Who contributed to Vercel on August 12, 2026?

12 developers shipped this update, including n1ckoates, joshualipman123, github-actions, gr2m, Dominik Ferber, github-actions[bot], unstubbable, and gnoff, and 4 more.

What were the notable Vercel updates?

feat(ui): add typed custom bodies to Completion APIs, feat (provider/xai): add reference audio for reference-to-video, and feat (provider/xai): support new grok imagine video 1.5 caps.

@vercel

Next.js + the Vercel platform - frontend infrastructure for the web

github ↗

Pick a date

Topics: JavaScript Full archive →

The Wire · Showcase

AI SDK ADDS TYPED COMPLETION BODIES, GROK VIDEO 1.5 SUPPORT LANDS

By RepoJournal · Filed 06:56 UTC on August 12, 2026 · About Vercel

12 people shipped this

github-actions[bot] @github-actions[bot] 4 cited

anthonyshew @anthonyshew 4 cited

joshualipman123 @joshualipman123 2 cited

n1ckoates @n1ckoates 1 cited

Dominik Ferber @dferber90 1 cited

unstubbable @unstubbable 1 cited

gnoff @gnoff 1 cited

kelvin @kelvinampofo 1 cited

lukesandberg @lukesandberg 1 cited

github-actions @github-actions

gr2m @gr2m

mischnic @mischnic

Vercel/ai shipped typed generics for custom completion bodies across all frameworks, while xAI Grok Imagine Video 1.5 integration arrives with 1080p native output and reference voice support.

The completion APIs now accept an optional body generic [1] that lets teams enforce type safety on custom payloads across React, Vue, Svelte, and Angular, defaulting to `object` for backward compatibility. In parallel, xAI provider support matured with two critical additions: reference-to-video gained preset voice IDs [2] for character voicing on up to 3 voices, and the new Grok Imagine Video 1.5 caps feature [3] adds native 1080p output for text-to-video and image-to-video while fixing a routing bug where empty reference arrays triggered malformed requests. The flags SDK bumped all providers [ref:7,8,9,10] after resolving four Dependabot security alerts [4] including infinite-loop DoS in nanoid and host-confusion XSS in fast-uri. Next.js fixed literal export support in `'use cache'` files [9], resolved a Turbopack runtime initialization race [10], added module type generics to import.meta.glob [11], and reduced cache footprint by retaining only one stale version for 3 days [12]. Turborepo validated all examples with turbo [13], added pytest task discovery for uv workspaces [14], upgraded to TypeScript 7.0.2 [15], and rotated daily Eve example maintenance [16]. Activity: 53 commits, 56 PRs, 18 releases across all desks.

Action items

→ Upgrade vercel/ai to pick up typed completion bodies if you validate custom payloads vercel/ai [plan]
→ Apply flags SDK patches immediately to close Dependabot alerts (nanoid DoS, fast-uri XSS) vercel/flags [immediate]
→ Test Next.js literal exports in 'use cache' files if you use file-level caching directives vercel/next.js [monitor]
→ Pull Turborepo's TypeScript 7.0.2 bump and pytest discovery if you maintain Python test suites vercel/turborepo [plan]

References

[1] feat(ui): add typed custom bodies to Completion APIs ↗ vercel/ai
[2] feat (provider/xai): add reference audio for reference-to-video ↗ vercel/ai
[3] feat (provider/xai): support new grok imagine video 1.5 caps ↗ vercel/ai
[4] chore: fix open Dependabot security alerts (#470) vercel/flags
[5] @flags-sdk/statsig@0.3.1 ↗ vercel/flags
[6] @flags-sdk/launchdarkly@1.1.1 ↗ vercel/flags
[7] @flags-sdk/hypertune@0.3.4 ↗ vercel/flags
[8] @flags-sdk/growthbook@0.3.1 ↗ vercel/flags
[9] Allow literal exports in `'use cache'` files ↗ vercel/next.js
[10] Fix shared Turbopack runtime initialization race ↗ vercel/next.js
[11] Turbopack: accept a module type argument for import.meta.glob (#96991) vercel/next.js
[12] [turbopack] Retain fewer stale cache versions and use a TTL ↗ vercel/next.js
[13] chore: Validate updated examples with turbo ↗ vercel/turborepo
[14] feat: Add pytest task discovery ↗ vercel/turborepo
[15] chore: Upgrade TypeScript to 7.0.2 ↗ vercel/turborepo
[16] feat: Rotate Eve example maintenance daily ↗ vercel/turborepo

Quick answers

What shipped in Vercel on August 12, 2026?: Vercel/ai shipped typed generics for custom completion bodies across all frameworks, while xAI Grok Imagine Video 1.5 integration arrives with 1080p native output and reference voice support. In total, 53 commits, 56 pull requests, and 18 releases landed.
Who contributed to Vercel on August 12, 2026?: 12 developers shipped this update, including n1ckoates, joshualipman123, github-actions, gr2m, Dominik Ferber, github-actions[bot], unstubbable, and gnoff, and 4 more.
What were the notable Vercel updates?: feat(ui): add typed custom bodies to Completion APIs, feat (provider/xai): add reference audio for reference-to-video, and feat (provider/xai): support new grok imagine video 1.5 caps.

WORKFLOW FIXES CORRUPTED EVENT LOG RACE CONDITION WHILE NEXT.JS ROLLS OUT RDC COMPRESSION

Workflow patched a critical event-log corruption bug in hook-resume staging that could lose data under concurrent replays, while Next.js adds controlled rollout flags for Resume Data Cache compression.

javascript 185 shipped

@vercel Aug 11

WORKFLOW EVENT ORDERING FIXED, SANDBOX PINNING NODE VERSIONS, EVE GAINS SANDBOX STOPS

Vercel shipped fixes across three major systems overnight: workflow's event consumer now respects delivery timing, sandbox images pin exact Node versions, and Eve lets authored code stop sandboxes mid-execution.

+12

javascript 126 shipped

@vercel Aug 10

PHOTON SETUP FIXED, AI SDK PATCHES REASONING TOKEN EDGE CASE

Eve's Photon integration now correctly detects dedicated projects and skips redundant registration, while the AI SDK patches a token-counting bug hitting reasoning models on Baseten.

javascript 25 shipped

@vercel Aug 9

TURBOPACK FIXES DEADLOCK THAT COULD HANG YOUR BUILDS

Luke Sandberg patched a critical threading bug in turbopack's scope_and_block primitive that could cause builds to hang under load.

javascript 10 shipped

Elsewhere on the wire

AI Agents Aug 13

ANTHROPIC INTEGRATION FIXES TOOL BLOCKS, OPENHANDS UNBLOCKS CI, CLINE PATCHES COMPOSER WEDGE

LangChain's Anthropic integration corrected tool result handling and model profiles [ref:1], while OpenHands fixed a critical test infrastructure failure blocking all PRs and shipped 1.13.0 with conversation archives [ref:6][ref:9].