TanStack
OSS
ai
Sign in / Sign up
Open main menu
ai
GitHub
Overview
Runs
Analytics
Loading workspace stats
Loading workspace insights...
Statistics interval
7 days
30 days
Latest CI Pipeline Executions
Status
Fix filter
Filter
Fuzzy
Filter range
Sort by
Sort by
Start time
Sort ascending
Sort descending
Succeeded
682-claude-structured-output-hard-fails-on-large-schemas-instead-of-falling-back-compiled-grammar-is-too-large
68dee5f2 Merge dcbc2ac8d875f8623cfa30c389560399e8033358 into 769a6c798d7984acdc12da395515f3c8be12f726
by Tom Beckenham
T
Succeeded
682-claude-structured-output-hard-fails-on-large-schemas-instead-of-falling-back-compiled-grammar-is-too-large
68dee5f2 docs: document Anthropic structured-output schema complexity limits (#682) Large/complex schemas are rejected by Anthropic's grammar compiler ("Schema is too complex for compilation" / "compiled grammar is too large"), directly and for anthropic/* models via OpenRouter. Document the errors, the common schema constructs that trigger them (optionals, catch/default wrappers, unions, unconstrained strings), and link to Anthropic's docs for current limits rather than copying numbers that change. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
by Tom Beckenham
T
Succeeded
682-claude-structured-output-hard-fails-on-large-schemas-instead-of-falling-back-compiled-grammar-is-too-large
cdd9f336 feat(ai): detect Anthropic's documented "Schema is too complex" rejection Widen the structured-output schema-rejection predicate beyond the observed "compiled grammar is too large" message to also match Anthropic's canonical 400 message "Schema is too complex for compilation", so `structuredOutput: 'auto'` falls back on the documented error too. Applies to both @tanstack/ai-anthropic and @tanstack/ai-openrouter. Document the boundary in the structured-outputs overview and skill: detection is reactive (we never pre-check schemas against Anthropic's caps), `'auto'` recovers from grammar-size rejections only, and the numeric limits are linked out rather than copied (they change and aren't all published). Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
by Tom Beckenham
T
Succeeded
682-claude-structured-output-hard-fails-on-large-schemas-instead-of-falling-back-compiled-grammar-is-too-large
cdd9f336 Merge bc43c3070ed2668364495450bcc14d2949ff73e2 into 8aabeec5fa4e75575dad7c26e6900d0e66165f91
by Tom Beckenham
T
Succeeded
682-claude-structured-output-hard-fails-on-large-schemas-instead-of-falling-back-compiled-grammar-is-too-large
d5ad7d4e feat(ai): fall back to forced-tool structured output on schema rejection Add `structuredOutput: 'auto' | 'native' | 'tool'` to `chat()` (default 'auto'). When a provider rejects a large structured-output schema (Anthropic "compiled grammar is too large", directly or for anthropic/* via OpenRouter), the engine transparently retries via the lenient forced-tool path instead of hard-failing. Detection is delegated to a new optional adapter predicate `isStructuredOutputSchemaError`; @tanstack/ai-anthropic and @tanstack/ai-openrouter implement it, and OpenRouter gains a forced-tool structured-output mode. Closes #682. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
by Tom Beckenham
T
Succeeded
682-claude-structured-output-hard-fails-on-large-schemas-instead-of-falling-back-compiled-grammar-is-too-large
d5ad7d4e Merge a07718d9c11e03ee5047383cd7394dc97eb89730 into 8aabeec5fa4e75575dad7c26e6900d0e66165f91
by Tom Beckenham
T
Previous page
Previous
Next
Next page