, refreshTokenByProvider } from "open-sse";

async function handleChat(request, credentials) {
  const body = await request.json();
  
  // Auto refresh if expiring
  if (isTokenExpiringSoon(credentials)) {
    const newTokens = await refreshTokenByProvider(
      credentials.provider,
      credentials,
      console
    );
    credentials = { ...credentials, ...newTokens };
  }
  
  return await handleChatCore(body, credentials, console);
}

Format Translation

import { translateRequest, translateResponse } from "open-sse";

// OpenAI → Claude
const claudeRequest = await translateRequest(openAIRequest, "openai", "claude");

// Claude → OpenAI
const openAIResponse = await translateResponse(claudeResponse, "claude", "openai");

Combo Models

import { handleComboChat } from "open-sse/services/combo.js";

const models = [
  { provider: "claude", model: "claude-3-5-sonnet-20241022" },
  { provider: "openai", model: "gpt-4" }
];

const response = await handleComboChat(request, models, getCredentials, console);

Main Exports

// Handlers
import { handleChatCore, isTokenExpiringSoon } from "open-sse";

// Services
import { getModelInfoCore, parseModel } from "open-sse";
import { buildProviderUrl, buildProviderHeaders, detectFormat } from "open-sse";
import { refreshTokenByProvider, refreshClaudeOAuthToken } from "open-sse";
import { checkFallbackError, isAccountUnavailable } from "open-sse";

// Translation
import { translateRequest, translateResponse, needsTranslation } from "open-sse";

// Utils
import { errorResponse } from "open-sse";
import { createSSETransformStreamWithLogger } from "open-sse";

Configuration

Environment Variables

# Enable detailed request/response logging (default: false)
ENABLE_REQUEST_LOGS=true

When enabled, logs are saved to logs/ directory with structure:

logs/
  └── {sourceFormat}_{targetFormat}_{model}_{timestamp}/
      ├── 0_client_raw_request.json
      ├── 1_raw_request.json
      ├── 2_converted_request.json
      ├── 3_raw_response.json
      └── 4_converted_response.json

Provider Models

import { PROVIDER_MODELS, getProviderModels } from "open-sse";

const claudeModels = getProviderModels("claude");

Constants

import { PROVIDERS, OAUTH_ENDPOINTS, CACHE_TTL } from "open-sse";

Examples

Next.js API Route

// app/api/chat/route.js
import { handleChatCore, getModelInfoCore } from "open-sse";

export async function POST(request) {
  const body = await request.json();
  const modelInfo = await getModelInfoCore(body.model);
  
  const credentials = {
    provider: modelInfo.provider,
    accessToken: process.env.API_TOKEN
  };
  
  return await handleChatCore(body, credentials, console);
}

Express.js

import express from "express";
import { handleChatCore, getModelInfoCore } from "open-sse";

const app = express();

app.post("/api/chat", async (req, res) => {
  const modelInfo = await getModelInfoCore(req.body.model);
  
  const credentials = {
    provider: modelInfo.provider,
    accessToken: process.env.API_TOKEN
  };
  
  const response = await handleChatCore(req.body, credentials, console);
  return res.send(response);
});

Supported Providers

• OpenAI (GPT-4, GPT-3.5)
• Anthropic Claude (Sonnet, Opus, Haiku)
• Google Gemini
• GitHub Copilot
• Codex
• Qwen

License

MIT