LOAM

LOAM is a system for off-grid, local-first communication.

It allows groups of people to communicate without internet access by connecting to a nearby device acting as a WiFi hotspot. Messages are shared locally and can be relayed between devices using low-bandwidth radios like LoRa, allowing communication to spread across a wider area.

The system is designed for use in situations where traditional networks are unavailable, unreliable, or unsafe, such as emergencies, large events, or areas with limited infrastructure.

LOAM prioritises:

• simplicity (no accounts, no setup)
• privacy (anonymous, ephemeral identities)
• resilience (works with very low bandwidth and intermittent connectivity)

Users connect via their phone's browser to a local web app (PWA), where they can send short messages, view announcements, and coordinate with others nearby. The interface is intentionally minimal and works across languages, including support for right-to-left scripts.

LOAM is transport-agnostic, meaning it can operate over different communication methods (WiFi, LoRa, or others), while maintaining the same user experience.

The goal is to provide a reliable, easy-to-use communication layer that works even when the internet does not.

Workspace

This repository is a small pnpm workspace.

• apps/client: Preact client application and local demo surface
• apps/server: backend server application and API surface
• packages/avatar: deterministic avatar generation from a shared SVG template
• packages/display-name: deterministic anonymous display-name generation from an id
• packages/qr: QR-related shared utilities
• packages/schema: shared data and schema definitions

Development

Install dependencies:

pnpm install

Run the workspace build:

pnpm build

Run the workspace test suite:

pnpm test

Optional local configuration

LOAM runs without a config file. Optional identity and LLM features can be enabled by creating .loam/config.json, or by setting LOAM_CONFIG_FILE to another JSON file path.

See config.example.json for a complete example. With Ollama running locally, the important settings are:

{
  "identity": {
    "allowUserDisplayNameEdit": true,
    "allowUserAvatarEdit": true,
    "allowUserAvatarUpload": true,
    "allowAdminUserEdit": true
  },
  "llm": {
    "ollama": {
      "enabled": true,
      "baseUrl": "http://localhost:11434",
      "model": "gemma4",
      "botId": "llm.ollama.gemma4",
      "botDisplayName": "Gemma"
    }
  }
}

When enabled, the Ollama bot appears as a direct-message contact. Sending it a DM creates a streaming assistant response in the same LOAM conversation. If the config is absent or llm.ollama.enabled is false, no LLM user or LLM routes are active.

When allowUserAvatarUpload is enabled, users can choose an image in the settings screen, crop it locally in the browser, and upload only the final 256 x 256 PNG/WebP avatar. Original image files are never sent to the server.