last30days

last30days for pi

A pi-native package bridge for Matt Van Horn's excellent mvanhorn/last30days-skill.

This is not Matt's original repo and it should not be confused with it. Matt built the underlying last30days research engine and Agent Skill. This repo packages that engine for the pi coding agent ecosystem with native pi tools, commands, docs, and release hygiene.

If you want the canonical upstream project, use Matt's repo:

👉 mvanhorn/last30days-skill

If you use pi and want /last30days as a first-class pi package, this repo is for you.

---

What this adds for pi users

This repo keeps the upstream Python research engine intact, then adds a pi-focused surface:

• last30days_research pi tool
• last30days_diagnose pi tool
• /last30days command
• /last30days-doctor command
• /last30days-config command
• /last30days-open command
• concise pi-native skill at /skill:last30days
• pi package metadata so pi install works directly
• pi packaging verification with npm run verify:pi

The main improvement: pi models do not need to parse the long Claude-specific skill contract from upstream. They get a small pi wrapper that calls the engine through typed tools.

---

Install

pi install https://github.com/gvkhosla/last30days-pi

Then inside pi:

/reload
/last30days OpenAI --quick

Local checkout install:

git clone https://github.com/gvkhosla/last30days-pi
pi install /absolute/path/to/last30days-pi

---

Usage in pi

/last30days OpenAI --quick
/last30days "Claude Code vs OpenClaw" --deep --days 14
/last30days Cursor IDE --emit html
/last30days OpenAI --competitors 2
/last30days-doctor
/last30days-config
/last30days-open
/skill:last30days cursor vs windsurf

Supported pi command flags:

• --quick, --balanced, --deep
• --depth quick|balanced|deep
• --days N / --days=N
• --lookback N / --lookback=N
• --emit compact|json|md|html
• --search reddit,youtube,github
• --web-backend auto|brave|exa|serper|parallel|none
• --competitors / --competitors N
• --no-auto-resolve

Reports are saved to LAST30DAYS_MEMORY_DIR, which defaults to ~/Documents/Last30Days/.

---

Why pi users may prefer this package

Native tools instead of prompt gymnastics

The upstream skill is optimized for Agent Skills hosts, especially Claude Code. This package gives pi native tool definitions:

• structured arguments
• tool output truncation that respects pi's tool-output budget
• details metadata with report paths/config
• command UX that works directly in pi

Setup diagnostics built in

Run:

/last30days-doctor

It shows which sources are currently active and what to unlock next.

Config opens from pi

Run:

/last30days-config

This opens:

~/.config/last30days/.env

Saved reports are easy to find

Run:

/last30days-open

This opens the local report directory.

Python fallback for pi installs

The extension first looks for Python 3.12+ (python3.14, python3.13, python3.12, python3). If none is available, it falls back to:

uv run --project <repo> python

---

Source unlocks

Zero-config sources usually work immediately:

• Reddit
• Hacker News
• Polymarket
• GitHub

Optional unlocks:

# X / Twitter
FROM_BROWSER=auto
# or
XAI_API_KEY=...
# or
AUTH_TOKEN=...
CT0=...

# TikTok, Instagram, Threads, Pinterest, comments
SCRAPECREATORS_API_KEY=...
INCLUDE_SOURCES=tiktok,instagram,threads

# Web backends
BRAVE_API_KEY=...
EXA_API_KEY=...
SERPER_API_KEY=...
PARALLEL_API_KEY=...

# Perplexity / Deep Research through OpenRouter
OPENROUTER_API_KEY=...

# Bluesky
BSKY_HANDLE=you.bsky.social
BSKY_APP_PASSWORD=...

Digg AI-1000 clusters unlock when digg-pp-cli is on your PATH.

---

Developer checks

npm run verify:pi
pi -e ./extensions/index.ts --skill ./pi-skills/last30days --mode json '/last30days-doctor'
uv run pytest -q

---

Relationship to upstream

This repo is a pi-focused packaging fork/bridge.

• Original engine and product: mvanhorn/last30days-skill
• Pi-focused package repo: gvkhosla/last30days-pi

Credit for the research engine, source integrations, scoring pipeline, and original skill belongs to Matt Van Horn and contributors to the upstream repo.

This repo's contribution is the pi ecosystem layer: installability, typed tools, slash commands, concise pi skill wrapper, docs, and package verification.

License

MIT, following the upstream project.