Neore Chat

An AI chat application: multi-model conversations, tools, document and image handling, and a browser extension. Built with TanStack Start (React), a Lunora backend on Cloudflare Workers, and a set of Hono and Rust Workers for media and document processing.

License: FSL-1.1-ALv2 (Functional Source License). This is source-available, not open source under the OSI definition.

Daniel Bannert's work is supported by the community on GitHub Sponsors.

Prerequisites

• Node.js 22.22 (see .nvmrc)

• pnpm 12 (pinned via packageManager in package.json)

• Rust toolchain, needed to start the backend the first time, because the document-parser Worker is built from Rust:

  rustup target add wasm32-unknown-unknown
  cargo install worker-build --version 0.8.7 --locked

  Skip this if services/document-parser/build/ already exists.

• At least one AI provider key (for example an OpenRouter key) to get replies.

Repository layout

apps/
├── web/                # Main web app (TanStack Start, React)
├── browser-extension/  # Browser extension (CRXJS + React)
└── native/             # Tauri v2 desktop and mobile shell around the web app
backend/                # Lunora backend: Durable Object shards, D1, R2, Better Auth
packages/
├── ai/                 # Model registry, prompts, design presets, tool definitions
├── chat-ui/            # Chat primitives (composer, thread)
├── service-sdk/        # Typed clients for the Workers services
└── ui/                 # Base UI component library
services/               # Workers: llm-gateway, document-parser, browser-renderer,
                        #   embeddings, nsfw-checker
docs/                   # User-facing documentation
plans/                  # Feature and design plans

Getting started

pnpm install
pnpm dev:setup     # generates local keys and writes backend/.dev.vars and apps/web/.env
pnpm dev:chat      # starts the backend, the web app, and the LLM gateway

Then open http://localhost:5173.

Service	URL
Web app	http://localhost:5173
Backend	http://localhost:8788
LLM gateway	http://localhost:8787

To run one part on its own, use pnpm dev:app, pnpm dev:backend, or pnpm dev:llm-gateway. The backend runs as a daemon that outlives its terminal, so stop it with cd backend && ./node_modules/.bin/lunora dev stop.

Configuration

Templates live next to the code that reads them:

• backend/.dev.vars.example: backend secrets and provider keys
• apps/web/.env.example: web app build-time variables (VITE_*)
• services/llm-gateway/.env.example and packages/ai/.env.example: gateway and model settings

pnpm dev:setup fills in the required local values. Anything you add by hand stays in the git-ignored .dev.vars and .env files. Never commit them.

For local development the backend needs only the core variables (auth secret, encryption key, site URL). Email, OAuth, storage, and tool API keys have development stubs and are optional.

Scripts

Command	What it does
pnpm dev:setup	First-time setup: keys and env files
pnpm dev:chat	Backend, web app, and LLM gateway together
pnpm dev:app	Web app only
pnpm dev:backend	Backend only
pnpm build:packages	Build all shared packages
pnpm build:chat	Build the web app
pnpm lint	Prettier, ESLint, and dedupe checks
pnpm lint:types	TypeScript checks (run separately; lint does not include types)
pnpm lint:eslint:fix	Fix ESLint issues
pnpm lint:prettier:fix	Fix formatting
pnpm lint:secrets	Scan for committed secrets
pnpm env:generate-encryption	Generate an encryption key
pnpm clean	Remove build artifacts and node_modules

Tests run through the task runner: pnpm vis run test for everything, or pnpm test inside a single project directory.

Deployment

Deployment is handled by Alchemy (alchemy.run.ts) and runs in CI. Local deploys are for maintainers only:

DEPLOY_ENV=preview pnpm run deploy

Use pnpm run deploy, not a bare pnpm deploy: since pnpm 11 a script named deploy shadows pnpm's built-in command.

Contributing

Read CONTRIBUTING.md and CODE_OF_CONDUCT.md before opening a pull request. Report security issues privately as described in SECURITY.md, not in public issues.

Tech stack

• Web: TanStack Start, React 19, Tailwind CSS v4, TanStack Router and Query
• Backend: Lunora on Cloudflare Workers (Durable Object shards, D1, R2, live queries)
• Auth: Better Auth, served by the backend
• AI: Multiple providers routed through the LLM gateway
• Services: Hono and Rust Workers (document parsing, browser rendering, moderation, embeddings)
• Infrastructure: Cloudflare Workers, D1, R2, Vectorize, Queues; deployed with Alchemy
• Monorepo: pnpm workspaces with the vis task runner

License

This project is licensed under the Functional Source License, Version 1.1, ALv2 Future License (FSL-1.1-ALv2). See LICENSE.md. Individual packages may carry their own license; check the package.json in each directory.