ce-session-extract

To extract conversation skeletons from large JSONL session logs, you can run a parsing script that filters user and assistant turns while collapsing tool calls into a compact digest. This significantly reduces the context size needed for analysis.

Yes, you can surface error signals in Claude Code or Cursor session files by running an errors mode extraction. This process parses the JSONL data to isolate failed tool calls and error outputs into a succinct digest.

The best way to reduce context size when analyzing session logs is to extract a filtered digest of the conversation. By emitting only the narrative skeleton or error signals, you load megabytes less data into your analysis context.

Yes, you need Python to parse JSONL files for error extraction because the process relies on Python scripts to read the structured logs and emit the filtered digest with a trailing meta line.

Yes, session extraction works with Codex JSONL files. The parsing scripts apply to Claude Code, Codex, and Cursor file formats to generate a compact digest of either conversation skeletons or error signals.