K2Think AI Client & API Proxy

OpenAI-compatible client library and API proxy for K2Think AI platform. Use it as a library (like OpenAI SDK) or as HTTP API proxy - no manual authentication needed!

Two Ways to Use

1️⃣ As a Library (No Server Needed!)

const K2ThinkClient = require('k2think-api-nodejs');
const client = new K2ThinkClient();

const response = await client.chat.completions.create({
  model: 'MBZUAI-IFM/K2-Think-v2',  // Use v2 model
  messages: [{ role: 'user', content: 'Hello!' }]
});

2️⃣ As HTTP API Proxy

npm start
curl -X POST http://localhost:3000/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{"model":"MBZUAI-IFM/K2-Think-v2","messages":[{"role":"user","content":"Hello"}]}'

3️⃣ Multi-Account Mode (NEW!)

Automatic account rotation to avoid rate limits:

const K2ThinkMultiClient = require('./client_multi');
const client = new K2ThinkMultiClient();

// Add accounts via CLI: node accounts-cli.js add

// Rotation happens automatically!
const response = await client.chat.completions.create({
  model: 'MBZUAI-IFM/K2-Think-v2',
  messages: [{ role: 'user', content: 'Hello!' }]
});
console.log(`Used account: ${response._accountId}`);

See docs/MULTI_ACCOUNT.md for full guide.

Features

• Zero-configuration: Set credentials once in .env
• OpenAI-compatible: Same API as OpenAI
• Automatic authentication: Token management handled automatically
• Context preservation: Multi-turn conversations work perfectly
• Two usage modes: Library or HTTP API
• Multi-account support: Automatic rotation to avoid rate limits (NEW!)

Installation

npm install

Prerequisites

Get K2Think Account

Before using this library, you need a K2Think account:

1. Sign up: https://www.k2think.ai/auth?mode=signup
2. Login: https://www.k2think.ai/auth
3. Save your email and password for configuration

Configure Credentials

Create .env file from .env.example:

cp .env.example .env

Edit .env with your K2Think credentials:

K2THINK_EMAIL=your-email@example.com
K2THINK_PASSWORD=your-password
PORT=3000

2. Use in your code

require('dotenv').config();
const K2ThinkClient = require('./client');

const client = new K2ThinkClient();

// Simple question
async function ask(question) {
  const response = await client.chat.completions.create({
    model: 'MBZUAI-IFM/K2-Think-v2',  // Updated model name
    messages: [{ role: 'user', content: question }]
  });
  return response.choices[0].message.content;
}

const answer = await ask('What is 2+2?');
console.log(answer);

That's it! No server needed.

Quick Start - API Proxy Mode

1. Configure credentials (same .env)

2. Start server

npm start

3. Use HTTP API

Simple Chat

curl -X POST http://localhost:3000/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{
    "model": "MBZUAI-IFM/K2-Think-v2",
    "messages": [
      {"role": "user", "content": "Hello! What is 2+2?"}
    ]
  }'

Chat with Context

curl -X POST http://localhost:3000/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{
    "model": "MBZUAI-IFM/K2-Think-v2",
    "messages": [
      {"role": "user", "content": "My name is Alice"},
      {"role": "assistant", "content": "Nice to meet you, Alice!"},
      {"role": "user", "content": "What is my name?"}
    ]
  }'

List Available Models

curl http://localhost:3000/v1/models

API Endpoints

• GET / - Health check
• POST /v1/chat/completions - Chat completions (OpenAI compatible)
• GET /v1/models - List available models (OpenAI compatible)

Which Mode to Use?

Library Mode (client.js)

Best for:

• Node.js applications and scripts
• Direct integration in your code
• When you don't need HTTP API

Example:

const client = new K2ThinkClient();
await client.chat.completions.create({...});

API Proxy Mode (index.js)

Best for:

• Web applications (frontend)
• Multiple services/languages accessing the API
• When you need REST API

Example:

curl http://localhost:3000/v1/chat/completions ...

How It Works

1. Automatic Authentication: Logs in to K2Think API using credentials from .env
2. Token Management: Automatically refreshes expired tokens
3. Request Translation: Converts OpenAI-format requests to K2Think format
4. Response Formatting: Returns responses in OpenAI-compatible format

Response Format

Standard OpenAI-compatible format:

{
  "id": "chatcmpl-...",
  "object": "chat.completion",
  "created": 1762533698,
  "model": "MBZUAI-IFM/K2-Think",
  "choices": [{
    "index": 0,
    "message": {
      "role": "assistant",
      "content": "Response here"
    },
    "finish_reason": "stop"
  }],
  "usage": {
    "prompt_tokens": 431,
    "completion_tokens": 69,
    "total_tokens": 500
  }
}

Testing

Run the test suite:

# Test library mode
node test_library.js

# Test library with integration tests
node tests/library_integration.test.js

# Test API proxy mode (start server first)
npm start &
node tests/api_integration.test.js

See docs/TEST_RESULTS.md for detailed test results.

Troubleshooting

• Authentication errors: Check credentials in .env
• Model not found: Use MBZUAI-IFM/K2-Think-v2 (not MBZUAI-IFM/K2-Think)
• Port already in use: Change PORT in .env or stop other service on port 3000
• No response: Check K2Think service availability and internet connection
• Proxy errors: The library bypasses system proxy automatically

Available Models

• MBZUAI-IFM/K2-Think-v2 - Latest K2 Think model with enhanced reasoning

Files

• client.js - Library mode - K2ThinkClient class for direct use
• client_multi.js - Multi-account mode - Automatic account rotation
• index.js - API proxy mode - HTTP server with endpoints
• auth_manager.js - Authentication and token management
• accounts-cli.js - CLI utility for managing multiple accounts
• quick-add.js - Quick account addition script (recommended)
• lib/account_manager.js - Account storage and management
• lib/account_rotator.js - Account rotation logic
• examples/ - Usage examples for library and multi-account modes
• tests/ - Integration tests for library and API modes
• test_library.js - Test for library mode
• data/accounts.json - Multi-account storage (encrypted)
• docs/TEST_RESULTS.md - Detailed test results and troubleshooting
• docs/MULTI_ACCOUNT.md - Multi-account system guide
• .env - Configuration (create from .env.example)

Documentation

• LIBRARY.md - Complete library documentation with examples
• USAGE.md - HTTP API usage guide
• QUICKSTART.md - Quick start guide (Russian)
• MULTI_ACCOUNT.md - Multi-account system guide
• ACCOUNT_SCHEMA.md - JSON storage schema

Quick Commands

# Add new account (recommended)
node quick-add.js email@example.com password123 "Account Name"

# List all accounts
node accounts-cli.js list

# View statistics
node accounts-cli.js stats

# Test multi-account system
node tests/test_multi_account.js

# Run example
node examples/multi_account_example.js

Support

If you find this project useful, consider supporting its development:

🪙 USDT (BNB Smart Chain)

0x30C93CEB10c53db8B01ae311db83C2287C431ECd

🪙 USDT (TON Network)

UQDUm3wVVrkcdHgqAZnEIXRbtPwt9KV52M20C6vsMiheKmKV

🪙 USDT (Tron / TRC20)

TJLpsWFGkr26hbpRdHxHwMMzNjUECWKSQc

🪙 USDT (Ethereum)

0x80dCc2DA8ad2A8283F63AAaD94dD490373a48885

🪙 TON (TON Network)

UQDUm3wVVrkcdHgqAZnEIXRbtPwt9KV52M20C6vsMiheKmKV

See SUPPORT.md for more ways to help.

License

MIT