Feature request: Incremental generation for large archives

[ludaesch]

Problem

When using claude-code-transcripts all with a large archive (17+ projects, 50+ sessions), regeneration takes a while even when only 1-2 sessions have changed. Every session's JSONL is re-parsed and every HTML file is regenerated.

Proposed Solution

Add incremental generation that skips sessions whose transcripts are already up-to-date (output HTML newer than source JSONL).

Suggested Implementation

1. Incremental mode (default for all): Before calling generate_html(), check if the output directory exists and its newest HTML file is newer than the source JSONL. If so, skip.

2. --force flag: Override incremental mode and regenerate everything (current behavior).

3. project command (optional): Regenerate transcripts for a single project directory, useful for quick updates:

   claude-code-transcripts project ~/.claude/projects/-Users-me-myproject -o ./archive

4. index command (optional): Regenerate only the index files without re-parsing sessions:

   claude-code-transcripts index -o ./archive

Use Case

I maintain a private transcript archive (all projects) and a public one (curated subset). Running update-transcripts daily means regenerating ~50 sessions when typically only 2-3 have changed. Incremental mode would reduce this significantly.

Implementation Notes

I've sketched out a patch (~100 lines) that adds:

• _is_session_up_to_date() helper function
• incremental=True parameter to generate_batch_html()
• --force flag to all command
• Optional project and index commands

Happy to submit a PR if this approach sounds reasonable. Just wanted to check if you'd welcome this feature and if the design makes sense before doing the work.

[nmpowell]

+1
I wanted a way to idempotently re-run all without regenerating everything, and this idea seems close.
This PR also introduces (from what I can see) incremental html, so some overlap conceptually.