Replace the entire @mana/local-llm engine with a transformers.js-based
implementation backed by Google's Gemma 4 E2B (released 2026-04-02).
The external API of LocalLLMEngine — load(), generate(), prompt(),
extractJson(), classify(), onStatusChange(), isSupported() — is
preserved 1:1, so the /llm-test page, the playground module, and the
Svelte 5 reactive bindings in svelte.svelte.ts need no changes
beyond updating the default model key.
Why the engine swap: MLC has not (and as of today still hasn't)
published Gemma 4 builds for WebLLM. The webml-community team and
HuggingFace's onnx-community already have Gemma 4 E2B running in
the browser via transformers.js + WebGPU, with a documented
Gemma4ForConditionalGeneration class shipped in @huggingface/transformers
v4.0.0. Going through the ONNX route gets us the latest Google model
six days after release instead of waiting on MLC compilation.
Trade-offs accepted (discussed before this commit):
- transformers.js is a more generic ONNX runtime, so per-token
throughput will be ~20-40% lower than WebLLM would deliver for the
same model size. For a 2B model on a modern WebGPU device that's
still well above interactive latency.
- The JS bundle gains ~2-3 MB (the ONNX runtime). Negligible compared
to the 500 MB model download.
- transformers.js v4 is brand new (released alongside Gemma 4) so the
Gemma4ForConditionalGeneration code path has very little battle
testing yet. The risk is partially offset by webml-community's
reference implementation.
What changed file by file:
- packages/local-llm/package.json: drop @mlc-ai/web-llm, add
@huggingface/transformers ^4.0.0; bump version 0.1.0 → 0.2.0; rewrite
description.
- packages/local-llm/src/types.ts: add `dtype` field to ModelConfig
('fp32' | 'fp16' | 'q8' | 'q4' | 'q4f16') so each model can request
the quantization that matches its uploaded ONNX shards.
- packages/local-llm/src/models.ts: replace the old Qwen 2.5 + Gemma 2
registry with a single `gemma-4-e2b` entry pointing at
onnx-community/gemma-4-E2B-it-ONNX with q4f16 quantization. Future
models can be added by appending entries — the /llm-test picker
reads MODELS dynamically and picks them up automatically.
- packages/local-llm/src/cache.ts: replace the WebLLM-specific
hasModelInCache helper with a generic Cache API probe that looks for
`https://huggingface.co/{model_id}/resolve/main/tokenizer.json` in
any open cache. tokenizer.json is small, downloaded first, and
always present, so its presence is a reliable proxy for "model has
been loaded before".
- packages/local-llm/src/engine.ts: full rewrite. Internally we now
hold a transformers.js model + processor pair (created via
AutoProcessor.from_pretrained + Gemma4ForConditionalGeneration.from_pretrained
with `device: 'webgpu'`), and translate our LoadingStatus union from
the library's `progress_callback` shape. generate() applies Gemma's
chat template via the processor, runs model.generate() with optional
TextStreamer for streaming, then slices the prompt tokens off the
output tensor to compute per-call usage. The convenience methods
(prompt, extractJson, classify) are unchanged because they only call
generate() under the hood.
- packages/local-llm/src/generate.ts and status.svelte.ts: deleted.
These were orphaned from a much earlier engine API (referenced
`getEngine()` / `subscribe()` / `LlmState` symbols that haven't
existed for a while) and were never re-exported from index.ts —
they only showed up because `tsc --noEmit` was crawling the src
tree. Their functionality lives in engine.ts + svelte.svelte.ts now.
- apps/mana/apps/web/package.json: swap the direct dep from
@mlc-ai/web-llm to @huggingface/transformers. This is the same
trick we used for the previous adapter-node externals warning —
having it as a direct dep makes adapter-node's Rollup pass treat
it as external automatically.
- apps/mana/apps/web/vite.config.ts: swap ssr.external entry from
@mlc-ai/web-llm to @huggingface/transformers. Add a comment
explaining the why so the next person doesn't wonder.
- apps/mana/apps/web/src/routes/(app)/llm-test/+page.svelte: change
the default selectedModel from 'qwen-2.5-1.5b' to 'gemma-4-e2b'.
All other model display strings come from the MODELS registry, so
this is the single hard-coded reference that needed updating.
- pnpm-lock.yaml: regenerated. Confirmed @mlc-ai/web-llm is gone (0
references) and @huggingface/transformers is in (4 references).
CSP: no header changes needed. We already opened connect-src for
huggingface.co + cdn-lfs.huggingface.co + raw.githubusercontent.com
when fixing the WebLLM blockers earlier today, and 'wasm-unsafe-eval'
is already in script-src — both transformers.js (ONNX runtime) and
WebLLM (MLC runtime) need that. If transformers.js spawns its
inference into a Web Worker via a blob URL we may need to add
`worker-src 'self' blob:` once we hit the first runtime test, but
the existing CSP should be enough for the synchronous path.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
|
||
|---|---|---|
| .. | ||
| apps | ||
| .gitignore | ||
| CLAUDE.md | ||
| README.md | ||
Mana Apps
A unified application ecosystem built on a shared authentication system, supporting multiple branded applications across web and mobile platforms.
Overview
Mana Apps is a monorepo containing web and mobile applications that provide organization management, team collaboration, and credit transfer capabilities. The system supports multiple branded applications (Memoro, Cards, Storyteller, Mana) through a flexible multi-tenant architecture.
Applications
- Web App (
apps/web) - SvelteKit-based web application - Mobile App (
apps/mobile) - React Native (Expo) app for iOS, Android, and web - Landing (
apps/landing) - Landing page (planned)
Features
- 🔐 Unified authentication with Supabase
- 🏢 Organization management with role-based access
- 👥 Team collaboration and member management
- 💰 Mana credit system with transfers and balance tracking
- 🎨 Multi-brand support with configurable themes
- 📱 Cross-platform (Web, iOS, Android)
- 🔄 Real-time updates across all platforms
- 🧪 Comprehensive testing with Vitest and Playwright
Quick Start
Prerequisites
- Node.js 20+
- pnpm (for web app)
- npm (for mobile app)
- Supabase account with project configured
- Expo CLI (for mobile development)
Setup
-
Clone the repository
git clone <repository-url> cd mana-core-apps -
Web App Setup
cd apps/web pnpm install cp .env.example .env # Edit .env with your Supabase credentials pnpm dev -
Mobile App Setup
cd apps/mobile npm install cp .env.example .env # Edit .env with your Supabase credentials npm start
Project Structure
mana-core-apps/
├── apps/
│ ├── web/ # SvelteKit web application
│ │ ├── src/
│ │ │ ├── routes/ # File-based routing
│ │ │ │ ├── (auth)/ # Public auth pages
│ │ │ │ └── (app)/ # Protected pages
│ │ │ ├── lib/
│ │ │ │ ├── components/
│ │ │ │ ├── config/ # Multi-app configuration
│ │ │ │ ├── server/ # Server-only utilities
│ │ │ │ └── types/
│ │ │ └── hooks.server.ts # Auth middleware
│ │ └── package.json
│ │
│ ├── mobile/ # React Native (Expo) app
│ │ ├── app/ # File-based routing (Expo Router)
│ │ │ ├── (drawer)/ # Drawer navigation
│ │ │ ├── auth/ # Auth screens
│ │ │ └── _layout.tsx # Root layout with auth
│ │ ├── components/ # React components
│ │ ├── utils/ # Utilities (Supabase, storage)
│ │ └── package.json
│ │
│ └── landing/ # Landing page (planned)
│
├── CLAUDE.md # Developer documentation
└── README.md # This file
Technology Stack
Web App (apps/web)
| Category | Technology |
|---|---|
| Framework | SvelteKit 2 with Svelte 5 (Runes) |
| Language | TypeScript |
| Styling | TailwindCSS 3 with PostCSS |
| Database | Supabase (PostgreSQL) |
| Auth | Supabase Auth with SSR |
| Testing | Vitest (unit) + Playwright (E2E) |
| Build Tool | Vite |
Mobile App (apps/mobile)
| Category | Technology |
|---|---|
| Framework | Expo 52 with React Native 0.76 |
| Language | TypeScript |
| Routing | Expo Router 4 (file-based) |
| Styling | NativeWind (TailwindCSS for RN) |
| Navigation | React Navigation (drawer, tabs) |
| Database | Supabase |
| Build | EAS Build |
| Platforms | iOS, Android, Web |
Development
Web App Commands
cd apps/web
# Development
pnpm dev # Start dev server (http://localhost:5173)
pnpm build # Build for production
pnpm preview # Preview production build
# Code Quality
pnpm check # Type-check with svelte-check
pnpm check:watch # Type-check in watch mode
pnpm lint # Check formatting and lint
pnpm format # Format code with Prettier
# Testing
pnpm test # Run unit tests (Vitest)
pnpm test:ui # Run tests with UI
pnpm test:e2e # Run E2E tests (Playwright)
Mobile App Commands
cd apps/mobile
# Development
npm start # Start Expo dev server
npm run ios # Run on iOS simulator
npm run android # Run on Android emulator
npm run web # Run web version (http://localhost:19006)
# Building
npm run build:dev # Build dev client
npm run build:preview # Build for internal testing
npm run build:prod # Build for production
# Code Quality
npm run lint # Lint and check formatting
npm run format # Fix linting and format code
# Setup
npm run prebuild # Generate native projects
Environment Configuration
Both apps require Supabase configuration. Create .env files based on .env.example:
Web App (apps/web/.env)
PUBLIC_SUPABASE_URL=your_supabase_project_url
PUBLIC_SUPABASE_ANON_KEY=your_supabase_anon_key
MIDDLEWARE_URL=https://mana-middleware-111768794939.europe-west3.run.app
PUBLIC_APP_NAME=Mana Web
NODE_ENV=development
Mobile App (apps/mobile/.env)
EXPO_PUBLIC_SUPABASE_URL=your_supabase_project_url
EXPO_PUBLIC_SUPABASE_ANON_KEY=your_supabase_anon_key
Architecture
Multi-Tenant System
The system supports multiple branded applications sharing the same authentication backend:
- Memoro - Voice recordings and memory management
- Cards - AI-powered flashcard learning
- Storyteller - Creative writing with AI assistance
- Mana - Central account and organization management
App configurations are centralized in apps/web/src/lib/config/apps.ts, defining branding, features, and routing for each application.
Authentication Flow
Web (SvelteKit):
- Server-side authentication using
@supabase/ssr - Middleware in
hooks.server.tshandles session validation - Protected routes in
(app)group require authentication - JWT validation via
safeGetSession()before allowing access
Mobile (Expo):
- Client-side authentication using
@supabase/supabase-js - Custom memory storage for session persistence
AuthProviderinapp/_layout.tsxmanages auth state- Automatic navigation based on authentication status
Database Schema
Key tables:
users- User profiles (linked viaauth_idto Supabase Auth)organizations- Organization entitiesuser_roles- User-organization relationships with rolesteams- Team entities within organizationsteam_members- User-team membershipscredit_transactions- Mana credit transfer history
See CLAUDE.md for detailed architecture documentation.
Testing
Web App
cd apps/web
# Unit tests
pnpm test # Run all tests
pnpm test:ui # Open Vitest UI
# E2E tests
pnpm test:e2e # Run Playwright tests
pnpm test:e2e --ui # Run with Playwright UI
Mobile App
Mobile testing is primarily done through Expo Go or development builds:
cd apps/mobile
npm start # Start dev server
# Then press 'i' for iOS or 'a' for Android
Deployment
Web App
Vercel (Recommended):
cd apps/web
vercel
Netlify:
cd apps/web
netlify deploy
Mobile App
iOS and Android (via EAS):
cd apps/mobile
# Preview build (internal testing)
npm run build:preview
# Production build
npm run build:prod
Configure EAS in eas.json with your build profiles.
Contributing
- Create a feature branch from
main - Make your changes
- Run linting and tests
- Submit a pull request
Code Style
- Use TypeScript for type safety
- Follow ESLint and Prettier configurations
- Write tests for new features
- Use conventional commit messages
Documentation
- CLAUDE.md - Comprehensive developer guide for Claude Code
- apps/web/README.md - Web-specific documentation
- Individual component documentation in source files
Support
For questions or issues, please contact the development team or open an issue in the repository.
License
Private - All rights reserved