mirrors/hermes-agent
1
0
Fork 0
mirror of https://github.com/NousResearch/hermes-agent.git synced 2026-04-25 00:51:20 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

fix(gateway): prevent duplicate messages on no-message-id platforms

Browse source 
Platforms that don't return a message_id after the first send (Signal,
GitHub webhooks) were causing GatewayStreamConsumer to re-enter the
"first send" path on every tool boundary, posting one platform message
per tool call (observed as 155 PR comments on a single response).

Fix: treat _message_id == "__no_edit__" as a sentinel meaning "platform
accepted the send but cannot be edited". When a tool boundary arrives
in that state, skip the message_id/accumulated/last_sent_text reset so
all continuation text is delivered once via _send_fallback_final rather
than re-posted per segment.

Also make prompt_toolkit imports in hermes_cli/commands.py optional so
gateway and test environments that lack the package can still import
resolve_command, gateway_help_lines, and COMMAND_REGISTRY.
This commit is contained in:
KUSH42 2026-04-10 01:25:49 +02:00 • committed by Teknium
parent b1e2b5ea74
commit 5dea7e1ebc
3 changed files with 65 additions and 7 deletions
Download patch file Download diff file Expand all files Collapse all files

19
gateway/stream_consumer.py
View file

@ -205,11 +205,20 @@ class GatewayStreamConsumer:
await self._send_or_edit(self._accumulated)
return
# Tool boundary: the should_edit block above already flushed
# accumulated text without a cursor. Reset state so the next
# text chunk creates a fresh message below any tool-progress
# messages the gateway sent in between.
if got_segment_break:
# Tool boundary: reset message state so the next text chunk
# creates a fresh message below any tool-progress messages.
#
# Exception: when _message_id is "__no_edit__" the platform
# never returned a real message ID (e.g. Signal, webhook with
# github_comment delivery). Resetting to None would re-enter
# the "first send" path on every tool boundary and post one
# platform message per tool call — that is what caused 155
# comments under a single PR. Instead, keep all state so the
# full continuation is delivered once via _send_fallback_final.
# (When editing fails mid-stream due to flood control the id is
# a real string like "msg_1", not "__no_edit__", so that case
# still resets and creates a fresh segment as intended.)
if got_segment_break and self._message_id != "__no_edit__":
self._message_id = None
self._accumulated = ""
self._last_sent_text = ""