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: remove synthetic error message injection, fix session resume after repeated failures

Browse source 
Two changes to the error handler in the agent loop:

1. Remove the 'if not pending_handled' block that injected fake
   [System error during processing: ...] messages into conversation
   history.  These polluted history, burned tokens on retries, and
   could violate role alternation by injecting as role=user.
   The tool_calls error-result path (role=tool) is preserved.

2. Append the error final_response as an assistant message when
   hitting the iteration limit, so session resume doesn't produce
   consecutive user messages.
This commit is contained in:
Teknium 2026-03-21 06:33:05 -07:00
parent 10d719ac1b
commit 779619f742
No known key found for this signature in database
1 changed files with 9 additions and 11 deletions
Download patch file Download diff file Expand all files Collapse all files

20
run_agent.py
View file

@ -6807,20 +6807,18 @@ class AIAgent:
pending_handled = True
break
if not pending_handled:
# Error happened before tool processing (e.g. response parsing).
# Choose role to avoid consecutive same-role messages.
last_role = messages[-1].get("role") if messages else None
err_role = "assistant" if last_role == "user" else "user"
sys_err_msg = {
"role": err_role,
"content": f"[System error during processing: {error_msg}]",
}
messages.append(sys_err_msg)
# Non-tool errors don't need a synthetic message injected.
# The error is already printed to the user (line above), and
# the retry loop continues. Injecting a fake user/assistant
# message pollutes history, burns tokens, and risks violating
# role-alternation invariants.
# If we're near the limit, break to avoid infinite loops
if api_call_count >= self.max_iterations - 1:
final_response = f"I apologize, but I encountered repeated errors: {error_msg}"
# Append as assistant so the history stays valid for
# session resume (avoids consecutive user messages).
messages.append({"role": "assistant", "content": final_response})
break
if final_response is None and (