fix(openai): stream compact responses to Codex clients

[specialpointcentral]

Summary

• preserve whether a compact request originally asked for streaming before compact body normalization removes stream
• keep upstream /responses/compact requests non-streaming, but wrap successful compact JSON responses as a single response.completed SSE event for streaming Codex clients
• cover OpenAI OAuth passthrough compact responses where upstream returns text/plain JSON

Close: #3875

Tests

• docker run --rm -v /tmp/sub2api-compact-pr:/workspace -w /workspace/backend golang:1.26.5 go test ./internal/service ./internal/handler -run 'TestOpenAIGatewayService_OAuthPassthrough_CompactClientStreamWrapsJSONAsSSE|TestOpenAIGatewayService_OAuthPassthrough_CompactOnlyModelMappingOverridesUpstreamModel|TestOpenAIGatewayService_Forward_CompactOnlyModelMappingOverridesOAuthUpstreamModel|TestNormalizeOpenAIResponsesCompactRequest_BodySignalPromoted|TestNormalizeOpenAIResponsesCompactRequest_PathBasedNoDoubleSuffix'

[Copilot]

Pull request overview

This PR adjusts the OpenAI /responses/compact flow to preserve whether a client originally requested streaming (even though the compact body normalization drops stream), and to support Codex clients that expect an SSE stream by wrapping successful compact JSON responses into a single response.completed SSE event.

Changes:

• Preserve the original compact request’s stream: true signal on the gin.Context before compact-body normalization removes it.
• When a compact request was originally streaming, return the non-streaming upstream compact JSON as a single response.completed SSE event (including OAuth passthrough where upstream may return text/plain JSON).
• Add test coverage for the SSE wrapping behavior and for propagation of the stream-requested flag.

Reviewed changes

Copilot reviewed 7 out of 7 changed files in this pull request and generated 1 comment.

Show a summary per file

File	Description
backend/internal/service/openai_gateway_service.go	Adds a context key to track whether a compact request originally requested streaming.
backend/internal/service/openai_gateway_response_handling.go	Wraps compact JSON responses as response.completed SSE for streaming Codex clients.
backend/internal/service/openai_gateway_request_body.go	Adds helpers to store/read the compact “client stream requested” flag on gin.Context.
backend/internal/service/openai_gateway_passthrough.go	Applies the same compact SSE wrapping behavior to OAuth passthrough non-streaming responses.
backend/internal/service/openai_compact_model_mapping_test.go	Adds a test validating SSE wrapping for OAuth passthrough compact responses (including text/plain JSON).
backend/internal/handler/openai_gateway_handler.go	Captures stream: true from compact request bodies before normalization drops stream.
backend/internal/handler/openai_gateway_compact_body_signal_test.go	Extends tests to assert the stream-requested flag is set when stream: true is present.

---

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

[specialpointcentral]

close because: #3880