fix(agent-core-v2): coerce string-typed tool arguments at the parse boundary#2175
Open
HelloWorldU wants to merge 3 commits into
Open
fix(agent-core-v2): coerce string-typed tool arguments at the parse boundary#2175HelloWorldU wants to merge 3 commits into
HelloWorldU wants to merge 3 commits into
Conversation
🦋 Changeset detectedLatest commit: 1cc8d4f The changes in this PR will be included in the next version bump. This PR includes changesets to release 1 package
Not sure what this means? Click here to learn what changesets are. Click here if you're a maintainer who wants to add another changeset to this PR |
Author
|
@codex review |
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 64fa779d34
ℹ️ About Codex in GitHub
Codex has been enabled to automatically 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 👍.
When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".
…ion in args normalization
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Related Issue
Resolve #2118
Problem
Some third-party models emit tool-call arguments as JSON strings (
"3") instead of the declared JSON types (3) — reported in the linked issue on 0.29.0 with Qwen. The arguments are decoded faithfully and validated strictly, so the call is rejected withInvalid args for tool ...and the turn stalls, even though the model's intent is unambiguous.What changed
The fix sits at the tool-call parse boundary — the earliest point the harness controls — and keeps validation itself strict. Two moves, as discussed in the issue:
A normalization layer before validation. It is generic and bounded by design, so it stays a single schema-driven rule rather than growing into a catalog of model-specific hacks: only lossless
string → integer / number / booleancoercions, only when the schema does not already accept strings at that position;$refnodes and ambiguous schemas are left untouched.A warning that keeps the provider's violation visible. Coerced calls succeed, but the tool result carries a model-visible note (rendered to the model at projection time, never shown in UIs) describing exactly what was coerced — the non-conforming output is absorbed without being silently hidden.
When coercion cannot fix the arguments, the rejection now says what actually arrived (e.g.
/line_offset must be integer (received string "3")) so the model can self-correct, and duplicate type errors from union branches are deduplicated.Scope: the v2 engine only, which is what the current CLI runs on. The v1 engine has an isomorphic parse/validate path; happy to mirror the change there if the team wants both covered.
Tests: unit coverage for the normalization rules and the new error wording, plus executor integration tests covering coercion with the warning note, note merging with the tool's own note, honest rejection, and clean passthrough. Roughly half of the +571/−11 diff is tests.
Checklist
gen-changesetsskill, or this PR needs no changeset.gen-docsskill, or this PR needs no doc update.