PicoWatch

PicoWatch is an ultra-lightweight, self-hosted observability and logging dashboard designed specifically for tracking LLM (Large Language Model) API usage, costs, and performance.

Built with Go and SvelteKit, it compiles down to a single, dependency-free binary that consumes merely ~24 MB of RAM while serving a fully-featured modern web interface.

Features

• Single Binary Deployment: The entire Svelte frontend is statically compiled and embedded directly into the Go backend executable (//go:embed). No Node.js, Docker, or external web servers required.
• Incredibly Lightweight:
  • Memory: Runs on ~24 MB of RAM (RSS).
  • Storage: The compiled standalone binary is only ~16 MB.
• LLM Cost & Pricing Tracking: Syncs directly with OpenRouter to pull the latest AI model pricing, automatically calculating your token costs across different models.
• Secure by Default: Built-in authentication system with secure API key management and /setup wizard for the initial admin account.
• SQLite Database: Zero configuration database using SQLite (with WAL mode enabled for performance). Just run the file and your data is persisted locally in picowatch.db.
• Data Retention: Configurable auto-cleanup for old logs to ensure the database stays small and snappy over time.

Tech Stack

• Backend: Go (Golang) + standard net/http router
• Database: SQLite3
• Frontend: SvelteKit (Static Adapter) + TailwindCSS
• Communication: RESTful JSON API

Building from Source

To build PicoWatch from source, you need Go (1.21+) and Node.js.

The simplest way is to use the Makefile:

make build

This will install frontend dependencies, build the Svelte UI, and compile everything into a single Go binary.

Alternatively, you can run the steps manually:

1. Build the Svelte Frontend

   cd ui
   npm install
   npm run build
   cd ..
   

   This compiles the Svelte frontend into static HTML/JS/CSS inside the ui/build directory.

2. Compile the Go Binary

   go build -o picowatch .
   

   The Go compiler will embed the ui/build directory directly into the compiled executable.

Running the Application

Simply execute the compiled binary. By default, the server will start on port 8080.

./picowatch

Optionally, you can override the port or database path using environment variables:

PORT=3000 DB_PATH=/data/picowatch.db ./picowatch

First Launch (Setup)

On the very first launch, open http://localhost:8080 in your browser. You will be redirected to the secure /setup page to create your master admin account and generate your first API key.

Docker

The easiest way to run PicoWatch is with Docker Compose:

docker compose up -d

This will build the image and start PicoWatch on port 8080. The database is persisted in a named Docker volume (picowatch-data).

To use a custom port:

# docker-compose.yml
ports:
  - "3000:8080"

API Usage Example

Once you have your API Key, you can log your LLM requests easily from anywhere:

curl -X POST http://localhost:8080/api/logs \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-4-turbo",
    "input_tokens": 120,
    "output_tokens": 45,
    "duration_ms": 1250
  }'

License

MIT License