fix(wasm): advertise typed schemas by default

[G7CNF]

Use the extracted WASM schema as the advertised schema when it is typed and small enough, instead of always showing permissive {} parameters.

Fixes #1303.

[gemini-code-assist]

Summary of Changes

Hello, I'm Gemini Code Assist¹! I'm currently reviewing this pull request and will post my feedback shortly. In the meantime, here's a summary to help you and other reviewers quickly get up to speed!

This pull request enhances the WASM tool loading mechanism by intelligently advertising typed schemas to the Large Language Model (LLM) by default, rather than always presenting a permissive empty object schema. This change improves the accuracy of tool parameter descriptions for the LLM, while falling back to a permissive schema only when the typed schema is too large or untyped.

Highlights

• Schema Advertising Logic: Implemented logic to advertise typed WASM schemas directly to the LLM if they are sufficiently small, moving away from a default permissive schema.
• Schema Size Constraint: Introduced a maximum size limit for advertised schemas to prevent performance issues with large schemas.
• Improved Fallback Messaging: Updated warning messages in the WASM loader to clarify when and why permissive schemas are used as a fallback.
• Enhanced Testing: Added and modified unit tests to validate the new schema advertising behavior for both typed and untyped WASM schemas.

Using Gemini Code Assist

The full guide for Gemini Code Assist can be found on our documentation page, here are some quick tips.

Invoking Gemini

You can request assistance from Gemini at any point by creating a comment using either /gemini <command> or @gemini-code-assist <command>. Below is a summary of the supported commands on the current page.

Feature	Command	Description
Code Review	/gemini review	Performs a code review for the current pull request in its current state.
Pull Request Summary	/gemini summary	Provides a summary of the current pull request in its current state.
Comment	@gemini-code-assist	Responds in comments when explicitly tagged, both in pull request comments and review comments.
Help	/gemini help	Displays a list of available commands.

Customization

To customize Gemini Code Assist for GitHub experience, repository maintainers can create a configuration file and/or provide a custom code review style guide (such as PEP-8 for Python) by creating and adding files to a .gemini/ folder in the base of the repository. Detailed instructions can be found here.

Limitations & Feedback

Gemini Code Assist may make mistakes. Please leave feedback on any instances where its feedback is incorrect or counter productive. You can react with 👍 and 👎 on @gemini-code-assist comments. If you're interested in giving your feedback about your experience with Gemini Code Assist for GitHub and other Google products, sign up here.

Footnotes

1. Review the Privacy Notices, Generative AI Prohibited Use Policy, Terms of Service, and learn how to configure Gemini Code Assist in GitHub here. Gemini can make mistakes, so double check it and use code with caution. ↩

[gemini-code-assist]

Code Review

This pull request is a great improvement that allows WASM tools to advertise their typed schemas by default, as long as they are not too large. This will help the LLM make better-informed tool calls. The logic for deciding when to use the typed schema is sound, and the changes are well-tested. I have one suggestion to improve the memory efficiency of the schema size check, aligning with our WASM performance guidelines.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: c3a4a987fd

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

Validate exported WASM schemas before advertising them

If a WASM extension's schema() export is only partially valid (for example, it has one typed property but also an array property without items or an orphaned required key), this branch now publishes that raw schema to the LLM/provider because WasmToolSchemas::new() only checks typed_property_count() and size. Unlike sidecar schemas in src/tools/wasm/loader.rs, exported schemas never go through validate_tool_schema(), and the provider-side normalizer in src/llm/rig_adapter.rs only adds strict-mode fields; it does not repair those structural errors. Before this change we fell back to the permissive {} schema, so malformed exports did not break function-calling for the tool.

Useful? React with 👍 / 👎.

[henrypark133]

Review: Advertise typed WASM schemas to LLMs by default

Good improvement for LLM tool-use quality. Previously all WASM tools sent permissive schemas and relied on tool_info hints; now typed schemas are promoted directly when they exist and fit within budget.

Positives:

• MAX_ADVERTISED_SCHEMA_BYTES = 64KB guard prevents context bloat from huge schemas
• Dual-gate: typed_property_count > 0 AND size check — untyped schemas correctly stay permissive with tool_info hint fallback
• Loader log messages updated to reflect the new behavior
• Both paths tested: test_typed_wasm_schema_is_advertised_to_llm and test_untyped_wasm_schema_stays_permissive

LGTM.

[henrypark133]

Follow-up from review: A competing PR (#1348) independently implemented the same fix with a tighter size limit of 8KB (MAX_ADVERTISED_SCHEMA_BYTES = 8 * 1024) vs your 64KB. Since these schemas get serialized into every LLM call alongside all other tools, the 8KB limit is more appropriate for context budget management.

Could you tighten the limit from 64 * 1024 to 8 * 1024 before merge? The rest of your implementation is clean and already approved.

Requested tightening

[Copilot]

Pull request overview

Updates WASM tool schema advertising so that the LLM sees a typed exported schema by default (when it’s typed and below a size threshold), fixing cases where tools were invoked with {} due to overly-permissive advertised schemas.

Changes:

• Advertise the extracted WASM schema directly when it has typed properties and is reasonably small; otherwise keep the permissive {} fallback.
• Adjust WASM wrapper tests to validate typed vs. untyped advertised schema behavior and the presence/absence of the tool_info hint.
• Update loader warning messages related to missing/invalid parameters in capabilities sidecars.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 2 comments.

File	Description
src/tools/wasm/wrapper.rs	Introduces size-gated typed schema advertising and updates/adds tests for typed vs. untyped behavior.
src/tools/wasm/loader.rs	Refines warning messages around capabilities-sidecar parameter schema handling.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[ilblackdragon]

Review: fix(wasm): advertise typed schemas by default

This PR is superseded — recommend closing.

PR #1699 ("fix(wasm): use typed WASM schema as advertised schema when available") was already merged to staging on 2026-03-28 (commit 27e8d6f8). It implements a strictly more sophisticated solution to the same issue #1303:

• This PR's approach: Clone the full discovery schema as advertised when typed properties exist and schema is < 64 KB.
• #1699's approach (merged): compact_schema() extracts required and enum/const properties from oneOf/anyOf/allOf variants, producing a smaller, more LLM-friendly schema.

Both wrapper.rs and loader.rs have merge conflicts with current staging. Merging this PR would regress the schema compaction logic.

[ilblackdragon]

Closing as superseded by #1699