fix: infer response media and Rust bodyless results from OpenAPI

[ChiragAgg5k]

Derive Accept and Rust response types from standard response content

PR 4/4. CLO-4377: https://linear.app/appwrite/issue/CLO-4377/openapi3-standards
Appwrite #13524 and canonical specs latest #110 / 2.0.x #111 are merged. Both published documents match the reviewed SHA256 7bf999d8faec746c1fb0ead1f0539dfaa8aead401405b2b697358e7059667f98 and contain no x-appwrite.produces. The branch is in sync with SDK-generator main; no parser release is required. Cloud's older server-ce pin remains unchanged; the owner explicitly approved publishing specs without waiting for that promotion.

Generated SDK changes

Actual generated Rust console account.delete_billing_address:

 pub async fn delete_billing_address(
     &self,
     billing_address_id: impl Into<String>,
-) -> crate::error::Result<serde_json::Value> {
+) -> crate::error::Result<()> {

Actual generated Rust account.delete now sends:

api_headers.insert("content-type".to_string(), "application/json".to_string());
api_headers.insert("accept".to_string(), "application/json".to_string());

Actual generated Node server adds this metadata for bodyless operations:

accept: 'application/json',

Spec evidence: genuine CE sitesDeleteLog after-generation still declares only:

"responses": {"204": {"description": "No content"}}

There is no response body, and no fallback extension. Explicit response content continues to determine Accept; binary/location and OAuth return branches are unchanged. Existing Swagger2 produces remains supported through the parser's standard response content conversion.

Breaking change / release notes

Rust methods previously inferred as Result<serde_json::Value> solely from the extension now return Result<()>. Callers binding those results must use () and must not expect JSON from a bodyless response. Keep this breaking API correction and the JSON Accept fallback in the eventual Rust SDK release notes; the changelog template is intentionally unchanged. Apply Rust SDK breaking-change version policy at publication.

Measured output

Fixed published canonical baseline (1,020 operations) has 117 bodyless operations: 49 with old extension and 68 without. Clean actual Rust generations show 11 client / 23 server / 38 console corrected signatures and 14 / 54 / 60 added Accept headers respectively. Counts depend on platform/exclusions/aliases. Other targets have only JSON Accept additions and necessary comma formatting; CLI/GraphQL output is unchanged. Rust has only approved signature/header changes, documentation signatures and strengthened integration harness.

Standard Swagger2 1.8.x server: explicit media and Rust signatures preserved; 39 JSON Accept insertions only where no produced content exists.

Validation

• No new unit tests. Added an output-shape check in the existing generation suite and strengthened the existing Rust integration call to require successful unit deserialization.
• New generation check observed red against baseline generator, green after. Full suite: 85 tests, 1,130 assertions, 4 existing skips.
• Docker e2e: Rust 1 test / 131 assertions; Node + PHP 2 tests / 298 assertions; Go 1 test / 134 assertions. PHP separately 1 / 142.
• PHP lint, Rector and Twig lint pass. Rust server cargo check and Rust client/server rustfmt pass.
• Go build/vet/test, Web/Node/React Native formatting/lint/types, Node build+830 tests, Dart, Kotlin/Android, Swift/Apple, C#/Unity, Python Black, Ruby RuboCop and PHP Pint/PHPStan/Rector pass.
• Existing published specs: 101 reference-valid documents; canonical and genuine CE before/after documents validator clean.

Existing gate limitations

• Rust console rustfmt reports unchanged long constants in usage_event_metric.rs; reproduced on baseline.
• Flutter formatter reports the same 20 files on baseline and after; no generated files were formatted in place.
• Unchanged CLI cannot build without missing dependency/go.sum entries; reproduced on baseline. CLI output itself has zero differences.
• Node's first test invocation lacked a build; rerun after npm run build passed all 830 tests.
• Canonical artifacts are now publicly available following genuine x86 Cloud generation and specs feat-flutter-dart-sdk-private-storage-files #110/fix-flutter-dart-code-quality #111. Generation suite rerun passes: 85 tests / 1,130 assertions. Remote CI and Validation checks passed at 788f49ae7932608a5269d5049cad10f3b6bea429; fresh checks follow the changelog-only revert. An additional local Rust e2e rerun initially encountered mock-server connectivity failures; one unchanged-code environment retry passed (1 test, 131 assertions).

Follow-through

Cloud pins sdk-generator 4.9.*; CE allows ^4.0. After choosing and publishing the generator version, plan a targeted Cloud dependency promotion if that range excludes it. No speculative version bump here.

Captured validation output

Actual command-output excerpts rendered locally and captured with Chromium. These are test-log screenshots, not live browser UI/e2e screenshots. Local configuration paths are excluded by selecting the public stdout excerpt; result text is unchanged.

Narrow regression and full generation suite passed at 788f49ae7932608a5269d5049cad10f3b6bea429, based on SDK-generator main f072d890699106748a9899a344a6932b31a41118. Follow-up b251f6eeb removes only the proposed changelog-template entries; Rust server was regenerated successfully. No new unit tests.

[greptile-apps]

Greptile Summary

This PR derives response media types from standard OpenAPI response content, defaults bodyless operations to JSON content negotiation, and updates Rust bodyless operations to return unit.

• Removes the legacy x-appwrite.produces fallback.
• Generates Accept: application/json when no response content type is declared.
• Maps bodyless Rust responses to crate::error::Result<()>.
• Strengthens generation and Rust integration coverage for the revised contract.
• The previously reported case mismatch is not an issue: the assertion consumes generated files normalized by strtolower(), and the thread was resolved after Greptile withdrew the finding.

Confidence Score: 5/5

The PR appears safe to merge with no outstanding correctness or repository-rule findings.

No new code changes were made after the previous review, and the sole previous finding was manually resolved after Greptile conceded it: GenerationTest::generate() lowercases generated file contents before the assertion, so the lowercase Rust return-type pattern is correct.

Important Files Changed

Filename	Overview
src/SDK/Extension/Appwrite.php	Removes the obsolete Appwrite-specific response media extension constant.
src/SDK/Language/Rust.php	Infers bodyless Rust unit results exclusively from standard response content.
src/SDK/SDK.php	Defaults generated Accept metadata to application/json when responses declare no content.
templates/rust/tests/tests.rs	Verifies successful unit deserialization for a generated bodyless Rust operation.
tests/generation/GenerationTest.php	Checks normalized generated Rust output for unit return types and JSON Accept headers.
tests/resources/spec-openapi3.json	Removes legacy produces metadata so tests exercise standard OpenAPI response inference.

_{Reviews (3): Last reviewed commit: "chore: leave Rust changelog template unc..." | Re-trigger Greptile}

[ChiragAgg5k]

@greptile review

[ChiragAgg5k]

@greptile review

Please reassess the resolved case-mismatch finding with the actual assertion input: GenerationTest::generate() at tests/generation/GenerationTest.php:194 applies strtolower() to every file before returning $files. The new test uses that normalized result, NOT raw Rust source. It correctly expects lowercase crate::error::result; changing it to uppercase would break the test. Current-head focused command vendor/bin/phpunit --testsuite Generation --filter testRustBodylessResponsesUseUnitWithJsonAccept passes (1 test, 3 assertions). See thread reply #1888 (comment). No source edit is needed for this false positive.