🧠 NeuroLink

Enterprise AI Development Platform with universal provider support, factory pattern architecture, and access to 100+ AI models through LiteLLM integration. Production-ready with TypeScript support.

NeuroLink is an Enterprise AI Development Platform that unifies 11 major AI providers with intelligent fallback and built-in tool support. Available as both a programmatic SDK and professional CLI tool. Features LiteLLM integration for 100+ models, plus 6 core tools working across all providers. Extracted from production use at Juspay.

🎉 NEW: LiteLLM Integration - Access 100+ AI Models

NeuroLink now supports LiteLLM, providing unified access to 100+ AI models from all major providers through a single interface:

• 🔄 Universal Access: OpenAI, Anthropic, Google, Mistral, Meta, and more
• 🎯 Unified Interface: OpenAI-compatible API for all models
• 💰 Cost Optimization: Automatic routing to cost-effective models
• ⚡ Load Balancing: Automatic failover and load distribution
• 📊 Analytics: Built-in usage tracking and monitoring

# Quick start with LiteLLM
pip install litellm && litellm --port 4000

# Use any of 100+ models through one interface
npx @juspay/neurolink generate "Hello" --provider litellm --model "openai/gpt-4o"
npx @juspay/neurolink generate "Hello" --provider litellm --model "anthropic/claude-3-5-sonnet"
npx @juspay/neurolink generate "Hello" --provider litellm --model "google/gemini-2.0-flash"

📖 Complete LiteLLM Integration Guide - Setup, configuration, and 100+ model access

🎉 NEW: SageMaker Integration - Deploy Your Custom AI Models

NeuroLink now supports Amazon SageMaker, enabling you to deploy and use your own custom trained models through NeuroLink's unified interface:

• 🏗️ Custom Model Hosting - Deploy your fine-tuned models on AWS infrastructure
• 💰 Cost Control - Pay only for inference usage with auto-scaling capabilities
• 🔒 Enterprise Security - Full control over model infrastructure and data privacy
• ⚡ Performance - Dedicated compute resources with predictable latency
• 📊 Monitoring - Built-in CloudWatch metrics and logging

# Quick start with SageMaker
export AWS_ACCESS_KEY_ID="your-access-key"
export AWS_SECRET_ACCESS_KEY="your-secret-key"
export SAGEMAKER_DEFAULT_ENDPOINT="your-endpoint-name"

# Use your custom deployed models
npx @juspay/neurolink generate "Analyze this data" --provider sagemaker
npx @juspay/neurolink sagemaker status  # Check endpoint health
npx @juspay/neurolink sagemaker benchmark my-endpoint  # Performance testing

📖 Complete SageMaker Integration Guide - Setup, deployment, and custom model access

🚀 Enterprise Platform Features

• 🏭 Factory Pattern Architecture - Unified provider management through BaseProvider inheritance
• 🔧 Tools-First Design - All providers include built-in tool support without additional configuration
• 🔗 LiteLLM Integration - 100+ models from all major providers through unified interface
• 🏗️ Enterprise Architecture - Production-ready with clean abstractions
• 🔄 Configuration Management - Flexible provider configuration with automatic backups
• ✅ Type Safety - Industry-standard TypeScript interfaces
• ⚡ Performance - Fast response times with streaming support and 68% improved status checks
• 🛡️ Error Recovery - Graceful failures with provider fallback and retry logic
• 📊 Analytics & Evaluation - Built-in usage tracking and AI-powered quality assessment
• 🎯 Real-time Event Monitoring - EventEmitter integration for progress tracking and debugging
• 🔧 MCP Integration - Model Context Protocol with 6 built-in tools and 58+ discoverable servers
• 🚀 Lighthouse Integration - Unified tool registration API supporting both object and array formats for seamless Lighthouse tool import

🚀 Quick Start

Install & Run (2 minutes)

# Option 1: LiteLLM - Access 100+ models through one interface
pip install litellm && litellm --port 4000
export LITELLM_BASE_URL="http://localhost:4000"
export LITELLM_API_KEY="sk-anything"

# Use any of 100+ models
npx @juspay/neurolink generate "Hello, AI" --provider litellm --model "openai/gpt-4o"
npx @juspay/neurolink generate "Hello, AI" --provider litellm --model "anthropic/claude-3-5-sonnet"

# Option 2: OpenAI Compatible - Use any OpenAI-compatible endpoint with auto-discovery
export OPENAI_COMPATIBLE_BASE_URL="https://api.openrouter.ai/api/v1"
export OPENAI_COMPATIBLE_API_KEY="sk-or-v1-your-api-key"
# Auto-discovers available models via /v1/models endpoint
npx @juspay/neurolink generate "Hello, AI" --provider openai-compatible

# Or specify a model explicitly
export OPENAI_COMPATIBLE_MODEL="claude-3-5-sonnet"
npx @juspay/neurolink generate "Hello, AI" --provider openai-compatible

# Option 3: Direct Provider - Quick setup with Google AI Studio (free tier)
export GOOGLE_AI_API_KEY="AIza-your-google-ai-api-key"
npx @juspay/neurolink generate "Hello, AI" --provider google-ai

# Option 3: Amazon SageMaker - Use your custom deployed models
export AWS_ACCESS_KEY_ID="your-access-key"
export AWS_SECRET_ACCESS_KEY="your-secret-key"
export SAGEMAKER_DEFAULT_ENDPOINT="your-endpoint-name"
npx @juspay/neurolink generate "Hello, AI" --provider sagemaker

# CLI Commands - No installation required
npx @juspay/neurolink generate "Explain AI"  # Auto-selects best provider
npx @juspay/neurolink gen "Write code"       # Shortest form
npx @juspay/neurolink stream "Tell a story" # Real-time streaming
npx @juspay/neurolink status                # Check all providers

# SDK Installation for using in your typescript projects
npm install @juspay/neurolink

Basic Usage

import { NeuroLink, AIProviderFactory } from "@juspay/neurolink";

// LiteLLM - Access 100+ models through unified interface
const litellmProvider = await AIProviderFactory.createProvider(
  "litellm",
  "openai/gpt-4o",
);
const result = await litellmProvider.generate({
  input: { text: "Write a haiku about programming" },
});

// Compare multiple models simultaneously
const models = [
  "openai/gpt-4o",
  "anthropic/claude-3-5-sonnet",
  "google/gemini-2.0-flash",
];
const comparisons = await Promise.all(
  models.map(async (model) => {
    const provider = await AIProviderFactory.createProvider("litellm", model);
    const result = await provider.generate({
      input: { text: "Explain quantum computing" },
    });
    return { model, response: result.content, provider: result.provider };
  }),
);

// Auto-select best available provider
const neurolink = new NeuroLink();
const autoResult = await neurolink.generate({
  input: { text: "Write a business email" },
  provider: "google-ai", // or let it auto-select
  timeout: "30s",
});

console.log(result.content);
console.log(`Used: ${result.provider}`);

Conversation Memory

NeuroLink supports automatic conversation history management that maintains context across multiple turns within sessions. This enables AI to remember previous interactions and provide contextually aware responses. Session-based memory isolation ensures privacy between different conversations.

// Enable conversation memory with configurable limits
const neurolink = new NeuroLink({
  conversationMemory: {
    enabled: true,
    maxSessions: 50, // Keep last 50 sessions
    maxTurnsPerSession: 20, // Keep last 20 turns per session
  },
});

🔗 CLI-SDK Consistency (NEW! ✨)

Method aliases that match CLI command names:

// All three methods are equivalent:
const result1 = await provider.generate({ input: { text: "Hello" } }); // Original
const result2 = await provider.generate({ input: { text: "Hello" } }); // Matches CLI 'generate'
const result3 = await provider.gen({ input: { text: "Hello" } }); // Matches CLI 'gen'

// Use whichever style you prefer:
const provider = createBestAIProvider();

// Detailed method name
const story = await provider.generate({
  input: { text: "Write a short story about AI" },
  maxTokens: 200,
});

// CLI-style method names
const poem = await provider.generate({ input: { text: "Write a poem" } });
const joke = await provider.gen({ input: { text: "Tell me a joke" } });

Enhanced Features

CLI with Analytics & Evaluation

# Basic AI generation with auto-provider selection
npx @juspay/neurolink generate "Write a business email"

# LiteLLM with specific model
npx @juspay/neurolink generate "Write code" --provider litellm --model "anthropic/claude-3-5-sonnet"

# With analytics and evaluation
npx @juspay/neurolink generate "Write a proposal" --enable-analytics --enable-evaluation --debug

# Streaming with tools (default behavior)
npx @juspay/neurolink stream "What time is it and write a file with the current date"

SDK with LiteLLM and Enhancement Features

import { NeuroLink, AIProviderFactory } from "@juspay/neurolink";

// LiteLLM multi-model comparison
const models = [
  "openai/gpt-4o",
  "anthropic/claude-3-5-sonnet",
  "google/gemini-2.0-flash",
];
const comparisons = await Promise.all(
  models.map(async (model) => {
    const provider = await AIProviderFactory.createProvider("litellm", model);
    return await provider.generate({
      input: { text: "Explain the benefits of renewable energy" },
      enableAnalytics: true,
      enableEvaluation: true,
    });
  }),
);

// Enhanced generation with analytics
const neurolink = new NeuroLink();
const result = await neurolink.generate({
  input: { text: "Write a business proposal" },
  enableAnalytics: true, // Get usage & cost data
  enableEvaluation: true, // Get AI quality scores
  context: { project: "Q1-sales" },
});

console.log("📊 Usage:", result.analytics);
console.log("⭐ Quality:", result.evaluation);
console.log("Response:", result.content);

Environment Setup

# Create .env file (automatically loaded by CLI)
echo 'OPENAI_API_KEY="sk-your-openai-key"' > .env
echo 'GOOGLE_AI_API_KEY="AIza-your-google-ai-key"' >> .env
echo 'AWS_ACCESS_KEY_ID="your-aws-access-key"' >> .env

# Test configuration
npx @juspay/neurolink status

JSON Format Support (Complete)

NeuroLink provides comprehensive JSON input/output support for both CLI and SDK:

# CLI JSON Output - Structured data for scripts
npx @juspay/neurolink generate "Summary of AI trends" --format json
npx @juspay/neurolink gen "Create a user profile" --format json --provider google-ai

# Example JSON Output:
{
  "content": "AI trends include increased automation...",
  "provider": "google-ai",
  "model": "gemini-2.5-flash",
  "usage": {
    "promptTokens": 15,
    "completionTokens": 127,
    "totalTokens": 142
  },
  "responseTime": 1234
}

// SDK JSON Input/Output - Full TypeScript support
import { createBestAIProvider } from "@juspay/neurolink";

const provider = createBestAIProvider();

// Structured input
const result = await provider.generate({
  input: { text: "Create a product specification" },
  schema: {
    type: "object",
    properties: {
      name: { type: "string" },
      price: { type: "number" },
      features: { type: "array", items: { type: "string" } },
    },
  },
});

// Access structured response
const productData = JSON.parse(result.content);
console.log(productData.name, productData.price, productData.features);

📖 Complete Setup Guide - All providers with detailed instructions

✨ Key Features

• 🔗 LiteLLM Integration - Access 100+ AI models from all major providers through unified interface
• 🔍 Smart Model Auto-Discovery - OpenAI Compatible provider automatically detects available models via /v1/models endpoint
• 🏭 Factory Pattern Architecture - Unified provider management with BaseProvider inheritance
• 🔧 Tools-First Design - All providers automatically include 6 direct tools (getCurrentTime, readFile, listDirectory, calculateMath, writeFile, searchFiles)
• 🔄 12 AI Providers - OpenAI, Bedrock, Vertex AI, Google AI Studio, Anthropic, Azure, LiteLLM, OpenAI Compatible, Hugging Face, Ollama, Mistral AI, SageMaker
• 💰 Cost Optimization - Automatic selection of cheapest models and LiteLLM routing
• ⚡ Automatic Fallback - Never fail when providers are down, intelligent provider switching
• 🖥️ CLI + SDK - Use from command line or integrate programmatically with TypeScript support
• 🛡️ Production Ready - Enterprise-grade error handling, performance optimization, extracted from production
• ✅ MCP Integration - Model Context Protocol with working built-in tools and 58+ discoverable external servers
• 🔍 Smart Model Resolution - Fuzzy matching, aliases, and capability-based search across all providers
• 🏠 Local AI Support - Run completely offline with Ollama or through LiteLLM proxy
• 🌍 Universal Model Access - Direct providers + 100,000+ models via Hugging Face + 100+ models via LiteLLM
• 📊 Analytics & Evaluation - Built-in usage tracking and AI-powered quality assessment

🛠️ MCP Integration Status ✅ BUILT-IN TOOLS WORKING

✅ Quick MCP Test (v1.7.1)

# Test built-in tools (works immediately)
npx @juspay/neurolink generate "What time is it?" --debug

# Disable tools for pure text generation
npx @juspay/neurolink generate "Write a poem" --disable-tools

# Discover available MCP servers
npx @juspay/neurolink mcp discover --format table

# Install popular MCP servers (NEW: Bitbucket support added!)
npx @juspay/neurolink mcp install filesystem
npx @juspay/neurolink mcp install github
npx @juspay/neurolink mcp install bitbucket  # 🆕 NEW

🔧 SDK Custom Tool Registration (NEW!)

Register your own tools programmatically with the SDK:

import { NeuroLink } from "@juspay/neurolink";
const neurolink = new NeuroLink();

// Register a simple tool
neurolink.registerTool("weatherLookup", {
  description: "Get current weather for a city",
  parameters: z.object({
    city: z.string().describe("City name"),
    units: z.enum(["celsius", "fahrenheit"]).optional(),
  }),
  execute: async ({ city, units = "celsius" }) => {
    // Your implementation here
    return {
      city,
      temperature: 22,
      units,
      condition: "sunny",
    };
  },
});

// Use it in generation
const result = await neurolink.generate({
  input: { text: "What's the weather in London?" },
  provider: "google-ai",
});

// Register multiple tools - Object format (existing)
neurolink.registerTools({
  stockPrice: {
    description: "Get stock price",
    execute: async () => ({ price: 150.25 }),
  },
  calculator: {
    description: "Calculate math",
    execute: async () => ({ result: 42 }),
  },
});

// Register multiple tools - Array format (Lighthouse compatible)
neurolink.registerTools([
  {
    name: "lighthouseTool1",
    tool: {
      description: "Lighthouse analytics tool",
      parameters: z.object({
        merchantId: z.string(),
        dateRange: z.string().optional(),
      }),
      execute: async ({ merchantId, dateRange }) => {
        // Lighthouse tool implementation with Zod schema
        return { data: "analytics result" };
      },
    },
  },
  {
    name: "lighthouseTool2",
    tool: {
      description: "Payment processing tool",
      execute: async () => ({ status: "processed" }),
    },
  },
]);

💰 Smart Model Selection

NeuroLink features intelligent model selection and cost optimization:

Cost Optimization Features

• 💰 Automatic Cost Optimization: Selects cheapest models for simple tasks
• 🔄 LiteLLM Model Routing: Access 100+ models with automatic load balancing
• 🔍 Capability-Based Selection: Find models with specific features (vision, function calling)
• ⚡ Intelligent Fallback: Seamless switching when providers fail

# Cost optimization - automatically use cheapest model
npx @juspay/neurolink generate "Hello" --optimize-cost

# LiteLLM specific model selection
npx @juspay/neurolink generate "Complex analysis" --provider litellm --model "anthropic/claude-3-5-sonnet"

# Auto-select best available provider
npx @juspay/neurolink generate "Write code" # Automatically chooses optimal provider

💻 Essential Examples

CLI Commands

# Text generation with automatic MCP tool detection (default)
npx @juspay/neurolink generate "What time is it?"

# Alternative short form
npx @juspay/neurolink gen "What time is it?"

# Disable tools for training-data-only responses
npx @juspay/neurolink generate "What time is it?" --disable-tools

# With custom timeout for complex prompts
npx @juspay/neurolink generate "Explain quantum computing in detail" --timeout 1m

# Real-time streaming with agent support (default)
npx @juspay/neurolink stream "What time is it?"

# Streaming without tools (traditional mode)
npx @juspay/neurolink stream "Tell me a story" --disable-tools

# Streaming with extended timeout
npx @juspay/neurolink stream "Write a long story" --timeout 5m

# Provider diagnostics
npx @juspay/neurolink status --verbose

# Batch processing
echo -e "Write a haiku\nExplain gravity" > prompts.txt
npx @juspay/neurolink batch prompts.txt --output results.json

# Batch with custom timeout per request
npx @juspay/neurolink batch prompts.txt --timeout 45s --output results.json

SDK Integration

// SvelteKit API route with timeout handling
export const POST: RequestHandler = async ({ request }) => {
  const { message } = await request.json();
  const provider = createBestAIProvider();

  try {
    // NEW: Primary streaming method (recommended)
    const result = await provider.stream({
      input: { text: message },
      timeout: "2m", // 2 minutes for streaming
    });

    // Process stream
    for await (const chunk of result.stream) {
      // Handle streaming content
      console.log(chunk.content);
    }

    // LEGACY: Backward compatibility (still works)
    const legacyResult = await provider.stream({ input: { text:
      prompt: message,
      timeout: "2m", // 2 minutes for streaming
    });
    return new Response(result.toReadableStream());
  } catch (error) {
    if (error.name === "TimeoutError") {
      return new Response("Request timed out", { status: 408 });
    }
    throw error;
  }
};

// Next.js API route with timeout
export async function POST(request: NextRequest) {
  const { prompt } = await request.json();
  const provider = createBestAIProvider();

  const result = await provider.generate({
    prompt,
    timeout: process.env.AI_TIMEOUT || "30s", // Configurable timeout
  });

  return NextResponse.json({ text: result.content });
}

🎬 See It In Action

No installation required! Experience NeuroLink through comprehensive visual documentation:

📱 Interactive Web Demo

cd neurolink-demo && node server.js
# Visit http://localhost:9876 for live demo

• Real AI Integration: All 9 providers functional with live generation
• Complete Use Cases: Business, creative, and developer scenarios
• Performance Metrics: Live provider analytics and response times
• Privacy Options: Test local AI with Ollama

🖥️ CLI Demonstrations

• CLI Help & Commands - Complete command reference
• Provider Status Check - Connectivity verification (now with authentication and model availability checks)
• Text Generation - Real AI content creation

🌐 Web Interface Videos

• Business Use Cases - Professional applications
• Developer Tools - Code generation and APIs
• Creative Tools - Content creation

📖 Complete Visual Documentation - All screenshots and videos

📚 Documentation

Getting Started

• 🔧 Provider Setup - Complete environment configuration
• 🖥️ CLI Guide - All commands and options
• 🏗️ SDK Integration - Next.js, SvelteKit, React
• ⚙️ Environment Variables - Full configuration guide

Advanced Features

• 🏭 Factory Pattern Migration - Guide to the new unified provider architecture
• 🔄 MCP Foundation - Model Context Protocol architecture
• ⚡ Dynamic Models - Self-updating model configurations and cost optimization
• 🧠 AI Analysis Tools - Usage optimization and benchmarking
• 🛠️ AI Workflow Tools - Development lifecycle assistance
• 🎬 Visual Demos - Screenshots and videos

Reference

• 📚 API Reference - Complete TypeScript API
• 🔗 Framework Integration - SvelteKit, Next.js, Express.js

🏗️ Supported Providers & Models

Tool Support Legend:

• ✅ Full: All tools working correctly
• ⚠️ Partial: Tools visible but may not execute properly
• ❌ Limited: Issues with model or configuration
• * Bedrock requires valid AWS credentials, Ollama requires specific models like gemma3n for tool support

✨ Auto-Selection: NeuroLink automatically chooses the best available provider based on speed, reliability, and configuration.

🔍 Smart Model Auto-Discovery (OpenAI Compatible)

The OpenAI Compatible provider includes intelligent model discovery that automatically detects available models from any endpoint:

# Setup - no model specified
export OPENAI_COMPATIBLE_BASE_URL="https://api.your-endpoint.ai/v1"
export OPENAI_COMPATIBLE_API_KEY="your-api-key"

# Auto-discovers and uses first available model
npx @juspay/neurolink generate "Hello!" --provider openai-compatible
# → 🔍 Auto-discovered model: claude-sonnet-4 from 3 available models

# Or specify explicitly to skip discovery
export OPENAI_COMPATIBLE_MODEL="gemini-2.5-pro"
npx @juspay/neurolink generate "Hello!" --provider openai-compatible

How it works:

• Queries /v1/models endpoint to discover available models
• Automatically selects the first available model when none specified
• Falls back gracefully if discovery fails
• Works with any OpenAI-compatible service (OpenRouter, vLLM, LiteLLM, etc.)

🎯 Production Features

Enterprise-Grade Reliability

• Automatic Failover: Seamless provider switching on failures
• Error Recovery: Comprehensive error handling and logging
• Performance Monitoring: Built-in analytics and metrics
• Type Safety: Full TypeScript support with IntelliSense

AI Platform Capabilities

• MCP Foundation: Universal AI development platform with 10+ specialized tools
• Analysis Tools: Usage optimization, performance benchmarking, parameter tuning
• Workflow Tools: Test generation, code refactoring, documentation, debugging
• Extensibility: Connect external tools and services via MCP protocol
• 🆕 Dynamic Server Management: Programmatically add MCP servers at runtime

🔧 Programmatic MCP Server Management [Coming Soon]

Note: External MCP server activation is in development. Currently available:

• ✅ 6 built-in tools working across all providers
• ✅ SDK custom tool registration
• 🔍 MCP server discovery (58+ servers found)
• 🚧 External server activation (one-line fix pending)

Manual MCP configuration (.mcp-config.json) support coming soon.

🤝 Contributing

We welcome contributions! Please see our Contributing Guidelines for details.

Development Setup

git clone https://github.com/juspay/neurolink
cd neurolink
pnpm install
npx husky install          # Setup git hooks for build rule enforcement
pnpm setup:complete        # One-command setup with all automation
pnpm test:adaptive         # Intelligent testing
pnpm build:complete       # Full build pipeline

Enterprise Developer Experience

NeuroLink features enterprise-grade build rule enforcement with comprehensive quality validation:

# Quality & Validation (required for all commits)
pnpm run validate:all      # Run all validation checks
pnpm run validate:security # Security scanning with gitleaks
pnpm run validate:env      # Environment consistency checks
pnpm run quality:metrics   # Generate quality score report

# Development Workflow
pnpm run check:all         # Pre-commit validation simulation
pnpm run format           # Auto-fix code formatting
pnpm run lint             # ESLint validation with zero-error tolerance

# Environment & Setup (2-minute initialization)
pnpm setup:complete        # Complete project setup
pnpm env:setup             # Safe .env configuration
pnpm env:backup            # Environment backup

# Testing (60-80% faster)
pnpm test:adaptive         # Intelligent test selection
pnpm test:providers        # AI provider validation

# Documentation & Content
pnpm docs:sync             # Cross-file documentation sync
pnpm content:generate      # Automated content creation

# Build & Deployment
pnpm build:complete        # 7-phase enterprise pipeline
pnpm dev:health            # System health monitoring

Build Rule Enforcement: All commits automatically validated with pre-commit hooks. See Contributing Guidelines for complete requirements.

📖 Complete Automation Guide - All 72+ commands and automation features

📄 License

MIT © Juspay Technologies

🔗 Related Projects

• Vercel AI SDK - Underlying provider implementations
• SvelteKit - Web framework used in this project
• Model Context Protocol - Tool integration standard

Built with ❤️ by Juspay Technologies