Turns out Claude Code retention policy can not be trusted, it happened to me after an upgrade (I assume), random session and sub-agent transcripts were wiped which were newer than 30 days, while others, older than 30 days, are still in storage. Looking at the CC repo, there are so many reports around this topic: https://github.com/anthropics/claude-code/issues?q=is%3Aissue%20cleanupPeriodDays
I propose we implement in cctop an archival mechanism where we automatically backup sessions that are not active for more than 24h (configurable). We should extend the history UI to include stats from the archive.
We need to decide how are we going to implement the archival process, given that we want to query old sessions without having to perform expensive decompression operations. We could compile that stats and store them in a sqlite db for long-term storage, separate from the full session archive. I'm opened to suggestions.
⚠️ Please DO NOT open a PR for this, first we need to decide all the details and have a well defined spec for this feature.
Turns out Claude Code retention policy can not be trusted, it happened to me after an upgrade (I assume), random session and sub-agent transcripts were wiped which were newer than 30 days, while others, older than 30 days, are still in storage. Looking at the CC repo, there are so many reports around this topic: https://github.com/anthropics/claude-code/issues?q=is%3Aissue%20cleanupPeriodDays
I propose we implement in cctop an archival mechanism where we automatically backup sessions that are not active for more than 24h (configurable). We should extend the history UI to include stats from the archive.
We need to decide how are we going to implement the archival process, given that we want to query old sessions without having to perform expensive decompression operations. We could compile that stats and store them in a sqlite db for long-term storage, separate from the full session archive. I'm opened to suggestions.