Overview

7 days30 days

Latest CI Pipeline Executions

Fuzzy

Succeeded
1133-featai-grok-catch-up-grok-imagine-image-video-and-voice-modes
95166346 fix(ai-grok): normalize JSON-null options and name endpoints in failures Re-applies the review residuals on top of the rebased branch: - Treat JSON-serialized null / empty-array values for duration and the reference fields as unset on every path: the generation branch no longer sends "duration": null or empty reference arrays, and null reference fields no longer trip the v1.0 reference gate; edit/extend destructure the reference fields out of the passthrough so no known key can leak into those bodies. - Name the actual endpoint in create-job failure messages so a failed edit/extension no longer reports as a 'video generation' failure. - Type-tie GROK_VIDEO_REFERENCE_MODELS / GROK_VIDEO_SOURCE_MODELS to the per-model options map so a typo or an unlisted future capable model is a compile error. - Sync the media-generation skill with the current contract (v1.0-only edit/extend, reference caps, 720p reference cap, 1080p on 1.5 only, edit/extend option rejections). - Tests: null-duration strip, null/empty reference normalization, endpoint-named edit failure (124 total).
by Tom Beckenham
Succeeded
1133-featai-grok-catch-up-grok-imagine-image-video-and-voice-modes
95166346 Merge 4ae0d3d66c663588a15d85a29e1e4a3a22aa65ec into fe42d8255d89f4b087caff5e1f52ea4f98acf77b
by Tom Beckenham
Succeeded
1133-featai-grok-catch-up-grok-imagine-image-video-and-voice-modes
111b3fc6 fix(ai-anthropic): add opus-5 models to the tool-capabilities map
by Alem Tuzlak
Succeeded
1133-featai-grok-catch-up-grok-imagine-image-video-and-voice-modes
111b3fc6 Merge 148d92b80ee51c268798abac06f4e3bee938c5cb into fe42d8255d89f4b087caff5e1f52ea4f98acf77b
by Tom Beckenham
Succeeded
1133-featai-grok-catch-up-grok-imagine-image-video-and-voice-modes
2d701f79 fix(ai-grok): restrict edit/extend to grok-imagine-video and reject mixed modes
by Alem Tuzlak
Failed
1133-featai-grok-catch-up-grok-imagine-image-video-and-voice-modes
2d701f79 Merge bd96ec58b02d01ff3e774ba2169a01aa52cf87d1 into fe42d8255d89f4b087caff5e1f52ea4f98acf77b
by Tom Beckenham
Failed
1133-featai-grok-catch-up-grok-imagine-image-video-and-voice-modes
d70999f7 fix(ai-grok): harden video mode routing and per-model options after review Address /review-pr findings on #1136: - Reject unknown modelOptions.mode values instead of silently misrouting a source-video body to /videos/generations (wrong endpoint, wrong billing). - Stop leaking raw duration/aspect_ratio/resolution from modelOptions into edit/extend bodies: edit rejects any duration (output inherits the source clip's length) and both modes reject size/aspect_ratio/resolution, matching the documented inherit-from-source semantics instead of sending fields the API ignores. Extend still snaps its added-tail duration. - Hoist mode-exclusivity checks into a dedicated edit/extend path so error messages never give generation-path advice for mode requests. - Throw on unknown image metadata.role strings instead of silently dropping the part (roles arrive untrusted from JSON callers). - Gate reference inputs to grok-imagine-video-1.5 at runtime, and split the provider options per model (GrokVideoBaseProviderOptions vs the 1.5 shape) so v1.0 no longer advertises reference fields at compile time. - Extract GROK_DEFAULT_REALTIME_MODEL so the token issuer and client adapter can't drift apart on future default bumps. - Comment/doc accuracy fixes: 'character' role alias documented, stale xAI URL and file headers refreshed, create-response JSDoc covers all three endpoints, deprecated 1.0 voice ids named in docs, example now requests the native 1080p it advertises. - 10 new tests: unknown mode/role rejection, edit duration/size rejection, extend snapping (modelOptions + generic), combined start-frame+reference body, explicit reference_images override, v1.0 reference gating, 1080p T2V body assertion, realtime token default model.
by Tom Beckenham
Failed
1133-featai-grok-catch-up-grok-imagine-image-video-and-voice-modes
d70999f7 Merge 40eb80032ffd5c89ae103f0c0146c7baeb7d8910 into bc8c5e8684da159b08e63aba7cfc51b01289d4eb
by Tom Beckenham
Failed
1133-featai-grok-catch-up-grok-imagine-image-video-and-voice-modes
66564582 feat(ai-grok): catch up Grok Imagine image, video, and voice modes - Image: add grok-imagine-image-2.0 (xAI's recommended model) with the 2.0-only quality provider option ('low' | 'medium'). - Video: drop the stale image-to-video-only guard on grok-imagine-video-1.5 (it now supports text-to-video with native 1080p); add reference-to-video (role:'reference' image parts -> reference_images, preset voices via modelOptions.reference_audios, max 3); add edit/extend modes via a source video prompt part + modelOptions.mode ('edit' -> /v1/videos/edits, 'extend' -> /v1/videos/extensions with duration = added tail). - Voice: add grok-voice-think-fast-2.0 (current recommended) and the grok-voice-latest alias; move realtime token/adapter defaults off the deprecated 1.0 ids. - Fix the stale GROK_CHAT_MODELS test expectation broken by the #1048 model-metadata sync. - Update docs, media-generation skill, and example apps accordingly. Closes #1133
by Tom Beckenham
Failed
1133-featai-grok-catch-up-grok-imagine-image-video-and-voice-modes
66564582 Merge 8220372b355bb2d5fe0098d207b3220ca1808324 into bc8c5e8684da159b08e63aba7cfc51b01289d4eb
by Tom Beckenham