Echelon InboxAI models
Triage and summaries work out of the box on Cloudflare Workers AI. For drafting replies, you can use Workers AI or bring any OpenAI-compatible provider.
What uses which model
| Job | Default | Can you change it? |
|---|---|---|
| Triage | Cloudflare Clef Flash (@cf/cloudflare/clef-flash) | Server-wide only, by whoever runs the server |
| Summaries | Gemma on Workers AI | Yes, in Settings → AI for summaries |
| Drafting replies | The server's default, if it has one | Yes, in Settings → AI for writing replies |
A small model is plenty for summaries. Drafting is worth a bigger one, since it writes in your name.
Supported providers
- Cloudflare Workers AI (no API key needed)
- OpenAI
- OpenRouter
- Ollama Cloud
- Abacus AI (RouteLLM)
- OpenCode Go. Pick a model that supports chat completions, such as GLM, Kimi or DeepSeek.
- Any other service with an OpenAI-compatible
/chat/completionsendpoint. Choose Other and enter its base URL.
Set up a provider
- Open Settings (⌘ ,) and find the AI section you want to change.
- Choose a provider. The app links to the page where you can get an API key.
- Paste your key, then pick a model from the list.
- Click Test connection to check it works, then Save.
API keys are encrypted and stored on the server. The app never shows a saved key again; leave the field empty to keep it.