Timmy_Foundation/hermes-agent
15
0
Fork 0
Code Issues 248 Pull Requests 14 Actions 1 Packages Projects Releases Wiki Activity

fix: detect and warn on file re-read loops after context compression

Browse Source 
When context compression summarizes conversation history, the agent
loses track of which files it already read and re-reads them in a loop.
Users report the agent reading the same files endlessly without writing.

Root cause: context compression is lossy — file contents and read history
are lost in the summary. After compression, the model thinks it hasn't
examined the files yet and reads them again.

Fix (two-part):
1. Track file reads per task in file_tools.py. When the same file region
   is read again, include a _warning in the response telling the model
   to stop re-reading and use existing information.
2. After context compression, inject a structured message listing all
   files already read in the session with explicit "do NOT re-read"
   instruction, preserving read history across compression boundaries.

Adds 16 tests covering warning detection, task isolation, summary
accuracy, tracker cleanup, and compression history injection.
This commit is contained in:
0xbyt4
2026-03-08 20:44:42 +03:00
parent cd77c7100c
commit 9eee529a7f
3 changed files with 349 additions and 6 deletions
Download Patch File Download Diff File Expand all files Collapse all files

33
run_agent.py
View File

@@ -2463,7 +2463,7 @@ class AIAgent:
if messages and messages[-1].get("_flush_sentinel") == _sentinel:
messages.pop()
def _compress_context(self, messages: list, system_message: str, *, approx_tokens: int = None) -> tuple:
def _compress_context(self, messages: list, system_message: str, *, approx_tokens: int = None, task_id: str = "default") -> tuple:
"""Compress conversation context and split the session in SQLite.
Returns:
@@ -2478,6 +2478,25 @@ class AIAgent:
if todo_snapshot:
compressed.append({"role": "user", "content": todo_snapshot})
# Preserve file-read history so the model doesn't re-read files
# it already examined before compression.
try:
from tools.file_tools import get_read_files_summary
read_files = get_read_files_summary(task_id)
if read_files:
file_list = "\n".join(
f" - {f['path']} ({', '.join(f['regions'])})"
for f in read_files
)
compressed.append({"role": "user", "content": (
"[Files already read in this session — do NOT re-read these]\n"
f"{file_list}\n"
"Use the information from the context summary above. "
"Proceed with writing, editing, or responding."
)})
except Exception:
pass # Don't break compression if file tracking fails
self._invalidate_system_prompt()
new_system_prompt = self._build_system_prompt(system_message)
self._cached_system_prompt = new_system_prompt
@@ -2999,7 +3018,8 @@ class AIAgent:
for _pass in range(3):
_orig_len = len(messages)
messages, active_system_prompt = self._compress_context(
messages, system_message, approx_tokens=_preflight_tokens
messages, system_message, approx_tokens=_preflight_tokens,
task_id=effective_task_id,
)
if len(messages) >= _orig_len:
break # Cannot compress further
@@ -3461,7 +3481,8 @@ class AIAgent:
original_len = len(messages)
messages, active_system_prompt = self._compress_context(
messages, system_message, approx_tokens=approx_tokens
messages, system_message, approx_tokens=approx_tokens,
task_id=effective_task_id,
)
if len(messages) < original_len:
@@ -3528,7 +3549,8 @@ class AIAgent:
original_len = len(messages)
messages, active_system_prompt = self._compress_context(
messages, system_message, approx_tokens=approx_tokens
messages, system_message, approx_tokens=approx_tokens,
task_id=effective_task_id,
)
if len(messages) < original_len or new_ctx and new_ctx < old_ctx:
@@ -3848,7 +3870,8 @@ class AIAgent:
if self.compression_enabled and self.context_compressor.should_compress():
messages, active_system_prompt = self._compress_context(
messages, system_message,
approx_tokens=self.context_compressor.last_prompt_tokens
approx_tokens=self.context_compressor.last_prompt_tokens,
task_id=effective_task_id,
)
# Save session log incrementally (so progress is visible even if interrupted)