se-session-extract

You can extract a conversation skeleton from large session JSONL files by running the skeleton extraction script, which generates a concise narrative digest of user and assistant turns with collapsed tool-call summaries for lightweight analysis.

To identify tool errors in session data, run the error extraction mode to surface tool errors and failures directly from the JSONL file, outputting a focused error report alongside a meta summary of error counts for quick diagnosis.

The best way to reduce JSONL session file size for AI context windows is to extract lightweight narrative skeletons or error signals, producing compact digests that feed AI agents without exposing the full transcript.

Yes, you can extract both error signals and conversation summaries from the same JSONL file by running the skeleton and error extraction modes independently, each producing a distinct digest with a meta summary of relevant counts.

No, this session extraction tool requires no external dependencies to process JSONL data, relying entirely on deterministic extraction logic built into its Python scripts to generate skeleton digests and error reports.