Skip to content

bugfix: defer streaming chat role chunk until generation starts #1710

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
valarLip merged 1 commit into from
Jul 28, 2026
Merged

bugfix: defer streaming chat role chunk until generation starts #1710

Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
21 changes: 13 additions & 8 deletions atom/entrypoints/openai/serving_chat.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -86,11 +86,13 @@ async def stream_chat_response(
# finally aborts the still-running seq; normal completion flips it to False.
aborted = True
try:
# Send initial role chunk
yield create_chat_chunk(request_id, model, delta={"role": "assistant"})

role_sent = False
while True:
chunk_data = await stream_queue.get()

if not role_sent:
yield create_chat_chunk(request_id, model, delta={"role": "assistant"})
role_sent = True
new_text = chunk_data["text"]
num_tokens_output += len(chunk_data.get("token_ids", []))
_ct = chunk_data.get("num_cached_tokens", 0)
Expand Down Expand Up @@ -333,13 +335,16 @@ async def stream_chat_response_fanout(
# whichever siblings are still running.
aborted = True
try:
for i in range(n):
yield create_chat_chunk(
request_id, model, delta={"role": "assistant"}, index=i
)

role_sent = [False] * n
while not all(finished):
idx, chunk_data = await shared_queue.get()

if not role_sent[idx]:
yield create_chat_chunk(
request_id, model, delta={"role": "assistant"}, index=idx
)
role_sent[idx] = True

if finished[idx]:
# Defensive: should not happen, engine emits finished once per seq.
continue
Expand Down
Loading