Skip to content

docs: clarify streaming lifecycle guidance #132

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
rapids-bot merged 2 commits into from
Jul 28, 2026
Merged

docs: clarify streaming lifecycle guidance #132

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
74b2f86
docs: clarify streaming lifecycle guidance
lbliii Jul 28, 2026
bc49b26
Merge release/0.1 into docs/terminology-style-follow-up
zhongxuanwang-nv Jul 28, 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
9 changes: 5 additions & 4 deletions skills/integrations/consumer/nemo-fabric-integrate/SKILL.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -158,12 +158,13 @@ Pick the smallest lifecycle the consumer needs:
failures remain normalized `RunResult` values. If iteration stops early,
call `await stream.aclose()` before starting another turn. `aclose()` waits
for the turn to finish; it does not cancel the harness invocation. The SDK
intentionally exposes ATOF records generated by NeMo Relay only; adapter-native progressive
output is deferred to a future normalized NeMo Fabric contract. The listener
intentionally exposes only ATOF records generated by NeMo Relay. A future
normalized NeMo Fabric contract will address adapter-native progressive
output. The listener
limits each record to 1 MiB and its queue to 1,024 records or 16 MiB of
encoded data. It correlates records through the NeMo Fabric request ID for
in-process harnesses or NeMo Relay's turn-scope role and 1-based turn index for
gateway harnesses, then yields only the matched scope tree. Delayed
in-process harnesses. For gateway harnesses, it uses the NeMo Relay turn-scope
role and 1-based turn index. It yields only the matched scope tree. Delayed
prior-turn records therefore do not enter the next stream. If gateway and
NeMo Fabric turn sequences do not align, the SDK discards the uncorrelated records
and emits a `RuntimeWarning` after natural stream exhaustion. The listener
Expand Down
Loading