Python JavaScript

Home

# GraphBit - High Performance Agentic Framework

**Type-Safe AI Agent Workflows with Rust Performance**

Installation¶

npm install @infinitibit_gmbh/graphbit

Quick Start¶

import { init, LlmClient, LlmConfig } from '@infinitibit_gmbh/graphbit';

// Initialize GraphBit
init();

// Configure LLM provider
const config = LlmConfig.openai({
  apiKey: process.env.OPENAI_API_KEY,
  model: 'gpt-4o-mini'
});

// Create client and generate completion
const client = new LlmClient(config);
const response = await client.complete('What is AI?', 100);
console.log(response);

Features¶

🤖 Multi-LLM Support - OpenAI, Anthropic, Ollama
🔄 Workflow Automation - Build complex agentic workflows
🛠️ Tool Integration - Function calling and tool execution
📄 Document Processing - PDF, DOCX, text splitting
🎯 Type-Safe - Full TypeScript support
⚡ High Performance - Native Rust implementation via NAPI

Core APIs¶

LLM Client¶

const client = new LlmClient(config);
await client.complete(prompt, maxTokens);
await client.completeBatch(prompts, maxTokens, temperature, concurrency);

Workflows¶

const workflow = new WorkflowBuilder('My Workflow')
  .description('Description')
  .build();

const executor = new Executor(llmConfig);
const result = await executor.execute(workflow);

Embeddings¶

const embClient = new EmbeddingClient(embConfig);
const response = await embClient.embed(['text 1', 'text 2']);
const similarity = EmbeddingClient.similarity(emb1, emb2);

Document Processing¶

const loader = new DocumentLoader();
const doc = await loader.loadFile('./document.pdf', 'pdf');

const splitter = TextSplitter.recursive(500, 50);
const chunks = await splitter.split(doc.content);

Tool Calling (with Async Support)¶

const { ToolRegistry, registerAsync } = require('@infinitibit_gmbh/graphbit');

const registry = new ToolRegistry();

// Register async tools with proper timing tracking
registerAsync(registry, 'fetchData', 'Fetches external data', {}, async (args) => {
  const response = await fetch(args.url);
  return await response.json();
});

const result = await registry.execute('fetchData', { url: 'https://api.example.com' });
console.log(result.executionTimeMs); // Includes full async wait time

Documentation¶

Requirements¶

Node.js ≥ 16.0.0
Supported platforms: Windows, macOS, Linux (x64, ARM64)

License¶

Apache License, Version 2.0.

Links¶

🚀 Getting Started¶

Installation Guide (JS/TS) - Install GraphBit for Node.js
Dependency Installation - Install dependencies for different scenarios
Quick Start (JS/TS) - Build your first agent in 5 minutes
Basic Examples - Simple examples to get you started

📚 User Guide¶

Core Concepts - Understand workflows, agents, and nodes
Workflow Builder - Creating and connecting workflow nodes
Agent Configuration (JS/TS) - Agent configuration for JavaScript/TypeScript
Document Loader (JS/TS) - Document loading for JavaScript/TypeScript
Text Splitters (JS/TS) - Text splitting for JavaScript/TypeScript
LLM Providers (JS/TS) - LLM configuration for JavaScript/TypeScript
Embeddings (JS/TS) - Embeddings for JavaScript/TypeScript
Data Validation (JS/TS) - JSON validation for JavaScript/TypeScript
Dynamic Graph Generation - Auto-generating workflow structures
Performance Optimization - Tuning for speed and efficiency
Monitoring & Observability - Metrics collection and debugging
Reliability & Fault Tolerance - Circuit breakers, retries, and error handling

🔧 API Reference¶

Configuration Options - All configuration parameters
Node Types - Agent nodes

🔗 Connectors & Integrations¶

AWS Boto3 - Amazon Web Services integration
Azure - Microsoft Azure services integration
Google Cloud Platform - Google Cloud services integration
Vector Databases - Pinecone, Qdrant, ChromaDB, and more

🛠️ Development¶

Architecture Overview - System design and components
Contributing Guide - How to contribute to GraphBit
JavaScript Bindings Architecture - Node.js-Rust integration details
Debugging - Troubleshooting and debugging workflows

📋 Examples & Use Cases¶

Content Generation Pipeline - Multi-agent content creation
Data Processing Workflow - ETL pipelines with AI agents
LLM Integration - Working with different language models
Semantic Search - Building intelligent search systems
Comprehensive Pipeline - End-to-end workflow examples

What is GraphBit?¶

GraphBit is a declarative framework for building reliable AI agent workflows with strong type safety, comprehensive error handling, and predictable performance. It features:

🔒 Type Safety - Strong typing throughout the execution pipeline
🛡️ Reliability - Circuit breakers, retry policies, and error handling
🤖 Multi-LLM Support - OpenAI, Anthropic, Ollama
⚡ Performance - Rust core with Node.js bindings for optimal speed
📊 Observability - Built-in metrics and execution tracing
🔧 Resource Management - Concurrency controls and memory optimization

Architecture¶

GraphBit uses a three-tier architecture:

GraphBit JS Architecture

Community & Support¶

GitHub: github.com/InfinitiBit/graphbit
Issues: Report bugs and request features
Discussions: Ask Questions and Share ideas
Contributing: See Our contributing guide

Ready to build your first AI workflow? Start with GraphBit's Quick Start Tutorial!