Problem
Reasoning response items can include text inside a content array, but the transcript parser only reads text from summary arrays or string content.
That means reasoning emitted as content parts is silently dropped from parsed model steps, so TraceRoot can miss useful reasoning context in the generated spans.
Before
A reasoning item with content parts is parsed without reasoning text.
Result: step.reasoning is undefined.
Expected
The parser should capture text from reasoning content parts, the same way it already captures text from summary parts.
After
The same payload should produce the joined reasoning text:
Check the transcript shape.
Then emit the trace.
Why this matters
Reasoning text is useful observability context. Dropping it makes traces less complete and can make debugging Codex runs harder.
Problem
Reasoning response items can include text inside a
contentarray, but the transcript parser only reads text fromsummaryarrays or stringcontent.That means reasoning emitted as content parts is silently dropped from parsed model steps, so TraceRoot can miss useful reasoning context in the generated spans.
Before
A reasoning item with
contentparts is parsed without reasoning text.Result:
step.reasoningis undefined.Expected
The parser should capture text from reasoning content parts, the same way it already captures text from summary parts.
After
The same payload should produce the joined reasoning text:
Why this matters
Reasoning text is useful observability context. Dropping it makes traces less complete and can make debugging Codex runs harder.