Skip to content

fix: insert separator between Responses API phase commentary and final text #312173

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
ulugbekna merged 3 commits into from
Apr 24, 2026
+112 −0
Merged

fix: insert separator between Responses API phase commentary and final text #312173

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
40168d0
test: add failing test for missing separator between phase commentary…
ulugbekna Apr 23, 2026
76ead44
fix: insert separator between output items in Responses API stream
ulugbekna Apr 23, 2026
1e29b8b
test: use last accumulated text instead of brittle index-from-end
ulugbekna Apr 23, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
8 changes: 8 additions & 0 deletions extensions/copilot/src/platform/endpoint/node/responsesApi.ts
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -841,6 +841,8 @@ export class OpenAIResponsesProcessor {
private sawCompactionMessage = false;
private latestCompactionOutputIndex: number | undefined;
private latestCompactionItem: OpenAIContextManagementResponse | undefined;
/** Tracks the output_index of the last text delta to detect output item boundaries */
private lastTextDeltaOutputIndex: number | undefined;
/** Maps output_index to { name, callId, arguments } for streaming tool call updates */
private readonly toolCallInfo = new Map<number, { name: string; callId: string; arguments: string }>();

Expand Down Expand Up @@ -915,6 +917,12 @@ export class OpenAIResponsesProcessor {
return onProgress({ text: '', copilotErrors: [{ agent: 'openai', code: chunk.code || 'unknown', message: chunk.message, type: 'error', identifier: chunk.param || undefined }] });
case 'response.output_text.delta': {
const capiChunk: CapiResponsesTextDeltaEvent = chunk;
// When text arrives from a new output item, emit a paragraph
// separator so that e.g. commentary and final text don't fuse.
if (this.lastTextDeltaOutputIndex !== undefined && capiChunk.output_index !== this.lastTextDeltaOutputIndex) {
onProgress({ text: '\n\n' });
}
this.lastTextDeltaOutputIndex = capiChunk.output_index;
const haystack = new Lazy(() => new TextEncoder().encode(capiChunk.delta));
return onProgress({
text: capiChunk.delta,
Expand Down
104 changes: 104 additions & 0 deletions extensions/copilot/src/platform/endpoint/node/test/responsesApi.spec.ts
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1184,3 +1184,107 @@ describe('summarizedAtRoundId and stateful marker interaction', () => {
services.dispose();
});
});

describe('phase commentary followed by phase final_answer', () => {
it('inserts a separator between commentary and final_answer text in the stream', async () => {
const services = createPlatformServices();
const accessor = services.createTestingAccessor();
const instantiationService = accessor.get(IInstantiationService);
const logService = accessor.get(ILogService);
const telemetryService = new SpyingTelemetryService();
const accumulatedTexts: string[] = [];
const phases: string[] = [];

const commentaryText = 'Responding directly in commentary as requested. My name is GitHub Copilot.';
const finalText = 'My name is GitHub Copilot.';

// Real-world Responses API stream: commentary message (output_index 0)
// followed by final_answer message (output_index 1), with incremental
// text deltas for each.
const events = [
{ type: 'response.output_item.added', output_index: 0, item: { type: 'message', role: 'assistant', content: [], phase: 'commentary', status: 'in_progress' }, sequence_number: 2 },
{ type: 'response.content_part.added', output_index: 0, content_index: 0, item_id: 'item-0', part: { type: 'output_text', text: '', annotations: [], logprobs: [] }, sequence_number: 3 },
{ type: 'response.output_text.delta', output_index: 0, content_index: 0, item_id: 'item-0', delta: 'Respond', logprobs: [], sequence_number: 4 },
{ type: 'response.output_text.delta', output_index: 0, content_index: 0, item_id: 'item-0', delta: 'ing', logprobs: [], sequence_number: 5 },
{ type: 'response.output_text.delta', output_index: 0, content_index: 0, item_id: 'item-0', delta: ' directly', logprobs: [], sequence_number: 6 },
{ type: 'response.output_text.delta', output_index: 0, content_index: 0, item_id: 'item-0', delta: ' in', logprobs: [], sequence_number: 7 },
{ type: 'response.output_text.delta', output_index: 0, content_index: 0, item_id: 'item-0', delta: ' commentary', logprobs: [], sequence_number: 8 },
{ type: 'response.output_text.delta', output_index: 0, content_index: 0, item_id: 'item-0', delta: ' as', logprobs: [], sequence_number: 9 },
{ type: 'response.output_text.delta', output_index: 0, content_index: 0, item_id: 'item-0', delta: ' requested', logprobs: [], sequence_number: 10 },
{ type: 'response.output_text.delta', output_index: 0, content_index: 0, item_id: 'item-0', delta: '.', logprobs: [], sequence_number: 11 },
{ type: 'response.output_text.delta', output_index: 0, content_index: 0, item_id: 'item-0', delta: ' My', logprobs: [], sequence_number: 12 },
{ type: 'response.output_text.delta', output_index: 0, content_index: 0, item_id: 'item-0', delta: ' name', logprobs: [], sequence_number: 13 },
{ type: 'response.output_text.delta', output_index: 0, content_index: 0, item_id: 'item-0', delta: ' is', logprobs: [], sequence_number: 14 },
{ type: 'response.output_text.delta', output_index: 0, content_index: 0, item_id: 'item-0', delta: ' Git', logprobs: [], sequence_number: 15 },
{ type: 'response.output_text.delta', output_index: 0, content_index: 0, item_id: 'item-0', delta: 'Hub', logprobs: [], sequence_number: 16 },
{ type: 'response.output_text.delta', output_index: 0, content_index: 0, item_id: 'item-0', delta: ' Cop', logprobs: [], sequence_number: 17 },
{ type: 'response.output_text.delta', output_index: 0, content_index: 0, item_id: 'item-0', delta: 'ilot', logprobs: [], sequence_number: 18 },
{ type: 'response.output_text.delta', output_index: 0, content_index: 0, item_id: 'item-0', delta: '.', logprobs: [], sequence_number: 19 },
{ type: 'response.output_text.done', output_index: 0, content_index: 0, item_id: 'item-0', text: commentaryText, logprobs: [], sequence_number: 20 },
{ type: 'response.content_part.done', output_index: 0, content_index: 0, item_id: 'item-0', part: { type: 'output_text', text: commentaryText, annotations: [], logprobs: [] }, sequence_number: 21 },
{ type: 'response.output_item.done', output_index: 0, item: { type: 'message', role: 'assistant', content: [{ type: 'output_text', text: commentaryText, annotations: [], logprobs: [] }], phase: 'commentary', status: 'completed' }, sequence_number: 22 },
{ type: 'response.output_item.added', output_index: 1, item: { type: 'message', role: 'assistant', content: [], phase: 'final_answer', status: 'in_progress' }, sequence_number: 23 },
{ type: 'response.content_part.added', output_index: 1, content_index: 0, item_id: 'item-1', part: { type: 'output_text', text: '', annotations: [], logprobs: [] }, sequence_number: 24 },
{ type: 'response.output_text.delta', output_index: 1, content_index: 0, item_id: 'item-1', delta: 'My', logprobs: [], sequence_number: 25 },
{ type: 'response.output_text.delta', output_index: 1, content_index: 0, item_id: 'item-1', delta: ' name', logprobs: [], sequence_number: 26 },
{ type: 'response.output_text.delta', output_index: 1, content_index: 0, item_id: 'item-1', delta: ' is', logprobs: [], sequence_number: 27 },
{ type: 'response.output_text.delta', output_index: 1, content_index: 0, item_id: 'item-1', delta: ' Git', logprobs: [], sequence_number: 28 },
{ type: 'response.output_text.delta', output_index: 1, content_index: 0, item_id: 'item-1', delta: 'Hub', logprobs: [], sequence_number: 29 },
{ type: 'response.output_text.delta', output_index: 1, content_index: 0, item_id: 'item-1', delta: ' Cop', logprobs: [], sequence_number: 30 },
{ type: 'response.output_text.delta', output_index: 1, content_index: 0, item_id: 'item-1', delta: 'ilot', logprobs: [], sequence_number: 31 },
{ type: 'response.output_text.delta', output_index: 1, content_index: 0, item_id: 'item-1', delta: '.', logprobs: [], sequence_number: 32 },
{ type: 'response.output_text.done', output_index: 1, content_index: 0, item_id: 'item-1', text: finalText, logprobs: [], sequence_number: 33 },
{ type: 'response.content_part.done', output_index: 1, content_index: 0, item_id: 'item-1', part: { type: 'output_text', text: finalText, annotations: [], logprobs: [] }, sequence_number: 34 },
{ type: 'response.output_item.done', output_index: 1, item: { type: 'message', role: 'assistant', content: [{ type: 'output_text', text: finalText, annotations: [], logprobs: [] }], phase: 'final_answer', status: 'completed' }, sequence_number: 35 },
{
type: 'response.completed',
response: {
id: 'resp_phase_test',
model: 'gpt-5.4-2026-03-05',
created_at: 1776962259,
usage: { input_tokens: 8432, output_tokens: 35, total_tokens: 8467, input_tokens_details: { cached_tokens: 0 }, output_tokens_details: { reasoning_tokens: 0 } },
output: [
{ type: 'message', content: [{ type: 'output_text', text: commentaryText, annotations: [], logprobs: [] }], phase: 'commentary', role: 'assistant', status: 'completed' },
{ type: 'message', content: [{ type: 'output_text', text: finalText, annotations: [], logprobs: [] }], phase: 'final_answer', role: 'assistant', status: 'completed' },
],
},
sequence_number: 36,
}
];

const sseBody = events.map(e => `data: ${JSON.stringify(e)}\n\n`).join('');
const response = createFakeStreamResponse(sseBody);
const telemetryData = TelemetryData.createAndMarkAsIssued({ modelCallId: 'model-call-phase-test' }, {});

const stream = await processResponseFromChatEndpoint(
instantiationService,
telemetryService,
logService,
response,
1,
async (text, _unused, delta) => {
accumulatedTexts.push(text);
if (delta.phase) {
phases.push(delta.phase);
}
return undefined;
},
telemetryData,
);

for await (const _ of stream) {
// consume stream
}

expect(phases).toEqual(['commentary', 'final_answer']);

// The accumulated text must separate commentary and final_answer text
const finalAccumulatedText = accumulatedTexts[accumulatedTexts.length - 1];
expect(finalAccumulatedText).toBe(
commentaryText + '\n\n' + finalText
);

accessor.dispose();
services.dispose();
});
});
Loading