#106106th of 149 in Skills, plugins and rules
sol-pi
Token-saving extension for the Pi coding agent with four opt-in mechanisms
Documentation ↗
(opens in a new tab)Website ↗
(opens in a new tab)Repository on GitHub ↗
(opens in a new tab)
- Stars
- 3.5k
- License
- MIT
- Last commit
- Oct 2026
- Language
- TypeScript
Missed the latest GitHub check; these facts may be older.
Overview
SoL-Pi is an NVIDIA extension for the Pi coding agent that cuts token use and agent turns. It adds Action Fusion (edit plus validation in one tool call), ObservationPack (handles for repeated large outputs), an evidence-checked log reducer, and context compaction at completed plan steps. It installs on unmodified Pi 0.85.1 and every mechanism is off until enabled in sol-pi.json.
Who it is for: Pi coding agent users who want lower token cost on long agent runs
Strengths
- Installs through Pi's public extension APIs; no Pi patches or vendored source
- All mechanisms disabled by default; config is a single sol-pi.json
- Reducer keeps a quotation only if it matches the archived source log
- Includes a skill example for calling it from Claude Code or Codex
Weaknesses
- Tested only against Pi 0.85.1; pi-v1 support still in development
- Archives persist locally and are not deleted when the session ends
- Reducer sends diagnostic log content to a remote model when enabled
- Requires Node.js 22.19 or newer; no Docker image
What it needs
- no GPU
- Needs Node.js 22.19+, npm, @earendil-works/pi-coding-agent 0.85.1
- Models: Uses the model and provider configured in Pi, Reducer provider/model route set in sol-pi.json
Also in Skills, plugins and rules
See all 149| Rank | Project | Score |
|---|---|---|
| 1 | cavemanCuts token use in coding agents with terse replies and compressed inputs | 95 out of 100 |
| 2 | skillsAgent skills for planning, TDD, debugging and code review | 94 out of 100 |
| 3 | eccAgents, skills, hooks and rules bundle for Claude Code, Codex and other harnesses | 94 out of 100 |
| #4 | impeccableDesign skill with 24 commands and a detector for AI-generated frontends | 94 out of 100 |
| #5 | ponytailSkill prompt that makes coding agents write less code and add tests | 92 out of 100 |