Skip to content

README.md

Latest commit

 

History

History
46 lines (33 loc) · 1.76 KB

README.md

File metadata and controls

46 lines (33 loc) · 1.76 KB

WASM Chat

Browser-based chat UI powered by Adam compiled to WebAssembly. The full agent loop runs in the browser — LLM API calls go through JavaScript fetch() bridged to Adam's HTTP layer via Emscripten Asyncify. No server needed.

Build & Run

make         # builds adam.js + adam.wasm (requires Emscripten)
make serve   # starts local server on http://localhost:8000

Then open http://localhost:8000 in your browser.

Usage

  1. Select a provider (Anthropic, OpenAI, or Gemini)
  2. Enter your API key
  3. Optionally set a model name
  4. Click Connect and start chatting

Your API key stays in the browser — it's only sent to the LLM provider's API, never to any other server.

How It Works

Browser                          LLM Provider
───────                          ────────────
User types message
  → adam_run() in WASM
    → adam_net_post_json()
      → EM_ASYNC_JS → fetch()  ──→  POST /v1/messages
      ← response JSON          ←──  200 OK + JSON
    → adam_json_parse_response()
  ← result.final_response
Display in chat UI

The key piece is the Emscripten net implementation (adam.c): adam_net_post_json calls an EM_ASYNC_JS function that does await fetch(). Emscripten's Asyncify allows the synchronous C code to suspend while JavaScript completes the async HTTP request, then resume with the response.

Streaming is not supported in the WASM build — the streaming path returns an error and the agent loop falls back to non-streaming HTTP automatically.

Requirements

  • Emscripten (emcc in PATH)
  • A modern browser (Chrome, Firefox, Safari, Edge)
  • An API key for any supported provider