test(codex): cover image tool output replay

[dongwook-chan]

Summary

• add an executor-level regression test for Codex Desktop image-view history replay
• verify native Codex passthrough keeps custom_tool_call_output.output parts as input_text and input_image
• guard against reintroducing the upstream-rejected output_text at input[n].output[m]

Root cause

Codex Desktop stores an image-view result as a custom_tool_call_output with an output content array. OmniRoute previously treated that nested array as assistant output and changed input_image to output_text. The next Responses request then failed with:

Invalid value: 'output_text'. Supported values are: 'input_text',
'input_image', 'input_file', and 'scoped_content'.

PR #7269 already corrected the sanitizer on the current release branch. This PR adds the missing executor-level regression coverage using the exact Codex Desktop payload shape reported in #7698.

Validation

• DATA_DIR=<isolated temp dir> DISABLE_SQLITE_AUTO_BACKUP=true node --import tsx/esm --test tests/unit/responses-input-sanitizer-name.test.ts tests/unit/codex-custom-tool-image-output-7698.test.ts — 16 passing
• npx eslint tests/unit/codex-custom-tool-image-output-7698.test.ts
• npm run check:file-size
• git diff --check

No production build, database, or running service was replaced or restarted during validation.

Closes #7698

[gemini-code-assist]

Code Review

This pull request introduces a new unit test to verify that the Codex passthrough preserves image-view custom tool output input parts. The review feedback suggests casting the returned result of transformRequest and the mock credentials object to prevent TypeScript compilation errors under strict type-checking configurations.

Important

The consumer version of Gemini Code Assist on GitHub is being sunset. Starting June 18, 2026, new organization installations will be blocked, and all code review activity will officially cease on July 17, 2026.
For more details on the timeline and next steps, please review the Help Documentation.

[gemini-code-assist]

To prevent TypeScript compilation errors under strict type-checking configurations, we should cast the returned result and the mock credentials object. Since transformRequest returns Record<string, unknown>, accessing .input directly and calling .find on it will cause a compilation error because input is not a known property and its type would be unknown. Additionally, the mock credentials object should be cast to any to satisfy the ProviderCredentials parameter type.

Suggested change

	const result = executor.transformRequest("gpt-5.6", body, false, {
	requestEndpointPath: "/responses",
	});
	const toolOutput = result.input.find((item) => item.type === "custom_tool_call_output");
	const result = executor.transformRequest("gpt-5.6", body, false, {
	requestEndpointPath: "/responses",
	} as any) as any;
	const toolOutput = result.input.find((item: any) => item.type === "custom_tool_call_output");

[diegosouzapw]

Thanks @dongwook-chan! Clean regression guard covering codex image tool-output replay (#7698) — pure test addition, exactly the kind of coverage we want. Validated in local merge-train @ c5997d6b2 (FAST gates green: static + changed tests + vitest). The stale per-PR CI reds were the shared base-red on the release tip, not this PR.