Skip to content

AGENTS.md

Latest commit

 

History

History
48 lines (34 loc) · 2.96 KB

AGENTS.md

File metadata and controls

48 lines (34 loc) · 2.96 KB

RunAnywhere Web Example — AGENTS.md

Overview

This is the browser validation app for the Web SDK. It consumes the Swift-aligned public facade from @runanywhere/web, backend registration packages from @runanywhere/web-llamacpp and @runanywhere/web-onnx, and browser helpers from @runanywhere/web/browser.

The example may break when the SDK facade changes; update it to the latest API rather than preserving old compatibility imports.

Commands

Run from examples/web/RunAnywhereAI/.

npm run typecheck
npm run build
npm run dev -- --host 127.0.0.1

SDK Surfaces By View

Tab View File Current SDK Surface
Chat views/chat.ts RunAnywhere.generateStream, RunAnywhere.generateWithTools
Vision views/vision.ts VideoCapture, RunAnywhere.loadModel, RunAnywhere.visionLanguage.loadCurrentModel, RunAnywhere.processImage
Voice views/voice.ts Placeholder until STT/VAD/TTS ONNX/Sherpa artifacts are linked
Transcribe views/transcribe.ts AudioCapture, RunAnywhere.transcribe once ONNX/Sherpa artifacts are linked
Speak views/speak.ts RunAnywhere.synthesize, AudioPlayback once ONNX/Sherpa artifacts are linked
Documents views/documents.ts RunAnywhere.ragIngest, RunAnywhere.ragQuery, RAG diagnostics
Storage views/storage.ts RunAnywhere.storage, RunAnywhere.modelRegistry, RunAnywhere.loadModel
Solutions views/solutions.ts RunAnywhere.solutions
Settings views/settings.ts local UI settings

Browser Requirements

The Vite dev server sets COOP/COEP headers for SharedArrayBuffer. Runtime WASM assets are copied from the SDK workspace when present — there are now three independent WASM artifacts across three packages:

WASM artifact Owning package Loaded by Used by views
racommons.{js,wasm} @runanywhere/web (core) RunAnywhere.initialize() All views (commons facade state)
racommons-llamacpp.{js,wasm} (CPU) @runanywhere/web-llamacpp LlamaCPP.register() Chat, Vision, Documents (RAG embeddings)
racommons-llamacpp-webgpu.{js,wasm} (WebGPU) @runanywhere/web-llamacpp LlamaCPP.register({ acceleration: 'webgpu' }) — runtime capability check picks one Chat, Vision (when WebGPU+JSPI available)
racommons-onnx-sherpa.{js,wasm} @runanywhere/web-onnx ONNX.register() Voice, Transcribe, Speak

The legacy packages/onnx/wasm/sherpa/sherpa-onnx.wasm standalone bundle has been deleted. STT/TTS/VAD now run through the proto-byte adapters in @runanywhere/web core against the registered Sherpa vtable inside racommons-onnx-sherpa.wasm — there is no separate standalone speech provider path.

Validation Standard

A passing build or app launch is only smoke validation. End-to-end modality validation requires browser launch, model download, model load, real inference, and reviewed logs/screenshots per test_workflows/instructions/web/.