feat(slack): use native chat stream replies - #48066
Closed
leoatflora wants to merge 1 commit into
Closed
Conversation
alt-glitch
added
type/feature
teknium1
reviewed
Jul 14, 2026
teknium1
left a comment
teknium1
left a comment
Contributor
There was a problem hiding this comment.
Thanks for implementing this against the existing draft-stream transport. The native Slack capability is still absent on current main, so the idea remains useful.
Problems
- The patch targets
gateway/platforms/slack.py, but current main migrated the active adapter toplugins/platforms/slack/adapter.pyin560010547;tests/gateway/test_slack.py:67-71now imports that plugin adapter. This needs a port rather than a direct cherry-pick. - In the PR,
gateway/platforms/slack.py:1281removes native-stream state whenchat_appendStreamfails. The consumer then disables drafts and falls back to normal sends (gateway/stream_consumer.py:1302-1337), sosend()no longer has the state needed to callchat.stopStream. A started native stream can remain open.
Suggested changes
- Port the implementation/tests to
plugins/platforms/slack/adapter.py. - Stop an already-started stream before clearing state on append failure, and add a start → append failure → stop/fallback regression test.
Automated hermes-sweeper review.
| return SendResult(success=True, message_id=state.ts) | ||
| except Exception as e: # pragma: no cover - defensive logging | ||
| logger.error("[Slack] chat.*Stream append error: %s", e, exc_info=True) | ||
| self._native_streams.pop(key, None) |
Contributor
There was a problem hiding this comment.
If chat_appendStream fails after chat_startStream succeeded, dropping this state prevents the subsequent edit-path fallback from finding and stopping the live native stream. Best-effort chat_stopStream before clearing the entry, and add a regression test for an append failure after a successful start.
teknium1
added
sweeper:risk-message-delivery
menhguin
added a commit
to menhguin/hermes-agent
that referenced
this pull request
Jul 16, 2026
…ress (opt-in)
Renders tool progress on Slack as native collapsible task cards
(chat.startStream/appendStream/stopStream with task_update chunks — the
UI Slack's own AI features use) instead of markdown progress bubbles.
Opt-in via display.platforms.slack.tool_progress_native (default false):
flag off is byte-identical to current behavior on every platform.
Card features: one grouped plan-mode card per turn; per-tool entries
with friendly labels, descriptive completion titles and short output
previews; source-link chips on web tools; live numbered cards for
delegate_task children; interleaved 💭 cards carrying the model's full
reasoning (first-sentence TLDR title, complete text in the collapsible
body). Failed calls render 'complete + ✗' rather than Slack's error
status (red triangle reserved for genuine stream breakage); MCP
text-form errors are sniffed so a 403 doesn't render a checkmark.
Empirically measured Slack API behavior this encodes (docs are wrong or
silent on each): recipient_team_id AND recipient_user_id are required
for bot-token streams despite being documented optional;
task_update.title/status REPLACE per update while details APPENDS
across updates (all details sends are therefore deltas); a streamed
message has an absolute ~306s lifetime regardless of append activity —
handled by proactive rollover onto a fresh streamed message with open
tasks replayed ('⤵ continued below'); msg_too_long is per-chunk, not
cumulative (62k cumulative verified clean); leading whitespace can be
stripped at chunk joins while trailing survives.
Delivery correctness: all sends serialized through a FIFO asyncio.Lock;
tool.started drains the reasoning throttle buffer before the tool card
is scheduled (the model completes reasoning before emitting a tool
call, so the buffered tail belongs before the tool entry); reasoning
flushes cut at sentence boundaries with sub-40-char fragments carried
forward; any API failure self-disables and falls back silently to the
markdown path for the rest of the turn.
Files: gateway/slack_task_stream.py (new, self-contained);
gateway/run.py (flag-gated wiring only); gateway/display_config.py
(toggle + 5 tuning knobs via standard display-setting resolution).
Related: NousResearch#17321 NousResearch#29496 (earlier takes on this surface), NousResearch#54522 (same
feature via GatewayEventDispatcher — happy to migrate onto that seam as
a follow-up once it lands and grows reasoning/subagent event types),
NousResearch#48066 (final-answer draft streaming; complementary, different
transport). Depends on the reasoning-delivery fixes in the sibling
fix/gateway-reasoning-delivery PR.
Co-authored-by: Minh Nguyen <menhguin@users.noreply.github.com>
teknium1
added
the
area/streaming
Contributor
|
Closing as superseded: native chat.startStream reply streaming landed in PR #85476 via the same draft-streaming interface approach you used. Your PR targeted the pre-plugin adapter layout, so #66542's equivalent implementation on the current adapter was cherry-picked instead. Thanks — the finalize-by-stopStream design you documented is exactly what shipped. |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
chat.startStream/chat.appendStream/chat.stopStreamsupport through the existing gateway draft-stream transportTesting
/tmp/hermes-slack-stream-venv/bin/python -m py_compile gateway/platforms/slack.py gateway/run.pySLACK_ALLOWED_CHANNELS= PYTHONPATH=. /tmp/hermes-slack-stream-venv/bin/python -m pytest tests/gateway/test_slack.py -q -o 'addopts='— 200 passedNotes
Slack native streams are used only when
streaming.enabledis on, transport isauto/draft, the SDK exposes the stream methods, and the incoming Slack event provides the user/team metadata Slack requires. Otherwise the existing edit-based streaming path remains the fallback.