Align Custom OAI tool_search availability

[PenguinDOOM]

This aligns tool_search static applicability with the existing gpt-5.4 and gpt-5.5 capability gate for Custom OAI Responses.

What changed:

• add gpt-5.4 and gpt-5.5 to the tool_search model allowlist
• add focused coverage for capability gating, agent tool availability, Responses request construction, and prompt alignment
• keep responsesApi.ts unchanged while proving native tool_search injection end to end

How this differs from #313803:

• Add toolSearch support for Custom OAI responses models #313803 introduced a broader Custom OAI configuration path by adding a new toolSearch setting and propagating it through schema, provider mapping, endpoint metadata, request handling, and Responses request shaping
• this PR keeps the scope narrower and treats the existing supportsToolSearch capability as the source of truth
• instead of adding new Custom OAI configuration or metadata plumbing, this PR only aligns the static tool_search allowlist with the capability path that already admits gpt-5.4 and gpt-5.5
• responsesApi.ts stays unchanged here because the request-level path already works once tool_search reaches the request tool list

Validation:

• npm run gulp compile-extensions
• npm run test:unit -- src/platform/endpoint/test/node/chatModelCapabilities.spec.ts
• npm run test:unit -- src/extension/intents/node/test/backgroundTodoEnablement.spec.ts
• npm run test:unit -- src/platform/endpoint/node/test/responsesApiToolSearch.spec.ts
• npm run test:unit -- src/extension/prompts/node/agent/test/agentPrompt.spec.tsx

[Copilot]

Pull request overview

Aligns tool_search availability for Custom OAI Responses with the existing gpt-5.4 / gpt-5.5 capability gating, and adds focused tests to validate the end-to-end behavior across capability checks, tool enablement, request construction, and prompt guidance.

Changes:

• Added gpt-5.4 and gpt-5.5 to the model allowlist for the tool_search tool definition.
• Expanded unit/integration coverage for capability gating, agent tool enablement, and Responses API request body tool injection behavior.
• Added prompt-level tests ensuring deferred-tool guidance and tool-search sections are gated by endpoint.supportsToolSearch.

Reviewed changes

Copilot reviewed 6 out of 6 changed files in this pull request and generated 1 comment.

Show a summary per file

File	Description
extensions/copilot/src/platform/endpoint/test/node/chatModelCapabilities.spec.ts	Refines/extends tests for modelSupportsToolSearch including explicit gpt-5.4/5.5 flag-enabled/disabled cases.
extensions/copilot/src/platform/endpoint/node/test/responsesApiToolSearch.spec.ts	Adds coverage for client tool_search injection behavior for CustomOAI Responses on gpt-5.4/5.5 and omission when upstream tools exclude it.
extensions/copilot/src/extension/tools/node/toolSearchTool.ts	Allows tool_search to apply to exact gpt-5.4 and gpt-5.5 model IDs.
extensions/copilot/src/extension/tools/node/test/testToolsService.ts	Updates test tools service to surface applicable model-specific tool definitions during tool enablement.
extensions/copilot/src/extension/prompts/node/agent/test/agentPrompt.spec.tsx	Adds tests to ensure tool-search guidance/list sections are gated by endpoint.supportsToolSearch for gpt-5.4/5.5.
extensions/copilot/src/extension/intents/node/test/backgroundTodoEnablement.spec.ts	Adds tests validating tool_search presence/absence in agent tools for CustomOAI gpt-5.4/5.5 based on endpoint capability.

[PenguinDOOM]

#313098
Pull request for this issue