fix(misc): remove duplicate docs page, update clopus 4.7

[waleedlatif1]

Summary

• remove duplicate docs page, update clopus 4.7

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
docs	Ready	Preview, Comment	Apr 16, 2026 7:35pm

[cursor]

PR Summary

Medium Risk
Changes LLM request payload composition and model capability metadata, which can alter Claude Opus 4.7 behavior (temperature/thinking options) at runtime. Docs-only changes are low risk.

Overview
Removes the duplicate custom-tools docs page from the top-level docs nav and consolidates content under tools/custom-tools.mdx, adding a Custom Tools vs MCP Tools comparison and updating the docs to state custom tool deletion requires Admin.

Updates Anthropic request construction to only send temperature when the selected model supports it, and revises the claude-opus-4-7 model definition by removing temperature capability and adding the xhigh thinking level.

^{Reviewed by Cursor Bugbot for commit 7ab2e69. Configure here.}

[greptile-apps]

Greptile Summary

This PR does two things: removes a duplicate custom-tools docs page (the content already lives at tools/custom-tools.mdx and is registered in tools/meta.json) and adds claude-opus-4-7 to the model registry with adaptive-thinking-only support and a new xhigh effort level.

• claude-opus-4-7 is correctly wired through supportsAdaptiveThinking() in core.ts, which already checks for opus-4-7 strings, so adaptive thinking flows end-to-end.
• The new xhigh effort level is unique to Opus 4.7 among Anthropic models (Opus 4.6 / Sonnet 4.6 only go up to max). If xhigh is not a recognized value in Anthropic's output_config.effort API, any user who selects it will receive an API error.

Confidence Score: 4/5

Safe to merge after confirming 'xhigh' is a valid Anthropic API effort value for Opus 4.7.

All changes are straightforward — docs deduplication is clean and the model wiring through supportsAdaptiveThinking is correct. The only open question is whether 'xhigh' is a valid Anthropic API value for output_config.effort; if it is not, users selecting that thinking level will hit a hard API error in production.

apps/sim/providers/models.ts — verify the 'xhigh' thinking level is a recognized Anthropic API effort value.

Important Files Changed

Filename	Overview
apps/sim/providers/models.ts	Adds claude-opus-4-7 with contextWindow 1M, maxOutputTokens 128k, nativeStructuredOutputs, and thinking levels ['low','medium','high','xhigh','max'] — the 'xhigh' level is new and not present in any other Anthropic model's thinking levels.
apps/sim/providers/anthropic/core.ts	supportsAdaptiveThinking() already includes 'opus-4-7' check; buildThinkingConfig correctly routes Opus 4.7 to the adaptive path; no logic changes required beyond the models.ts addition.
apps/docs/content/docs/en/meta.json	Removes the top-level 'custom-tools' page reference; content remains accessible via the 'tools' section entry which already lists custom-tools in its own meta.json.
apps/docs/content/docs/en/tools/custom-tools.mdx	Authoritative custom tools documentation page; unchanged, confirming the deleted custom-tools/index.mdx was indeed a duplicate.

Flowchart

%%{init: {'theme': 'neutral'}}%%
flowchart TD
    A[Request with thinkingLevel] --> B{buildThinkingConfig}
    B --> C{supportsAdaptiveThinking?}
    C -- "opus-4-7 ✓" --> D["thinking: { type: 'adaptive' }\noutputConfig: { effort: thinkingLevel }"]
    C -- "No (e.g. opus-4-5)" --> E["thinking: { type: 'enabled', budget_tokens: N }"]
    D --> F["Anthropic API: output_config.effort = 'low'|'medium'|'high'|'xhigh'|'max'"]
    E --> G["Anthropic API: budget_tokens = 2048/8192/32768"]
    F --> H{xhigh valid API value?}
    H -- Yes --> I[Success]
    H -- No --> J[API Error returned to user]

Loading

_{Reviews (1): Last reviewed commit: "fix(misc): remove duplicate docs page, u..." | Re-trigger Greptile}

[waleedlatif1]

Re: Greptile review — xhigh is confirmed valid per Anthropic's official migration guide and adaptive thinking docs. It's documented as an exclusively Opus 4.7 effort level (output_config.effort) that maps 1:1 to the API — no translation layer needed. Inline thread resolved.

[waleedlatif1]

Re: Cursor review — The medium-risk assessment is correct. Both the temperature guard (supportsTemperature check before setting payload temperature) and the xhigh effort level have been verified against Anthropic's official docs. The xhigh value is a documented Anthropic API value for output_config.effort on Opus 4.7, and the temperature deprecation is a confirmed breaking change per the migration guide. All good to ship.