# callstackincubator/ai

On-device LLM execution in React Native with Vercel AI SDK compatibility

Repository: https://github.com/callstackincubator/ai
Canonical: https://ross.abutalabs.com/products/callstackincubator-ai
Homepage: https://react-native-ai.dev
Language: TypeScript
License: MIT
License Family: permissive
Topics: ai, llm, on-device-ai, react-native
Last push: 2026-07-07T08:48:30+00:00

## Health v2 (maintenance only)
Score: 76/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 91, release rhythm 68, longevity 55
- inputs: {"age_days": 775, "days_push": 57, "days_rel": 217, "gap_med": 12, "n_releases_24m": 12}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1387, forks 58 (observed 2026-08-28T04:04:35.105891+00:00)

## What it is
A collection of on-device AI primitives for React Native offering Vercel AI SDK-compatible APIs for running LLMs, embeddings, transcription, and speech synthesis directly on mobile devices. It supports Apple Foundation Models, Google Gemini Nano via ADK, GGUF models via llama.rn, and MLC LLM runtimes on iOS and Android.

## Use cases
- run llm inference offline in a react native app
- add privacy-preserving on-device chat to a mobile app
- use apple foundation models from react native
- run gguf models locally on android and ios
- generate embeddings on-device for semantic search
- add speech-to-text and text-to-speech to a mobile app
- replace cloud llm api calls with local inference to cut costs

## When to choose
- you need offline-capable, low-latency AI features in a React Native app
- user data privacy requires that inference never leaves the device
- you want Vercel AI SDK APIs like streamText and generateText backed by local models
- you want to avoid server and API costs for AI features

## When to avoid
- you need frontier-scale model quality that only large cloud models provide
- your app is not React Native or targets web/desktop only
- you need broad model choice without managing GGUF or MLC model downloads
- you require features like large-context reasoning beyond on-device model limits

## Facets
- artifact type: library
- maturity: active
- function: llm-inference, sdk, speech-recognition, tts, rag, agent-framework
- domain: large-language-models, mobile-development, artificial-intelligence, cross-platform
- platform: -
- tags: on-device-ai, react-native, vercel-ai-sdk, apple-foundation-models, gemini-nano, gguf, mlc-llm, privacy-preserving, edge-ai, devtools-profiler, ios, android, mobile, typescript

## Member repositories
- callstackincubator/ai (main) score 76

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:04:35.105891+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T04:39:49.331473+00:00, confidence not recorded.
  - readme: https://github.com/callstackincubator/ai (fetched 2026-08-28T04:04:35.105891+00:00, sha fa4656f27df8)
  - homepage: https://react-native-ai.dev (fetched 2026-08-29T11:55:36.786300+00:00, sha ee2847066fb6)
  - site_page: https://www.react-native-ai.dev/docs (fetched 2026-08-29T11:55:36.795404+00:00, sha 0776aee95d78)
- Data as of 2026-08-30T08:39:29.467469+00:00.