mirrors/hermes-agent
1
0
Fork 0
mirror of https://github.com/NousResearch/hermes-agent.git synced 2026-06-23 10:42:00 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions 7

fix: improve context compression quality — named constants, tool tracking, degradation warning

Browse source 
Three targeted improvements to the compression system:

1. Replace hardcoded truncation limits with named class constants
   (_CONTENT_MAX=6000, _CONTENT_HEAD=4000, _CONTENT_TAIL=1500,
   _TOOL_ARGS_MAX=1500, _TOOL_ARGS_HEAD=1200). Previous limits
   (3000/500) heavily truncated the summarizer's input — a 200-line
   edit got cut to 3000 chars before the summarizer ever saw it.

2. Add '## Tools & Patterns' section to both compression prompt
   templates (first-pass and iterative). Preserves working tool
   invocations, preferred flags, and tool-specific discoveries
   across compaction boundaries.

3. Warn users on 2nd+ compression: 'Session compressed N times —
   accuracy may degrade. Consider /new to start fresh.'

Ref #499
This commit is contained in:
SHL0MS 2026-03-30 23:34:58 -04:00 • committed by Teknium
parent af4abd2f22
commit 8567031433
2 changed files with 36 additions and 12 deletions
Download patch file Download diff file Expand all files Collapse all files

9
run_agent.py
View file

@ -6013,6 +6013,15 @@ class AIAgent:
except Exception as e:
logger.warning("Session DB compression split failed — new session will NOT be indexed: %s", e)
# Warn on repeated compressions (quality degrades with each pass)
_cc = self.context_compressor.compression_count
if _cc >= 2:
self._vprint(
f"{self.log_prefix}⚠️ Session compressed {_cc} times — "
f"accuracy may degrade. Consider /new to start fresh.",
force=True,
)
# Update token estimate after compaction so pressure calculations
# use the post-compression count, not the stale pre-compression one.
_compressed_est = (