Skip to content
cannonade-aiPublic

About

Local-first desktop app for building LLM test suites and running them against many local or cloud models at once

Topics

Resources

Stars

7 stars

Watchers

0 watching

Forks

Latest commit

 

History

334 Commits

Folders and files

NameName
Last commit message
Last commit date
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Cannonade

cannonade

cannonade.app - Docs

Cannonade is a cross-platform local-first desktop app for building test suites, running them against multiple models at once, and comparing the results without writing an eval harness or shipping your prompts to a third-party service.

Cannonade test run

Features

  • Local models: view, load, unload, and delete the local models of your configured LLM providers.
  • External models: view cloud model details and pricing from your configured LLM providers.
  • Playground: chat with any configured model, tweak sampling parameters, and inspect token and timing stats.
  • Prompt library: keep system prompts in one place and reuse them across the playground and test suites.
  • Test suites: define test cases and configure evaluators once, then run them against any number of local or cloud models.
  • Test runs: every run is saved locally with results, scores, and stats for each case.
  • Model downloads: download models from the model registry of a configured provider or from huggingface.
  • Local first: prompts, test suites, and test runs are all stored locally in JSON files.

Supported providers

Provider Notes
LM Studio Local or remote LM Studio server
Ollama Local or remote Ollama server
llama.cpp Local or remote llama-server instance
Unsloth Local or remote Unsloth Studio server
Custom Any OpenAI-compatible local API endpoint
OpenRouter Access hundreds of models via OpenRouter
Vercel Access hundreds of models via AI Gateway
More More supported providers are coming soon

Download

Grab the build for your platform from the latest release:

Platform Build
Windows .exe installer
macOS .dmg (Apple silicon)
Linux .AppImage and .deb

Note

The macOS build is not code-signed or notarized, so Gatekeeper reports it as damaged on first launch. Copy Cannonade.app into /Applications, then remove the quarantine flag before opening it:

xattr -dr com.apple.quarantine /Applications/Cannonade.app

Full instructions, including what to do if macOS still blocks the app, are in the macOS setup docs.

Development

Built with Electron, Vue 3, TypeScript, Vite, Pinia, and Sass.

npm ci
npm run dev

Checks

npm run validate # typecheck + build + lint:fix + tests
npm run build
npm test
npm run lint:fix

Packaging

npm run build:win
npm run build:mac
npm run build:linux

Contributing

Bug reports and feature requests: GitHub Issues.

About

Local-first desktop app for building LLM test suites and running them against many local or cloud models at once

Topics

Resources

Stars

7 stars

Watchers

0 watching

Forks

Releases

Packages

Contributors

Languages