fix: deduplicate context pressure warnings in gateway mode#6309
Closed
duan78 wants to merge 1 commit into
Closed
Conversation
In gateway mode, a new AIAgent is created per message. The instance-level _context_pressure_warned flag was always False on fresh instances, causing the 'Context compaction approaching' warning to fire on every message when context was already above 85% of the compaction threshold. Fix: add a class-level dict _context_pressure_last_warned keyed by session_id that tracks the timestamp of the last warning emission. If less than 5 minutes have passed since the last warning for a given session, the warning is suppressed. The entry is cleared after compression drops context below the 85% threshold, allowing a fresh warning cycle. - run_agent.py: add class-level dedup dict + cooldown logic - tests/run_agent/test_context_pressure.py: add 3 dedup tests
teknium1
added a commit
that referenced
this pull request
Apr 9, 2026
Combines the approaches from PR #6309 (duan78) and PR #5963 (KUSH42): Tiered warnings (from #5963): - Replaces boolean _context_pressure_warned with float _context_pressure_warned_at - Fires at 85% (orange) and re-fires at 95% (red/critical) - Adds 'compacting context...' status message before compression Gateway dedup (from #6309): - Class-level dict _context_pressure_last_warned survives across AIAgent instances (gateway creates a new instance per message) - 5-minute cooldown per session prevents warning spam - Higher-tier warnings bypass the cooldown (85% → 95% always fires) - Compression reset clears the dedup entry for the session - Stale entries evicted (older than 2x cooldown) to prevent memory leak Does NOT inject into messages — purely user-facing via _safe_print (CLI) and status_callback (gateway). Zero prompt cache impact. Fixes #6309. Fixes #5963.
pull Bot
pushed a commit
to gmhl000/hermes-agent
that referenced
this pull request
Apr 9, 2026
…earch#6411) Combines the approaches from PR NousResearch#6309 (duan78) and PR NousResearch#5963 (KUSH42): Tiered warnings (from NousResearch#5963): - Replaces boolean _context_pressure_warned with float _context_pressure_warned_at - Fires at 85% (orange) and re-fires at 95% (red/critical) - Adds 'compacting context...' status message before compression Gateway dedup (from NousResearch#6309): - Class-level dict _context_pressure_last_warned survives across AIAgent instances (gateway creates a new instance per message) - 5-minute cooldown per session prevents warning spam - Higher-tier warnings bypass the cooldown (85% → 95% always fires) - Compression reset clears the dedup entry for the session - Stale entries evicted (older than 2x cooldown) to prevent memory leak Does NOT inject into messages — purely user-facing via _safe_print (CLI) and status_callback (gateway). Zero prompt cache impact. Fixes NousResearch#6309. Fixes NousResearch#5963.
saxster
pushed a commit
to saxster/hermes-agent
that referenced
this pull request
Apr 9, 2026
…earch#6411) Combines the approaches from PR NousResearch#6309 (duan78) and PR NousResearch#5963 (KUSH42): Tiered warnings (from NousResearch#5963): - Replaces boolean _context_pressure_warned with float _context_pressure_warned_at - Fires at 85% (orange) and re-fires at 95% (red/critical) - Adds 'compacting context...' status message before compression Gateway dedup (from NousResearch#6309): - Class-level dict _context_pressure_last_warned survives across AIAgent instances (gateway creates a new instance per message) - 5-minute cooldown per session prevents warning spam - Higher-tier warnings bypass the cooldown (85% → 95% always fires) - Compression reset clears the dedup entry for the session - Stale entries evicted (older than 2x cooldown) to prevent memory leak Does NOT inject into messages — purely user-facing via _safe_print (CLI) and status_callback (gateway). Zero prompt cache impact. Fixes NousResearch#6309. Fixes NousResearch#5963.
Tommyeds
pushed a commit
to Tommyeds/hermes-agent
that referenced
this pull request
Apr 12, 2026
…earch#6411) Combines the approaches from PR NousResearch#6309 (duan78) and PR NousResearch#5963 (KUSH42): Tiered warnings (from NousResearch#5963): - Replaces boolean _context_pressure_warned with float _context_pressure_warned_at - Fires at 85% (orange) and re-fires at 95% (red/critical) - Adds 'compacting context...' status message before compression Gateway dedup (from NousResearch#6309): - Class-level dict _context_pressure_last_warned survives across AIAgent instances (gateway creates a new instance per message) - 5-minute cooldown per session prevents warning spam - Higher-tier warnings bypass the cooldown (85% → 95% always fires) - Compression reset clears the dedup entry for the session - Stale entries evicted (older than 2x cooldown) to prevent memory leak Does NOT inject into messages — purely user-facing via _safe_print (CLI) and status_callback (gateway). Zero prompt cache impact. Fixes NousResearch#6309. Fixes NousResearch#5963.
angelburgosrosado
pushed a commit
to angelburgosrosado/hermes-agent
that referenced
this pull request
Apr 27, 2026
…earch#6411) Combines the approaches from PR NousResearch#6309 (duan78) and PR NousResearch#5963 (KUSH42): Tiered warnings (from NousResearch#5963): - Replaces boolean _context_pressure_warned with float _context_pressure_warned_at - Fires at 85% (orange) and re-fires at 95% (red/critical) - Adds 'compacting context...' status message before compression Gateway dedup (from NousResearch#6309): - Class-level dict _context_pressure_last_warned survives across AIAgent instances (gateway creates a new instance per message) - 5-minute cooldown per session prevents warning spam - Higher-tier warnings bypass the cooldown (85% → 95% always fires) - Compression reset clears the dedup entry for the session - Stale entries evicted (older than 2x cooldown) to prevent memory leak Does NOT inject into messages — purely user-facing via _safe_print (CLI) and status_callback (gateway). Zero prompt cache impact. Fixes NousResearch#6309. Fixes NousResearch#5963.
02356abc
pushed a commit
to 02356abc/hermes-agent
that referenced
this pull request
May 14, 2026
…earch#6411) Combines the approaches from PR NousResearch#6309 (duan78) and PR NousResearch#5963 (KUSH42): Tiered warnings (from NousResearch#5963): - Replaces boolean _context_pressure_warned with float _context_pressure_warned_at - Fires at 85% (orange) and re-fires at 95% (red/critical) - Adds 'compacting context...' status message before compression Gateway dedup (from NousResearch#6309): - Class-level dict _context_pressure_last_warned survives across AIAgent instances (gateway creates a new instance per message) - 5-minute cooldown per session prevents warning spam - Higher-tier warnings bypass the cooldown (85% → 95% always fires) - Compression reset clears the dedup entry for the session - Stale entries evicted (older than 2x cooldown) to prevent memory leak Does NOT inject into messages — purely user-facing via _safe_print (CLI) and status_callback (gateway). Zero prompt cache impact. Fixes NousResearch#6309. Fixes NousResearch#5963.
olympus-terminal
pushed a commit
to olympus-terminal/hermes-agent
that referenced
this pull request
May 16, 2026
…earch#6411) Combines the approaches from PR NousResearch#6309 (duan78) and PR NousResearch#5963 (KUSH42): Tiered warnings (from NousResearch#5963): - Replaces boolean _context_pressure_warned with float _context_pressure_warned_at - Fires at 85% (orange) and re-fires at 95% (red/critical) - Adds 'compacting context...' status message before compression Gateway dedup (from NousResearch#6309): - Class-level dict _context_pressure_last_warned survives across AIAgent instances (gateway creates a new instance per message) - 5-minute cooldown per session prevents warning spam - Higher-tier warnings bypass the cooldown (85% → 95% always fires) - Compression reset clears the dedup entry for the session - Stale entries evicted (older than 2x cooldown) to prevent memory leak Does NOT inject into messages — purely user-facing via _safe_print (CLI) and status_callback (gateway). Zero prompt cache impact. Fixes NousResearch#6309. Fixes NousResearch#5963.
gweeteve
pushed a commit
to gweeteve/hermes-agent
that referenced
this pull request
Jun 2, 2026
…earch#6411) Combines the approaches from PR NousResearch#6309 (duan78) and PR NousResearch#5963 (KUSH42): Tiered warnings (from NousResearch#5963): - Replaces boolean _context_pressure_warned with float _context_pressure_warned_at - Fires at 85% (orange) and re-fires at 95% (red/critical) - Adds 'compacting context...' status message before compression Gateway dedup (from NousResearch#6309): - Class-level dict _context_pressure_last_warned survives across AIAgent instances (gateway creates a new instance per message) - 5-minute cooldown per session prevents warning spam - Higher-tier warnings bypass the cooldown (85% → 95% always fires) - Compression reset clears the dedup entry for the session - Stale entries evicted (older than 2x cooldown) to prevent memory leak Does NOT inject into messages — purely user-facing via _safe_print (CLI) and status_callback (gateway). Zero prompt cache impact. Fixes NousResearch#6309. Fixes NousResearch#5963.
Egavasyug
pushed a commit
to Egavasyug/hermes-agent
that referenced
this pull request
Jun 10, 2026
…earch#6411) Combines the approaches from PR NousResearch#6309 (duan78) and PR NousResearch#5963 (KUSH42): Tiered warnings (from NousResearch#5963): - Replaces boolean _context_pressure_warned with float _context_pressure_warned_at - Fires at 85% (orange) and re-fires at 95% (red/critical) - Adds 'compacting context...' status message before compression Gateway dedup (from NousResearch#6309): - Class-level dict _context_pressure_last_warned survives across AIAgent instances (gateway creates a new instance per message) - 5-minute cooldown per session prevents warning spam - Higher-tier warnings bypass the cooldown (85% → 95% always fires) - Compression reset clears the dedup entry for the session - Stale entries evicted (older than 2x cooldown) to prevent memory leak Does NOT inject into messages — purely user-facing via _safe_print (CLI) and status_callback (gateway). Zero prompt cache impact. Fixes NousResearch#6309. Fixes NousResearch#5963.
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
1 participant
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
In the Hermes gateway (Telegram, Discord, Slack, etc.), every incoming user message creates a new
AIAgentinstance (seegateway/run.py~line 4569 inrun_sync()). The_context_pressure_warnedflag is designed to prevent duplicate "Context compaction approaching" warnings within a single agent loop, but since it's initialized toFalsein__init__, it resets on every message.Result: If a session's context is already above 85% of the compaction threshold, the warning fires on every single message the user sends. This can spam users with 3+ notifications per minute.
Fix
Added class-level time-based deduplication on
AIAgent(persists across instances):_context_pressure_last_warned: dict[str, float]tracks the last warning timestamp persession_id_CONTEXT_PRESSURE_COOLDOWN = 300): same session won't receive the warning more than once per 5 minutes, even across newAIAgentinstancesChanges
run_agent.py: ~12 lines of logic (class var + cooldown check in emission + reset in compression)tests/run_agent/test_context_pressure.py: 3 new test cases:test_second_instance_within_cooldown_is_suppressed— verifies dedup works across instancestest_warning_fires_after_cooldown_expires— verifies warning returns after 5 mintest_compression_resets_dedup— verifies compression clears the cooldownValidation
run_agent.pyRelated
_context_pressure_warneddesign noted inreset_session_state()not clearing the flag (intentional, per comment at line 6015-6018)