Quiver Clone → Brand Studio

A QuiverAI-compatible REST API at /v1 plus a Brand Studio skills foundation — tokens, church vertical, agent runtime — ready to merge into Brand Studio.

Layer	Location
SVG API + playground	src/app/v1/, src/app/app/
Skills (422)	skills/manifest.json
Media Studio	/app/studio · docs/MEDIA_ENGINE.md
Brand Studio handoff	skills/BRAND_STUDIO_INTEGRATION.md

npm run skills:update-registry   # rebuild manifest after skill edits
npm run skills:validate-profiles

Features

• Text to SVG — POST /v1/svgs/generations
• Image to SVG — POST /v1/svgs/vectorizations
• Model discovery — GET /v1/models, GET /v1/models/{model}
• OpenAPI spec — GET /v1/openapi.json
• Playground UI, gallery, export (SVG/PNG)
• Media Studio — unified image/video/SVG generation (/v1/media/*, Higgsfield-style)

Quick start

npm install
npm run dev

Base URL: http://localhost:3000/v1

Default API key (local dev): qv_dev_local_key

Authentication

Authorization: Bearer qv_dev_local_key
Content-Type: application/json

Set custom keys in .env.local:

QUIVER_CLONE_API_KEYS=qv_live_key_1,qv_live_key_2
OPENAI_API_KEY=sk-...          # optional, for AI text-to-SVG
ANTHROPIC_API_KEY=sk-ant-...    # optional

API examples

List models

curl http://localhost:3000/v1/models \
  -H "Authorization: Bearer qv_dev_local_key"

Text to SVG

curl -X POST http://localhost:3000/v1/svgs/generations \
  -H "Authorization: Bearer qv_dev_local_key" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "arrow-1.1",
    "prompt": "a minimalist logo of a mountain",
    "stream": false
  }'

Image to SVG

curl -X POST http://localhost:3000/v1/svgs/vectorizations \
  -H "Authorization: Bearer qv_dev_local_key" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "arrow-1.1",
    "image": { "url": "https://example.com/logo.png" },
    "auto_crop": true
  }'

Streaming

Set "stream": true on generation or vectorization. The response is text/event-stream with event: lines (generating, reasoning, draft, content) and terminates with data: [DONE].

SDK compatibility

Drop-in replacement for QuiverAI SDK calls by changing the base URL:

import { QuiverAI } from "@quiverai/sdk";

const client = new QuiverAI({
  bearerAuth: "qv_dev_local_key",
  baseUrl: "http://localhost:3000", // instead of https://api.quiver.ai
});

Or use cURL with the same request bodies from QuiverAI docs.

Response shape

Non-streaming responses match QuiverAI:

{
  "id": "resp_...",
  "created": 1704067200,
  "credits": 20,
  "data": [{ "mime_type": "image/svg+xml", "svg": "<svg ...>" }],
  "usage": { "input_tokens": 0, "output_tokens": 0, "total_tokens": 0 }
}

Errors use the same envelope:

{
  "status": 401,
  "code": "invalid_api_key",
  "message": "Invalid API key",
  "request_id": "..."
}

Rate limits & credits

• Rate limit: 20 requests / 60 seconds per API key (headers: X-RateLimit-*)
• Credits: 10,000 starting balance per org (in-memory; resets on server restart)

Model	Generate	Vectorize
arrow-1	30	30
arrow-1.1	20	15
arrow-1.1-max	25	20

Legacy routes

Simplified playground routes still work without auth:

• GET /api/models
• POST /api/generate
• POST /api/vectorize

Note

Independent recreation inspired by QuiverAI. Uses local templates + ImageTracer for vectorization, with optional OpenAI/Anthropic for AI generation — not Quiver's proprietary Arrow models.