Skip to content

emrg: fix double-accumulated reasoning in llm.jsonl (O(n²) blowup)#943

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
argszero merged 2 commits into from
Aug 23, 2026
Merged

emrg: fix double-accumulated reasoning in llm.jsonl (O(n²) blowup) #943

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
083d09c
emrg: fix double-accumulated reasoning in llm.jsonl (O(n²) blowup, ra…
Aug 23, 2026
835ebb0
emrg: merge master into reasoning-fix branch (resolve #942 conflict)
Aug 23, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
8 changes: 7 additions & 1 deletion emrg/server/daemon.py
View file
Open in desktop
Original file line numberDiff line numberDiff line change
Expand Up@@ -2425,7 +2425,13 @@ async def _run_tool_loop(
return

full_content = "".join(content_parts)
full_reasoning = "".join(reasoning_parts) or None
# llm.py yields the ACCUMULATED reasoning snapshot on every chunk
# (contract: chunk["reasoning"] = full think text so far, see
# llm.py:366). Appending every snapshot and joining would
# double-accumulate → O(n²) blowup (29640 chars for 47 real tokens,
# up to 10MB records in llm.jsonl, rant 2026-08-23T10:15:06). Take the
# LAST snapshot = complete think text.
full_reasoning = reasoning_parts[-1] if reasoning_parts else None
logger.debug("round %d finish: %s, tool_calls=%d, content_len=%d%s",
round_num, final_finish, len(tc_by_index), len(full_content),
self._format_cache_pct(final_usage))
Expand Down
Loading