Back to directory

needle-rs

A Rust and WebAssembly runtime that runs Cactus Compute's Needle tool-calling models in a browser tab, Cloudflare Workers, Node.js, Python, C FFI or embedded no_std targets, with no server or API key.

needle-rs

needle-rs is a pure-Rust and WebAssembly runtime for Needle, a family of small transformer models that map a query and a tool list to a JSON function call. The same runtime and binary support all three Needle generations at once (121M, 45M and 26M parameters), reading the generation from the model container itself rather than from a file name, and none of it needs a server round trip or an API key.

The WebAssembly runtime is under 600KB, around 200KB over the wire after brotli compression, with one runtime dependency, and a full agent ships in 14MB with Needle 2 or 36MB with Needle 3. A Needle 2 session needs about 23MB of working memory, while a Needle 3 session adds a further 8.8MB of key-value cache at 512 tokens, or 2.3MB with an int8 cache.

More in Inference

See all 770
  • OpenDecision

    deepanwadhwa.github.io

    An open-source Python package that answers typed questions about application state or documents using a local natural-language-inference model, returning structured choice, score, yes/no or evidence-relation values instead of free text.

  • Lifeboat

    iterate.ai

    A local inference server that puts an OpenAI- and Anthropic-compatible API in front of open-weight language models running on hardware you control, from a laptop to a GPU cluster.

  • Claude S40

    github.com

    An unofficial Claude chat client for 2007-era Nokia Series 40 phones, built as a Java ME app plus a small Go relay server, with web search, reading mode and saved chats.