Chonkie.Embeddings 0.1.0-preview.87

Chonkie.Net - The Lightweight RAG Ingestion Library

Chonkie.Net is an experimental .NET/C# port of Python Chonkie, providing fast, efficient, and robust text chunking for Retrieval-Augmented Generation (RAG) systems. This is an independent port and is not officially affiliated with the original Chonkie project.

Key Features

• Fast & Efficient - 10-100x faster than Python implementations
• 11 Specialized Chunkers - Choose the right chunker for your data type
• 7 Embedding Providers - OpenAI, Azure, Gemini, Cohere, VoyageAI, Jina, and ONNX local models
• 9 Vector Database Integrations - Pinecone, Qdrant, Chroma, Weaviate, MongoDB, Pgvector, Elasticsearch, Milvus, Turbopuffer
• 5 LLM Providers - OpenAI, Azure, Groq, Cerebras, Gemini
• ONNX Support - Local embeddings with SentenceTransformers
• Complete RAG Pipeline - End-to-end document processing for RAG
• No Dependencies Bloat - Minimal, modular architecture
• Type-Safe - Full C# 14 nullable reference types support
• 900+ Tests - Comprehensive unit and integration test suite

Quick Start

Installation

dotnet add package Chonkie.Net

Basic Chunking (30 seconds)

using Chonkie.Chunkers;
using Chonkie.Tokenizers;

// Create a chunker
var chunker = new RecursiveChunker(
    tokenizer: new WordTokenizer(),
    chunkSize: 512
);

// Chunk your text
var text = "Your document here...";
var chunks = chunker.Chunk(text);

// Use the chunks
foreach (var chunk in chunks)
{
    Console.WriteLine($"Text: {chunk.Text}");
    Console.WriteLine($"Tokens: {chunk.TokenCount}");
}

With Embeddings & Vector Database

using Chonkie.Embeddings;
using Chonkie.Handshakes;

// Create embeddings
var embeddings = new OpenAIEmbeddings(
    apiKey: Environment.GetEnvironmentVariable("OPENAI_API_KEY")!
);

// Create vector database connection
var vectorDb = new PineconeHandshake(
    apiKey: "your-pinecone-key",
    indexName: "my-index",
    embeddingModel: embeddings
);

// Store chunks with embeddings (vectorDb embeds internally)
await vectorDb.WriteAsync(chunks);

Documentation

• Quick Start Guide - Get started in 5 minutes
• RAG System Tutorial - Build a complete RAG system
• Chunker Selection Guide - Choose the right chunker
• Vector Database Integration - Connect to any vector DB
• Python Migration Guide - Coming from Python Chonkie?

Chunkers (11 Types)

Embeddings (7 Providers)

• OpenAI
• Azure OpenAI
• Google Gemini
• Cohere
• VoyageAI
• Jina
• Local ONNX (SentenceTransformers)

LLM Providers (5 Types)

• OpenAI
• Azure OpenAI
• Groq (fast inference)
• Cerebras (ultra-fast)
• Google Gemini

Vector Databases (9 Integrations)

• Pinecone - Fully managed serverless
• Qdrant - Open-source vector search
• Chroma - Lightweight local embedding DB
• Weaviate - Open-source, flexible
• MongoDB - MongoDB Atlas Vector Search
• PostgreSQL - pgvector extension
• Elasticsearch - Search-optimized
• Milvus - High-performance distributed
• Turbopuffer - Real-time, edge-optimized

Common Use Cases

1. Document Ingestion for RAG

// Chunk documents, embed, and store in vector DB
var chunks = chunker.Chunk(document);
await vectorDb.WriteAsync(chunks);

2. Code Analysis

var codeChunker = new CodeChunker(
    tokenizer: new WordTokenizer(),
    chunkSize: 1024
);
var chunks = codeChunker.Chunk(sourceCode);

3. Semantic Search

var semanticChunker = new SemanticChunker(
    tokenizer: new WordTokenizer(),
    embeddingModel: embeddings,
    threshold: 0.5f
);
var chunks = semanticChunker.Chunk(text);
// Chunks grouped by semantic meaning

4. RAG Pipeline

var pipeline = new Pipeline()
    .ProcessWith("text")
    .ChunkWith("recursive", new { chunk_size = 1024 })
    .RunAsync(texts: documentText);

Why Chonkie.Net?

✅ Type Safety - Full C# 14 support
✅ Almost Production Ready - 900+ tests, zero warnings
✅ Extensively Documented - Tutorials and guides
✅ Complete Features - Feature parity with Python Chonkie, all major RAG components included

Minimum Requirements

• .NET 10.0 or higher
• C# 14 features enabled
• Windows, Linux, or macOS

Contributing

Contributions are welcome! Please visit GitHub Repository.

License

Licensed under Apache License 2.0. See LICENSE for details.

Learn More

• Official Repo: https://github.com/gianni-rg/Chonkie.Net
• Python Chonkie: https://github.com/chonkie-inc/chonkie
• Documentation: Check the /docs folder in the repository