feat(provider/quiverai): support SVG editing and animation
Author: gr2mCreated Sep 16, 2026Updated Sep 17, 2026
Labelstype:batchtask-identify-issue-type-donetask-feature-analysis-split-featurefactory-automatictask-identify-harness-labels-done
QuiverAI's Arrow 2 launch adds native SVG editing and animation. @ai-sdk/quiverai currently exposes generation and vectorization through generateImage; add these operations using the existing image-model interface.
Proposed scope:
- Extend
providerOptions.quiverai.operationwitheditandanimate, routing to/v1/svgs/editsand/v1/svgs/animations. - Use one source SVG in
prompt.images; expose edit reference images separately through a typed provider option. Verify URL, binary, and base64 inputs against the API. - Require an edit instruction; allow animation without an instruction. Map editing settings into the endpoint's nested
settingsobject and exposemaxReviewStepsand reasoning effort. - Return SVG bytes, including animated SVGs. Preserve animation timing fields (
loop_period_ms,opening_animation_ms) in per-image provider metadata. - Build operation-specific request bodies and validate source counts, required fields, and batching. Preserve existing generation/vectorization behavior.
- Confirm model/operation availability through the live catalog and API-key permissions.
Acceptance criteria: request/response and invalid-input tests in Node and Edge, live verification for both operations, examples under examples/ai-functions/src/generate-image/quiverai/, provider documentation, and a patch changeset. No new video-model interface is needed for SVG output.
References:
Source: vercel/ai