Summary

• log semantic response diagnostics for phase, commentary, final-answer, and reasoning-summary fields
• keep the diagnostics on the existing logging path without changing the response transport
• add focused coverage for semantic diagnostic logging

Stack

• base: #314

note: greptile review for oc-chatgpt-multi-auth. cite files like lib/foo.ts:123. confirm regression tests + windows concurrency/token redaction coverage.

Greptile Summary

adds semantic response diagnostics logging to the sse-to-json conversion path. after parseSseStream assembles the final response, the new logStreamDiagnostics function reads the computed phase, commentary_text, final_answer_text, and reasoning_summary_text fields, truncates each to 400 chars, and emits a "stream-diagnostics" log entry only when at least one semantic field is present. no changes to the response transport.

• truncateDiagnosticText: safe unknown → string helper with 400-char cap; correctly returns undefined for non-strings and empty strings.
• logStreamDiagnostics: guarded by LOGGING_ENABLED and isRecord; omitted from the null-response code path; spreads only present fields into the log payload.
• beforeEach mock clear added to response-handler-logging.test.ts — correct fix that prevents call-count contamination across tests.
• new test covers the full happy path (phase, commentary, final-answer, reasoning-summary) against a realistic multi-event sse fixture.
• truncation branch in truncateDiagnosticText has no vitest coverage (strings > 400 chars never exercised).

Confidence Score: 5/5

• safe to merge — pure additive logging on the happy path, no transport changes, and existing tests are not broken.
• change is minimal and isolated: two small helper functions, one call site after the null-response guard, and a focused test addition. no concurrency concerns (logging is synchronous and fire-and-forget). no windows filesystem or token-safety risk. the only gap is the untested truncation branch, which is a P2 follow-up.
• no files require special attention.

Important Files Changed

Sequence Diagram

sequenceDiagram
    participant caller
    participant convertSseToJson
    participant parseSseStream
    participant logStreamDiagnostics
    participant logRequest

    caller->>convertSseToJson: response (SSE stream), headers
    convertSseToJson->>convertSseToJson: consume stream chunks
    convertSseToJson->>logRequest: "stream-full" { fullContent }
    convertSseToJson->>parseSseStream: fullText
    parseSseStream-->>convertSseToJson: finalResponse (or null)
    alt finalResponse is null
        convertSseToJson-->>caller: raw SSE response (passthrough)
    else finalResponse present
        convertSseToJson->>logStreamDiagnostics: finalResponse
        logStreamDiagnostics->>logStreamDiagnostics: extract phase, commentaryText,\nfinalAnswerText, reasoningSummaryText
        opt any semantic field present
            logStreamDiagnostics->>logRequest: "stream-diagnostics" { responseId, phase, ... }
        end
        convertSseToJson-->>caller: JSON response
    end

This is a comment left during a code review.
Path: lib/request/response-handler.ts
Line: 352-357

Comment:
**Missing vitest coverage for truncation branch**

the `trimmed.length > maxLength` branch has no test exercising it. the new test only passes strings well under 400 chars (`"Thinking..."`, `"Done."`, `"Need more context."`), so the slice-and-append path is never hit. given the 80% coverage threshold, it's worth adding a test with a string > 400 chars to verify the truncated output matches expectations and the `...` suffix is appended correctly.

How can I resolve this? If you propose a fix, please make it concise.

_{Reviews (3): Last reviewed commit: "Merge main into feat/openai-parity-pr7" | Re-trigger Greptile}