Skip to content

Repository files navigation

AnyLM

Local-first LLM workspace for Ollama.
Projects with RAG, streaming multi-agent chat, and one OpenAI-compatible endpoint every local app can share — inference stays on your machine.

Download · Website · Releases · Launch kit

AnyLM

Why AnyLM

Most local tooling either gives you a runtime (Ollama) or a chat window. AnyLM adds the missing workspace layer:

  • Projects with instructions and attached reference docs
  • RAG — docs are chunked, embedded, and retrieved into chat
  • Multi-agent orchestration with a visible agent trail
  • One local API (http://127.0.0.1:3227/v1) so Cursor, Continue, and scripts share a resident runtime
  • Desktop installers for macOS, Windows, and Linux

Works with the models you already pulled. No cloud inference required.

Download

Platform Get it
macOS (Apple Silicon / Intel) anylm.app/download
Windows anylm.app/download
Linux (AppImage) anylm.app/download

Or grab assets from GitHub Releases. Prerequisites: Ollama running locally, with a chat model and (for RAG) nomic-embed-text.

ollama pull llama3.2
ollama pull nomic-embed-text

Repository layout

Path Role
app/ Electron desktop app (TypeScript, Bun)
web/ Marketing + download site (Next.js) — reads GitHub Releases
firebase/ Auth, Firestore rules, hosted sign-in page
docs/ Release notes, specs, launch kit
press-kit/ Logos and social images for Product Hunt / press
auth-backend/ Retired NestJS service (reference only — not shipped)

There is no inference server of yours to operate. Governance logic runs inside the app against Firestore under the signed-in user’s token; firebase/firestore.rules is the authorization layer.

Develop from source

Prerequisites: Bun 1.1+, Ollama as above.

./scripts/setup.sh
./scripts/dev.sh

Firebase emulators:

./scripts/dev.sh --emulator

Manual app setup is in app/README.md; Firebase deploy notes in firebase/README.md.

The OpenAI-compatible endpoint runs inside the app on http://127.0.0.1:3227/v1.

Configuration and secrets

File Contents Ships in the app?
app/.env Firebase project id, web API key, public OAuth client ids, local service hosts Yes — treat as public
.env.example (repo root) Code-signing + GitHub release credentials No — CI only

Provider client secrets for Google/GitHub live in the Firebase project. Outlook uses a public client with PKCE. User tokens are stored via the OS keystore (app/src/main/token-store.ts).

app/scripts/build-env.js refuses to bake keys matching SECRET / PASSWORD / PRIVATE / *_TOKEN / CSC_* / APPLE_* into the bundle.

What enforcement does and does not guarantee

Firestore rules stop a user raising their own token limit, editing org policy without the role for it, reading a colleague’s prompts, or rewriting usage history. They cannot compel a client to report usage, so limits are cooperative rather than adversarial-proof.

That trade is deliberate: anyone who can patch the app can also run Ollama directly. REST-shaped paths in app/src/main/api/ keep a seam if a server must return later.

Launch & community

Shipping on Product Hunt, HN, and community channels? Start here:

License

MIT © 2026 Yash Aryan and contributors.

About

Local-first LLM workspace for Ollama.

Topics

Resources

Contributing

Security policy

Stars

0 stars

Watchers

0 watching

Forks

Releases

Packages

Contributors

Languages