@mghodsizadeh/pi-recall0.1.0
Pull context from your earlier pi conversations into the current one.
Untested on omp
This is a pi package. No one has reported whether it works on omp yet. If you try it, add your result below.
Reports 0
No reports yet.
Readme
pi-recall
Pick up where you left off. Pull the useful parts of earlier pi conversations into your current one, without pasting raw transcripts.
/recall auth refactor
Why
Sometimes you start a new session: the cache expired, the old one got too long, or you're picking up a task from last week. The decisions you made are still in the old session. pi-recall finds that session and brings in only what matters.
Install
pi install npm:@mghodsizadeh/pi-recall
# or
pi install git:github.com/ghodsizadeh/pi-recall
Use it
/recall [query] opens a picker with the sessions from this project.
| Key | Action |
|---|---|
| type | filter live |
↑ ↓ |
move |
Tab |
select (you can pick several) |
Enter |
attach |
Esc |
cancel |
Then choose how to attach the sessions:
- Condensed: your messages, the assistant's replies and the names of the tools it called. Tool output and thinking are dropped. This step doesn't call a model, so it's instant and the wording is exact.
- Digest: one model call turns the selected sessions into a short summary of decisions, current state, files and open items. You can edit the summary before it's attached.
Or just ask. The agent has a recall tool, so a prompt like this works:
check how we handled rate limiting in an earlier session and do the same here
What counts as "this project"
- the git repo you're in, including its subdirectories
- its git worktrees, even ones that live outside the repo folder (
my-app-feature-x/)
Good to know
- Search runs over the condensed text, so words that only appeared in tool output won't match.
- Each attached session is capped at about 15k tokens. If a session is longer, the middle is cut, and the agent can still read that part with the tool.
- Everything runs locally. Your sessions are only read, never modified.
License
MIT