TanStackOSS
    ai
    fix(ai-react): reset structured-output state synchronously on user actions `useChat({ outputSchema })` only reset `partial`/`final` when the server's first chunk landed (`RUN_STARTED`). That left the previous run's values visible in the gap between dispatching a `sendMessage`/`reload`/`append`/ `clear` and the network response, which read as stale data in the UI — and never cleared at all when a request errored before any chunk arrived. Reset on the user action instead. The RUN_STARTED reset stays, since each agent-loop iteration emits its own RUN_STARTED and intermediate text from tool-decision runs must not leak into the final structured buffer. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
    nx run-many --targets=build --exclude=examples/**
Succeeded
CI Pipeline Execution

nx run-many --targets=build --exclude=examples/**

Click to copy
Linux
4 CPU cores
da486ebc576
88%

Cache hits

29 of 33 tasks used cache.

Get faster results

Learn how to enable distribution.

Atomizer enabled

33 groups of tasks optimized.

© 2026 - Nx Cloud

Terms of ServicePrivacy PolicyChangelogStatusDocsContact Nx CloudPricingCompany@NxDevTools