lolipopshock / pi-clm

# pi-clm
 [code] 
`pi-clm` is a Pi extension 
that lets the agent manage its own context: the language model can generate functions to 
modify the context (mapped as a file), and the edited version becomes its next input. 
Check our paper Context Language Models and research
codebase. 
## Install
 [code] 
## Quick start
| command | what it does |
|---------|--------------|
| `/clm` | open the panel: **overview** (context size over time + compaction points) **input** (what the next request contains) · **edits** (per-revision side-by-side diff) |
| `/clm status` | three lines: next request vs budget, edits, changed settings |
| `/clm config` | open **settings**; `/clm config ` changes one, `/clm config reset` drops this session's changes |
| `/clm-compact [instructions]` | ask the model to compact its own context now; anything you add (e.g. what to keep) is passed along. The result shows on the **edits** page |
| `/clm on` / `off` / `reset` | enable, use raw context, discard the projection |
In the panel: `1–4` or `Tab` switch pages, `← →` step through compaction points, `z`
zooms the chart, `Enter` opens the selection, `q` closes.
## Docs
- How it works — the mirror, what runs without the model, the panel, safety.
- Configuration — budget, reserve, reminders, guard, steering and the other settings.
- Architecture — the system as implemented, design notes, known limitations.
- Development — setup, checks, releasing.
## Citation
If you use pi-clm in your research, please cite
Context Language Models:
 [code]