fix(translator): support message.reasoning fallback in responses non-stream

[warelik]

Problem

When translating OpenAI Chat Completions responses to OpenAI Responses API in non-streaming mode (ConvertOpenAIChatCompletionsResponseToOpenAIResponsesNonStream), reasoning content was extracted solely from choices.0.message.reasoning_content. When upstream providers (such as o1/o3 via compatible gateways, Grok, DeepSeek, OpenRouter, or third-party OpenAI endpoints) return reasoning under choices.0.message.reasoning, the reasoning text was completely omitted from the converted Responses output.

In the same file, the streaming converter (ConvertOpenAIChatCompletionsResponseToOpenAIResponses) already fell back to reasoning when reasoning_content was absent or empty, causing an unintended discrepancy between streaming and non-streaming responses.

Fix

• internal/translator/openai/openai/responses/openai_openai-responses_response.go:889: Updated ConvertOpenAIChatCompletionsResponseToOpenAIResponsesNonStream to check choices.0.message.reasoning_content and fall back to choices.0.message.reasoning if missing or empty, establishing stream/non-stream parity while preserving reasoning_content priority.

Tests

• TestConvertOpenAIChatCompletionsResponseToOpenAIResponsesNonStream_ReasoningFallback:
  • reasoning_content field present: Verifies standard reasoning_content extraction.
  • reasoning fallback field present: Verifies fallback to reasoning when reasoning_content is absent.
  • both reasoning_content and reasoning present (reasoning_content priority): Verifies priority.
  • empty reasoning_content falls back to reasoning: Verifies fallback when reasoning_content is "".
  • neither field present without request reasoning: Verifies no reasoning item is emitted.
  • neither field present with request reasoning produces empty summary: Verifies empty summary when requested.

Reverse bite-check

Reverting internal/translator/openai/openai/responses/openai_openai-responses_response.go to the pre-fix state causes TestConvertOpenAIChatCompletionsResponseToOpenAIResponsesNonStream_ReasoningFallback to fail on both fallback subtests:

=== RUN   TestConvertOpenAIChatCompletionsResponseToOpenAIResponsesNonStream_ReasoningFallback
=== RUN   TestConvertOpenAIChatCompletionsResponseToOpenAIResponsesNonStream_ReasoningFallback/reasoning_content_field_present
=== RUN   TestConvertOpenAIChatCompletionsResponseToOpenAIResponsesNonStream_ReasoningFallback/reasoning_fallback_field_present
    openai_openai-responses_response_test.go:1331: reasoning found = false, want true; out={"id":"chatcmpl_r","object":"response","created_at":1773896263,"status":"completed","background":false,"error":null,"incomplete_details":null,"model":"o3-mini","output":[{"id":"msg_chatcmpl_r_0","type":"message","status":"completed","content":[{"type":"output_text","annotations":[],"logprobs":[],"text":"hello"}],"role":"assistant"}]}
=== RUN   TestConvertOpenAIChatCompletionsResponseToOpenAIResponsesNonStream_ReasoningFallback/both_reasoning_content_and_reasoning_present_(reasoning_content_priority)
=== RUN   TestConvertOpenAIChatCompletionsResponseToOpenAIResponsesNonStream_ReasoningFallback/empty_reasoning_content_falls_back_to_reasoning
    openai_openai-responses_response_test.go:1331: reasoning found = false, want true; out={"id":"chatcmpl_empty_rc","object":"response","created_at":1773896263,"status":"completed","background":false,"error":null,"incomplete_details":null,"model":"o3-mini","output":[{"id":"msg_chatcmpl_empty_rc_0","type":"message","status":"completed","content":[{"type":"output_text","annotations":[],"logprobs":[],"text":"hello"}],"role":"assistant"}]}
=== RUN   TestConvertOpenAIChatCompletionsResponseToOpenAIResponsesNonStream_ReasoningFallback/neither_field_present_without_request_reasoning
=== RUN   TestConvertOpenAIChatCompletionsResponseToOpenAIResponsesNonStream_ReasoningFallback/neither_field_present_with_request_reasoning_produces_empty_summary
--- FAIL: TestConvertOpenAIChatCompletionsResponseToOpenAIResponsesNonStream_ReasoningFallback (0.00s)
    --- PASS: TestConvertOpenAIChatCompletionsResponseToOpenAIResponsesNonStream_ReasoningFallback/reasoning_content_field_present (0.00s)
    --- FAIL: TestConvertOpenAIChatCompletionsResponseToOpenAIResponsesNonStream_ReasoningFallback/reasoning_fallback_field_present (0.00s)
    --- PASS: TestConvertOpenAIChatCompletionsResponseToOpenAIResponsesNonStream_ReasoningFallback/both_reasoning_content_and_reasoning_present_(reasoning_content_priority) (0.00s)
    --- FAIL: TestConvertOpenAIChatCompletionsResponseToOpenAIResponsesNonStream_ReasoningFallback/empty_reasoning_content_falls_back_to_reasoning (0.00s)
    --- PASS: TestConvertOpenAIChatCompletionsResponseToOpenAIResponsesNonStream_ReasoningFallback/neither_field_present_without_request_reasoning (0.00s)
    --- PASS: TestConvertOpenAIChatCompletionsResponseToOpenAIResponsesNonStream_ReasoningFallback/neither_field_present_with_request_reasoning_produces_empty_summary (0.00s)
FAIL
FAIL	github.com/router-for-me/CLIProxyAPI/v7/internal/translator/openai/openai/responses	0.372s
FAIL

Verification

TMPDIR=/Users/warelik/.cache/gotmp GOCACHE=/Users/warelik/.cache/gocache go build ./...
TMPDIR=/Users/warelik/.cache/gotmp GOCACHE=/Users/warelik/.cache/gocache go vet ./internal/translator/openai/openai/responses/...
gofmt -l internal/translator/openai/openai/responses/openai_openai-responses_response.go internal/translator/openai/openai/responses/openai_openai-responses_response_test.go
TMPDIR=/Users/warelik/.cache/gotmp GOCACHE=/Users/warelik/.cache/gocache go test -v -run "TestConvertOpenAIChatCompletionsResponseToOpenAIResponsesNonStream_ReasoningFallback" ./internal/translator/openai/openai/responses/...

Output:

=== RUN   TestConvertOpenAIChatCompletionsResponseToOpenAIResponsesNonStream_ReasoningFallback
=== RUN   TestConvertOpenAIChatCompletionsResponseToOpenAIResponsesNonStream_ReasoningFallback/reasoning_content_field_present
=== RUN   TestConvertOpenAIChatCompletionsResponseToOpenAIResponsesNonStream_ReasoningFallback/reasoning_fallback_field_present
=== RUN   TestConvertOpenAIChatCompletionsResponseToOpenAIResponsesNonStream_ReasoningFallback/both_reasoning_content_and_reasoning_present_(reasoning_content_priority)
=== RUN   TestConvertOpenAIChatCompletionsResponseToOpenAIResponsesNonStream_ReasoningFallback/empty_reasoning_content_falls_back_to_reasoning
=== RUN   TestConvertOpenAIChatCompletionsResponseToOpenAIResponsesNonStream_ReasoningFallback/neither_field_present_without_request_reasoning
=== RUN   TestConvertOpenAIChatCompletionsResponseToOpenAIResponsesNonStream_ReasoningFallback/neither_field_present_with_request_reasoning_produces_empty_summary
--- PASS: TestConvertOpenAIChatCompletionsResponseToOpenAIResponsesNonStream_ReasoningFallback (0.00s)
    --- PASS: TestConvertOpenAIChatCompletionsResponseToOpenAIResponsesNonStream_ReasoningFallback/reasoning_content_field_present (0.00s)
    --- PASS: TestConvertOpenAIChatCompletionsResponseToOpenAIResponsesNonStream_ReasoningFallback/reasoning_fallback_field_present (0.00s)
    --- PASS: TestConvertOpenAIChatCompletionsResponseToOpenAIResponsesNonStream_ReasoningFallback/both_reasoning_content_and_reasoning_present_(reasoning_content_priority) (0.00s)
    --- PASS: TestConvertOpenAIChatCompletionsResponseToOpenAIResponsesNonStream_ReasoningFallback/empty_reasoning_content_falls_back_to_reasoning (0.00s)
    --- PASS: TestConvertOpenAIChatCompletionsResponseToOpenAIResponsesNonStream_ReasoningFallback/neither_field_present_without_request_reasoning (0.00s)
    --- PASS: TestConvertOpenAIChatCompletionsResponseToOpenAIResponsesNonStream_ReasoningFallback/neither_field_present_with_request_reasoning_produces_empty_summary (0.00s)
PASS
ok  	github.com/router-for-me/CLIProxyAPI/v7/internal/translator/openai/openai/responses	0.389s

Tooling note

The jbcontext semantic-search CLI is installed but non-functional in this environment
(jbcontext search fails with the OS keychain is not accessible). The equivalent
review passes were performed with the repository's own tooling and manual inspection
instead; this is disclosed for transparency about how the change was reviewed.