# Godance

> Drop-in OpenAI-compatible REST API that routes to 16+ AI providers across text, image,
> video, and audio. One credit ledger, one billing line, one SDK. Pay with card, ACH, SEPA,
> or USDC. Pass-through provider pricing plus a flat 5% platform fee — no tiers, no seats,
> no subscriptions.

## Positioning

- vs. OpenRouter: Godance covers image and video too, not just text. Plus crypto top-ups.
- vs. Eden AI: simpler flat 5% fee (Eden charges 5.5%, tier-dependent).
- vs. Fireworks / Together: pure aggregator — no GPUs, no model weights, no lock-in.
- vs. B.AI / WorldClaw: crypto is a payment method here, not the brand identity. Targets
  developer teams, not crypto-native funds.

## API surface (OpenAI-compatible)

- POST /v1/chat/completions — streaming + non-streaming
- POST /v1/images/generations — sync; HQ variants (e.g. gpt-image-1-hd) auto-promote to async
- POST /v1/videos/generations — async only, signed webhook callback
- GET /v1/jobs/{id} — poll status of any async generation
- POST /v1/audio/speech — text-to-speech
- POST /v1/audio/transcriptions — speech-to-text
- POST /v1/embeddings — text embeddings
- GET /v1/models — catalog enumeration
- CRUD /v1/keys — API key management (plaintext returned exactly once)
- POST /v1/billing/topup — Stripe Checkout for card / ACH / SEPA
- POST /v1/billing/coinbase/topup — Coinbase Commerce for USDC / USDT
- GET /v1/webhooks/secret — signing secret (Stripe-style HMAC-SHA256)
- POST /v1/webhooks/secret/rotate — rotate with 24h dual-signature grace

## Models

### Text
- OpenAI: gpt-4o-mini, gpt-4o, gpt-4.1
- Anthropic: claude-3-5-sonnet-20241022, claude-3-5-haiku-20241022
- Google Vertex: gemini-1.5-pro, gemini-1.5-flash
- Roadmap: DeepSeek V4, Kimi K2.6, Qwen 3.6, GLM 4.7, MiniMax, Grok

### Image
- OpenAI: gpt-image-1 (sync), gpt-image-1-hd (auto-async)
- Black Forest Labs: flux-1.1-pro, flux-2

### Video (async-only by design — durations 1–5 minutes; HTTP sync would time out)
- OpenAI: sora-2 ($0.15/sec)
- Volcengine Ark: seedance-1.0 ($0.08/sec, Alicloud-allowed)
- Roadmap: Veo 3.1, Kling, Runway Gen-4

## Pricing

Provider list price plus a flat 5% platform fee — same percentage across text, image,
video. No tiers, no seats, no minimums. Top up the ledger; spend across every model.

- Per-model rates: https://godance.ai/models
- Pricing page: https://godance.ai/pricing

## Deployment

Two-region: DigitalOcean primary (NYC3 / FRA1), Alibaba Cloud secondary (HK / SG) for
disaster recovery and low-latency to Chinese model providers. DNS-level failover via
Cloudflare health checks on /healthz.

Per-region provider allowlist enforced in code. The Alicloud region routes only DeepSeek,
Moonshot (Kimi), Alibaba (Qwen), Zhipu (GLM), MiniMax, and Volcengine Ark (Doubao /
Seedream / Seedance). Western frontier providers (Anthropic, OpenAI, Google Vertex, BFL
Flux, ElevenLabs, Cartesia, Runway) never route through Alicloud — compliance with
upstream terms of service for Chinese-hosted infrastructure.

## Fallback chains

Configure per-model fallbacks (e.g. claude-3-5-sonnet → claude-3-5-haiku → gpt-4o).
On primary 5xx / timeout / 429, Godance transparently retries the next entry. Response
carries fallback_used:true. Each fallback re-checks the regional allowlist.

## Webhook signing

Header: X-Godance-Signature: t=<unix>,v1=<hex>[,v1=<hex>...]

Multiple v1=... entries appear during the 24-hour grace window after a secret rotation.
Verifiers walk all v1 entries and accept the request if ANY matches the stored secret
(Stripe-compatible pattern). Per-user secret retrievable via GET /v1/webhooks/secret;
rotate via POST /v1/webhooks/secret/rotate.

## Status

Real-time per-provider availability and latency:
https://godance.ai/status (launches with public beta)