ax
The pretty much "official" DSPy framework for Typescript
What is ax?
The pretty much "official" DSPy framework for Typescript
Overview of ax
What it is
Ax is a TypeScript-first programming framework for building with large language models (LLMs). It provides a unified programming model that compiles into verified libraries for TypeScript, Python, Java, C++, Go, and Rust. The framework is published as the npm package @ax-llm/ax.
Key features
- Signatures for typed structured generation using a string DSL, a fluent
f()builder, or any Standard Schema v1 validator (Zod, Valibot, ArkType). - Named deployment profiles that select wire behavior (native API, router, hosted inference) and resolve the model ID internally; the same signature works across providers like OpenAI, Anthropic, Google Gemini, Meta, Together, Fireworks, DeepSeek, and Grok.
- Agents with runtime execution, context budgets, checkpoints, action-log replay, discovery, memory, skills, and delegation.
- Flows as typed program graphs with branches, loops, feedback, cache behavior, parallel execution, and
.returns(...)projection. - Optimizers including GEPA, few-shot bootstrapping, portable optimizer artifacts, and evaluation/apply flows.
- Provider-native speed with streaming as the default; the hot path renders the signature, calls the provider, parses the result, and returns a typed value.
Use cases
- Structured extraction from text (e.g., sentiment, priority, ticket numbers).
- Constrained and nested object generation using a string grammar or fluent builder.
- Multi-language AI application development where the same program model runs across TypeScript, Python, Java, C++, Go, and Rust.
- Building agents with memory, delegation, and checkpointing.
- Creating typed program graphs (flows) with branching, looping, and parallel execution.
Getting started
typescript import { ai, ax } from '@ax-llm/ax'; const llm = ai({ name: 'openai', apiKey: process.env.OPENAI_APIKEY }); const classify = ax( 'review:string -> sentiment:class 'positive, negative, neutral'', ); const { sentiment } = await classify.forward(llm, { review: 'This product is amazing!', }); // sentiment: 'positive' — typed as the literal union For runnable examples in all supported languages, see src/examples/README.md in the repository.
License and pricing
The project is licensed under Apache-2.0 and is open source. Pricing is not stated in the provided metadata or README.
Based on the project's public README and repository metadata.