Repository navigation
feat: fast mode on lucy-vton-3.6; lucy-vton-latest resolves to 3.6 - #34
Merged
Merged
Conversation
This was referenced Oct 11, 2026
Merged
tomershlasky
force-pushed
the
vton-3-6-fast-latest
branch
from
October 11, 2026 12:24
14124ef to
8bdcdec
Compare
Mirrors the JS SDK. The API now serves lucy-vton-3.6 on the fast tier, billed at 2x like 3.5, and lucy-vton-latest moves from lucy-vton-3.5 to lucy-vton-3.6 server-side. - RealtimeModels.LUCY_VTON_3_6 advertises supportedSpeeds = setOf(Speed.FAST). - Fast set is now lucy-2.5, lucy-latest, lucy-vton-3.5, lucy-vton-3.6, lucy-vton-latest; Speed KDoc, README and the capability pin test updated. - lucy-vton-latest geometry is unchanged (realtime 1280x720@30, video 1280x720@20), so only comments/tests reflect the new resolution target.
tomershlasky
force-pushed
the
vton-3-6-fast-latest
branch
from
October 11, 2026 12:24
8bdcdec to
81d1ce7
Compare
AdirAmsalem
approved these changes
Oct 12, 2026
tomershlasky
added a commit
to DecartAI/sdk
that referenced
this pull request
Oct 12, 2026
…lves to 3.6 (#222) ## Description Two updates for `lucy-vton-3.6`, now that the API supports them: 1. **Fast mode on `lucy-vton-3.6`.** The realtime registry entry gains `supportedSpeeds: ["fast"]`. The API already serves 3.6 on the fast tier, billed at 2x the standard realtime rate like 3.5. Without this change, `speed: "fast"` on 3.6 logs a "model does not list it" warning. This applies to realtime only; batch has no speed tier. 2. **`lucy-vton-latest` → `lucy-vton-3.6`.** The alias resolves server-side, so the SDK only updates comments, docs and test names. 3.5 and 3.6 have identical geometry: realtime 1280x720 at `{ideal: 30, max: 30}`, video 1280x720 at 20 fps. The alias entries keep their values, and `lucy-vton-latest` keeps `supportedSpeeds: ["fast"]`. Following the AGENTS.md runbook, `supportedSpeeds` is added only now that the API actually serves 3.6 on the fast tier. ## Changes | File | Change | | --- | --- | | `src/shared/model.ts` | `supportedSpeeds: ["fast"]` on realtime `lucy-vton-3.6`; both `lucy-vton-latest` comments now point at 3.6 | | `src/realtime/client.ts`, `README.md`, `index.html` | fast-mode model lists now include `lucy-vton-3.6` | | `AGENTS.md` | removes "opt-in by name; no fast tier" from 3.6; `lucy-vton-latest` → `lucy-vton-3.6` | | `tests/unit.test.ts` | adds 3.6 to `fastRealtimeModels` (the pinned exact set); 3.6 asserts `["fast"]`; latest test titles say 3.6 | | `tests/realtime.unit.test.ts` | adds 3.6 to the "no speed warning" loop | ## Verification | Command | Result | | --- | --- | | `pnpm --filter @decartai/sdk typecheck` | pass | | `pnpm --filter @decartai/sdk test` | 386 passed (15 files) | | `pnpm --filter @decartai/sdk build` | pass | | `pnpm --filter @decartai/sdk test:package` | pass | | `pnpm --filter @decartai/sdk format:check` | 86 files, clean | The live e2e was not run locally because no `DECART_API_KEY` was available. ## Compatibility Additive. `models.realtime("lucy-vton-3.6").supportedSpeeds` changes from `undefined` to `["fast"]`. No ids, geometry or types change. ## Server-side - Fast mode for `lucy-vton-3.6` is already live on the API. - The `lucy-vton-latest` → 3.6 alias change is rolling out on the API separately. Only docs and comments here depend on it, so merge order doesn't matter. --- ### Companion PRs (lucy-vton-3.6: fast mode + `lucy-vton-latest` → 3.6) - JS SDK: #222 - Python: DecartAI/decart-python#73 - iOS: DecartAI/decart-ios#49 - Android: DecartAI/decart-android#34 - C++: DecartAI/decart-cpp#15 <!-- conductor-workspace-link --> --- [Open workspace in Conductor](https://app.conductor.build/decart/w/261)
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Mirrors the JS SDK (DecartAI/sdk) for two API-side changes:
lucy-vton-3.6(realtime only). The API now serves 3.6 on the fast tier, billed at 2x like 3.5.RealtimeModels.LUCY_VTON_3_6now advertisessupportedSpeeds = setOf(Speed.FAST), the same as 3.5.lucy-vton-latestnow resolves server-side tolucy-vton-3.6(it used to resolve to 3.5). Geometry is the same (realtime 1280x720@30, video 1280x720@20), so only comments and tests change for this part. 3.6 is no longer something you have to pick by name to get.Changes
RealtimeModels.kt:LUCY_VTON_3_6getssetOf(Speed.FAST). The latest-aliases comment now noteslucy-vton-latest -> lucy-vton-3.6.VideoModels.kt: same note on the latest-aliases comment. Video models have no speed tiers, so nothing else changes here.RealTimeClient.kt: theSpeedKDoc lists the new fast set.RealtimeModelsTest.kt: the fast-capability pin test now includes 3.6, so the set is lucy-2.5, lucy-latest, lucy-vton-3.5, lucy-vton-3.6 and lucy-vton-latest. The test is renamed to match, and thelucy-vton-latestgeometry comment now points at 3.6.RealTimeClientUrlTest.kt:LUCY_VTON_3_6+Speed.FASTis now covered by the "does not warn" case.README.md: the fast-mode prose lists the new fast set.No public API shape changes and no version bump.
Verification
./gradlew :sdk:testDebugUnitTestlocally (JDK 17 Corretto, Android SDK platform 35): BUILD SUCCESSFUL, 194 tests, 0 failures, 0 skipped. That includes the updated pin test and the URL/warning tests.Server-side
lucy-vton-3.6is already live on the API.lucy-vton-latest→ 3.6 alias change is rolling out on the API separately. This PR only changes docs/comments for the alias (geometry and fast capability are identical for 3.5 and 3.6), so merge order doesn't matter.Note
Low Risk
Registry metadata, docs, and tests only; billing behavior for fast mode follows existing patterns for other VTON models.
Overview
Aligns the Android SDK with backend changes for lucy-vton-3.6 and the
lucy-vton-latestalias.RealtimeModels.LUCY_VTON_3_6now advertisesSpeed.FASTinsupportedSpeeds(same as 3.5), soConnectOptions.speed = Speed.FASTis treated as supported and no longer triggers the unsupported-speed warning for that model. README andSpeedKDoc list lucy-vton-3.6 / lucy-vton-latest among fast-tier models (2× billing, US only).Comments in
RealtimeModelsandVideoModelsnow state thatlucy-vton-latestresolves server-side to lucy-vton-3.6 (was 3.5); output geometry is unchanged. Unit tests pin the fast-capability set and cover LUCY_VTON_3_6 + FAST in the no-warning path.No new public API surface or version bump.
Reviewed by Cursor Bugbot for commit 81d1ce7. Bugbot is set up for automated code reviews on this repo. Configure here.
Companion PRs (lucy-vton-3.6: fast mode +
lucy-vton-latest→ 3.6)