everyframe.studio

# EveryFrame miner

Run an EveryFrame miner on Bittensor **SN117** with `everycli` and an attested
cloud worker. The worker receives jobs from the EveryFrame coordinator, calls
your configured generation providers, and returns signed completion receipts.
It does not run generation models locally, so you do not need a GPU.

Commands are direct: `everycli doctor`, `everycli status`,
`everycli balances`, and so on. No command prefix is needed. Existing scripts using
the older namespace remain compatible. `everycli update` upgrades the CLI;
`everycli worker-update` updates the existing worker using its latest compatible
signed release.

New to mining? Run `everycli ui` for a guided setup page and status dashboard
in your browser (see [Set up in your browser](#set-up-in-your-browser-everycli-ui)).

This source tree includes the worker, the CLI with built-in updates, native
provider contracts, flexible Studio video and media contracts, and sixteen
provider integrations. See the [direct provider guide](DIRECT-PROVIDERS.md) and
[additional provider models](PROVIDER-EXPANSION.md).

Before joining, review and accept the [Miner Terms of Service](docs/miner-terms.md)
in the setup guide or with `everycli terms`. Miners must obtain required provider
permissions for resale and serving third-party requests, and contact providers
directly for approvals and discounts. A working API key is not approval.
See [acceptance and unattended setup](docs/miner-terms-acceptance.md).

The v1.0.0 integration also includes optional [automatic provider-price offers](docs/automatic-offers.md): explicit model selection, price previews, discount ceilings and estimated exposure limits. It stays off until enabled.

The direct Google integration adds Gemini Omni Flash, four Nano Banana image
models, Lyria 3.5, and both Gemini 3.8 TTS models. See the [Google provider
guide](GOOGLE.md) for supported settings, references and compatible-worker requirements.
Available jobs depend on the coordinator's
enabled models, current pricing, your credentials, and your active offers—not
every model in the registry is necessarily available for mining.

## What's new

v1.1.1 improves [automatic offers](docs/automatic-offers.md): check prices first, choose models that pass your limits, and see per-model reasons when a price check fails. It also adds longer bounded coordinator requests and VEED input-second pricing support. Update the CLI and reopen the UI; automatic-offer users should also install the matching signed worker.

v1.1.0 adds direct [Alibaba Cloud, BytePlus and Volcengine](ASIAN-PROVIDERS.md) providers for Wan, Qwen Image, Seedance and Seedream, including the reviewed download, update-recovery and provider-error fixes.

In v1.1.0, Fal readiness uses a cached authenticated key check.
Workers do not poll Fal's public model catalog. A failed metadata or key-check
request does not disable all Fal models; confirmed invalid keys and reviewed
model retirements still block work. See [Fal readiness](docs/fal-readiness.md).

v1.1.0 adds [WaveSpeedAI](WAVESPEED.md), including 85
reviewed endpoints, plus automatic offers for 10 Fal video contracts. New provider
capabilities require the coordinated server and signed worker update.

See the [latest release](https://github.com/everyframe-studio/everyframe-miner/releases/latest)
for what changed, how to upgrade and whether a worker update is needed. Notes for
earlier releases are on the [releases page](https://github.com/everyframe-studio/everyframe-miner/releases).

To upgrade, run `everycli update`, then reopen `everycli ui`. When a release needs
a new worker, choose **Update worker** in the UI (or run `everycli worker-update`).
Updating the CLI alone never changes a running worker. Saved keys, offers and
discounts are kept; newly added models need their own offers.

Provider guides: [direct providers](DIRECT-PROVIDERS.md), [OpenRouter](OPENROUTER.md),
[Google](GOOGLE.md), [Pika](PIKA.md), [WaveSpeedAI](WAVESPEED.md), [Kling 4.0 through Fal](KLING4.md) and
[additional provider models](PROVIDER-EXPANSION.md).

### Model IDs

The **EveryFrame contract ID** is used with `everycli offer --model`. The
**Provider API ID** is the underlying provider's ID; for example,
`direct/openrouter/veo-3.1` maps to `google/veo-3.1`. Never paste a
`studio/` or `direct/` contract ID into the provider's API.

The Models page lists **Candidates / retired** separately from supported models.
Retired models cannot receive new offers, but existing offers can still be
withdrawn and past jobs stay readable. A listed model is not proof of adequate
funds, capacity or successful generation.

### Job history and earnings

```sh
everycli jobs                            # newest 20 of your own attempts
everycli jobs --status failed --limit 20
everycli jobs --model direct/vivix/vivix-w1
everycli jobs --summary                  # all-time counts by model
everycli earnings
```

The same views are under **Job history** and **Earnings** in `everycli ui`.
Earnings separates pending review, approved work and reward-eligible value. The
eligible value is part of approved work, not an extra payment, and estimated
provider costs are not an invoice. These reports never approve work, change
offers or repeat generations.

### Recover a deleted worker or a rejected authorization

If you deleted a worker (including directly in Phala), or the coordinator rejects
your saved miner authorization (HTTP 401), run `everycli reconnect` in the same
profile, or use **Worker → Reconnect profile**:

```bash
everycli reconnect --wallet your-wallet --hotkey default
everycli doctor
```

Reconnect asks before retiring an old binding, needs a proof from the same local
hotkey and keeps your API keys, offers, discounts and history. It never creates,
deletes, starts or updates a VM. Afterwards, use **Deploy worker** if the old VM is
gone. A 401 is separate from a Phala API-token error; check your computer's clock first.

## Worker failure diagnostics and safe failo