ctx has reached major stable version 1.0. It's an open-source local CLI for searching your coding agent history.
1.0 brings several big changes:
- switched from sqlite to Tantivy, resulting in ~16x performance improvement
- introduced "lineage", so the CLI can understand how to trace deep subagents and forked sessions to where they came from
- introduced "ctx blame", which is like git blame but for agent sessions.
The performance changes are really significant: if you have a lot of agent history (I sure do) then a previously 40 minute ingestion now takes 2.5 minutes. And new activity is so cheap that it can finally run as a lite background process instead of on-demand.
Interestingly, I got the inspiration to try "source-backed indexing" here on HN from @malandin (of SereneDB) on our last post. So yes, HN is working as intended. (See: https://news.ycombinator.com/item?id=48763462#48778744 )
And I'm very excited about "ctx blame" which is what we decided to pursue for our paid add-on rather than a cloud service (it's fully local like ctx search). It works like git blame: you start with a line, commit, file, or PR and then it returns the exact agent session event and transcript that produced it.
It's really useful for digging up the previous context for why the code was implemented the way it was. Which tests were run at the time, what user conversation was had at the time, etc.
Happy to answer any questions or implement changes based on your feedback!
luca-ctx•51m ago
ctx has reached major stable version 1.0. It's an open-source local CLI for searching your coding agent history.
1.0 brings several big changes:
- switched from sqlite to Tantivy, resulting in ~16x performance improvement
- introduced "lineage", so the CLI can understand how to trace deep subagents and forked sessions to where they came from
- introduced "ctx blame", which is like git blame but for agent sessions.
The performance changes are really significant: if you have a lot of agent history (I sure do) then a previously 40 minute ingestion now takes 2.5 minutes. And new activity is so cheap that it can finally run as a lite background process instead of on-demand.
Interestingly, I got the inspiration to try "source-backed indexing" here on HN from @malandin (of SereneDB) on our last post. So yes, HN is working as intended. (See: https://news.ycombinator.com/item?id=48763462#48778744 )
And I'm very excited about "ctx blame" which is what we decided to pursue for our paid add-on rather than a cloud service (it's fully local like ctx search). It works like git blame: you start with a line, commit, file, or PR and then it returns the exact agent session event and transcript that produced it.
It's really useful for digging up the previous context for why the code was implemented the way it was. Which tests were run at the time, what user conversation was had at the time, etc.
Happy to answer any questions or implement changes based on your feedback!