axflow logo

axflow

Free

The TypeScript framework for AI development ![GitHub Repo stars](https://img.shields.io/github/stars/axflow/axflow?style=social)

FreeFree tier
Type
Open Source

About axflow

Axflow is a collection of TypeScript modules for building robust natural language powered applications. It takes a code-first, modular approach, allowing developers to adopt modules incrementally. The framework includes @axflow/models (a zero-dependency SDK with React hooks and streaming utilities), axgen (for connecting data to large language models), and axeval (for evaluating LLM output quality). Future modules include extract (data processing and chunking for vector search), serve (a serving framework for LLMs with middleware for throttling, analytics, and logging), and finetune (model fine-tuning). The library is open source under the MIT license and aims to deconstruct complex LLM workflows into manageable, intuitive components.

Key Features

Zero-dependency modular SDK for building natural language applications
React hooks and streaming utilities for AI application development
axgen module for connecting data to large language models
axeval module for evaluating LLM output quality
Planned modules for data extraction (extract), LLM serving (serve), and fine-tuning (finetune)
Code-first, modular, and scalable framework for TypeScript developers

Pros & Cons

Pros
  • Modular design allows incremental adoption of components
  • Zero-dependency SDK reduces integration complexity
  • Includes React hooks and streaming utilities out of the box
  • Open source under MIT license with active development on GitHub
  • TypeScript-native for type safety and developer experience
Cons
  • Some modules (extract, serve, finetune) are still in development and not yet released
  • Documentation may be limited as the project is relatively new
  • Fewer community resources compared to more established frameworks

Best For

Building robust natural language powered applications with TypeScriptConnecting custom data to large language models for retrieval-augmented generationEvaluating and benchmarking the quality of LLM outputsDeveloping AI features and products using a modular, code-first approach