Spettro Subscription

Spettro offers a paid subscription plan that provides:

• Inference credits — use Spettro-managed models through the spettro provider without configuring third-party API keys.
• Multi-model access — several models available under a single key, tiered by the plan (see Pricing).
• Overflow tier — pro/max plans continue working on a free-tier model once the monthly credit budget is exhausted (with automatic retry).

Signing in

Inside the TUI:

/login

This starts a device-flow login:

1. Spettro contacts the backend and gets a browser URL.
2. A message in the chat shows the URL. Copy it.
3. Open the URL in your browser and sign in.
4. Once authenticated, Spettro receives an ep_ API key and saves it encrypted to ~/.spettro/keys.enc.

The flow polls the backend every 2 seconds until you complete the browser step or the session expires.

You can also sign in directly from the onboarding wizard by selecting "Sign in to your Spettro subscription" as your provider.

Signing out

/logout

This:

• Removes the saved ep_ API key from the encrypted store.
• Clears the Spettro Subscription models from the provider manager.
• Stops any in-flight inference through the Spettro proxy.

You can sign back in at any time with /login.

Active model list

Once signed in, Spettro fetches the models available on your plan from GET /v1/models. These appear in the model selector under the Spettro provider group, labelled with your plan tier.

The model list is refreshed in the background periodically so new models or plan upgrades are picked up without re-logging.

Plan and credit info

When signed in, the top bar of the TUI shows your plan name, status, and credit usage.

Spettro (Pro)

If the plan is in a special state (e.g. past due, cancelled), the status is shown alongside the plan name.

Overflow rate limiting

When a pro or max account exhausts its monthly credit budget, inference continues on a free-tier model — but at a reduced rate. The Spettro backend returns 429 Too Many Requests with a Retry-After header. The provider manager transparently waits these out and retries, so a short spell of throttling never shows up as an error in the chat. You may notice slower response times when approaching or exceeding your credit limit.

The waits back off exponentially (about 1s, 2s, 4s, ...) with jitter, each capped at the Retry-After the backend sent (20 seconds without one), so parallel sessions throttled together do not retry in lockstep. Early retries can come before the Retry-After, since that is the worst-case refill time. Once a request has waited about 3 minutes in all, the 429 is reported as an error, and the model fallback chain, if configured, is offered. It is not retried again: a workflow sub-agent that hits it fails instead of being re-run from scratch.

Configuration

The Spettro Subscription state is stored in:

Data	Location
API key (ep_...)	~/.spettro/keys.enc (encrypted)
Email / plan / status (cache)	~/.spettro/config.json (spettro_email, spettro_plan, spettro_plan_status fields)

The cached plan info is displayed immediately on startup before the network refresh completes. It is updated asynchronously in the background.

API

The Spettro backend (overridable via the SPETTRO_API_URL environment variable) exposes:

Endpoint	Purpose
POST /auth/initiate	Register a login session, returns a browser URL
GET /auth/poll/:session	Poll until the user signs in, returns an ep_ key
GET /v1/models	List models available on the user's plan
GET /v1/account	Plan, status, and credit usage
POST /v1/chat/completions	OpenAI-compatible inference (used by the provider manager)

The default base URL is https://api.spettro.app.

Pricing

See https://spettro.app/pricing for current plans, credit limits, and model availability.