# sauravpanda/BrowserAI

Run local LLMs like llama, deepseek-distill, kokoro and more inside your browser

Repository: https://github.com/sauravpanda/BrowserAI
Canonical: https://ross.abutalabs.com/products/browserai
Homepage: https://browserai.dev
Language: TypeScript
License: MIT
License Family: permissive
Topics: ai, llm-inference, webgpu, agents, llm, tts, localllm, llama, local
Last push: 2026-07-21T02:47:31+00:00

## Health v2 (maintenance only)
Score: 78/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 93, release rhythm 80, longevity 43
- inputs: {"age_days": 603, "days_push": 43, "days_rel": 138, "gap_med": 4, "n_releases_24m": 20}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1449, forks 138 (observed 2026-08-28T04:04:45.670600+00:00)

## What it is
BrowserAI is a TypeScript library for running LLMs, speech recognition, text-to-speech, and audio separation models directly in the browser using WebGPU and WebAssembly, with no server required. It provides a unified SDK over multiple engines (MLC, Transformers, Flare, Demucs) with Web Worker support, structured output, and built-in storage.

## Use cases
- run llama or deepseek models locally in the browser without a server
- build a privacy-focused chatbot with no API keys or cloud costs
- add speech recognition and text-to-speech voice chat to a web app
- generate structured JSON output from an LLM client-side
- build AI agents that run entirely in the user's browser
- create an offline-capable AI web app
- separate vocals and stems from audio in the browser

## When to choose
- you need client-side LLM inference with zero server or API costs
- privacy requires that user data never leaves the device
- you want a simple SDK abstracting multiple browser AI engines
- you're building chat, voice, or agent experiences in a web app

## When to avoid
- you need large models that exceed browser memory or WebGPU limits
- you require server-grade throughput or heavy batch inference
- your users are on browsers without WebGPU/WASM support
- you need fine-tuning or training rather than inference

## Facets
- artifact type: library
- maturity: active
- function: llm-inference, speech-recognition, tts, agent-framework, rag, sdk, chatbot
- domain: large-language-models, artificial-intelligence, web-development, frontend, speech-processing
- platform: browser, wasm
- tags: webgpu, local-llm, in-browser-ai, privacy-first, web-workers, structured-output, mlc, transformers-js, audio-separation, indexeddb, ai-agents, web-server, nodejs

## Member repositories
- sauravpanda/BrowserAI (main) score 78

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:04:45.670600+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 2026-08-30T04:36:01.446267+00:00, confidence not recorded.
  - readme: https://github.com/sauravpanda/BrowserAI (fetched 2026-08-28T04:04:45.670600+00:00, sha 43181eaad6ee)
  - homepage: https://browserai.dev (fetched 2026-08-29T11:45:54.046188+00:00, sha da07ad3a1659)
  - site_page: https://docs.browserai.dev/ (fetched 2026-08-29T11:45:54.055487+00:00, sha a077ad69809d)
  - site_page: https://browserai.dev/about (fetched 2026-08-29T11:45:54.058621+00:00, sha 39bd75639948)
- Data as of 2026-08-30T08:39:29.467469+00:00.