Xubin Ren 1c71489121 fix(agent): count all message fields in token estimation ...

estimate_prompt_tokens() only counted the `content` text field, completely
missing tool_calls JSON (~72% of actual payload), reasoning_content,
tool_call_id, name, and per-message framing overhead. This caused the
memory consolidator to never trigger for tool-heavy sessions (e.g. cron
jobs), leading to context window overflow errors from the LLM provider.

Also adds reasoning_content counting and proper per-message overhead to
estimate_message_tokens() for consistent boundary detection.

Made-with: Cursor

2026-03-22 12:19:44 +08:00

..

__init__.py

refactor: unify instance path resolution and preserve workspace override

2026-03-08 02:58:25 +00:00

evaluator.py

refactor: replace <SILENT_OK> with structured post-run evaluation

2026-03-14 17:41:08 +08:00

helpers.py

fix(agent): count all message fields in token estimation

2026-03-22 12:19:44 +08:00