_ _ __ _(_)_ __ __ _(_) \ \ /\ / / | '_ \ _____ / _` | | \ V V /| | |_) |_____| (_| | | \_/\_/ |_| .__/ \__,_|_| |_|

What was I doing?

One command restores your context across all your git repos: recent activity, uncommitted work, unpushed commits. Plus a welcome-back brief from your own local LLM.

zero dependencies node ≥18 local-first · ollama license: MIT

git clone https://github.com/L-ubu/wip-ai.git && cd wip-ai && npm link

# the monday-morning command

20+ repos. Back from lunch, a meeting marathon, or a long weekend. wip is the save-state loader for your brain.

$ wip ▟ wip · what was I doing? (19 repos scanned) REPO ACTIVE BRANCH STATE ● wip-ai just now main ✎4 "feat: ollama brief" ● Terminup 2d ago master ✎7 ↑2 "tup: slime codex" ● jlr_mss 3d ago feat/vat ✎2 "fix: vat nullable" ● flipper-portals 2w ago main "eu portals batch 2" ● old-experiment 4mo ago main "wip: broken" … and 14 more (wip --all) ▸ pick up where you left off: wip-ai (~/Projects/wip-ai) uncommitted: bin/wip-ai.mjs, src/brief.mjs, README.md last commit: "feat: ollama brief" on main 🤖 welcome-back brief (ollama qwen2.5:7b) You were mid-way through the wip-ai Ollama integration. Pick up in src/brief.mjs, the prompt still needs tuning. Commit the working scanner before context-switching.

# features

Small tool, sharp edges.

multi-repo scan

Scans every git repo under your roots in parallel. 70 repos in about a second.

local-llm brief

Ollama writes a 3-line welcome-back: what you were doing, where to pick up, one next step. No cloud, no keys, no cost.

zero dependencies

Plain Node ≥18 + git. Nothing to audit, nothing to break, nothing to update.

state at a glance

Uncommitted files ✎, unpushed commits ↑, stashes ⚑, stale repos in red. The truth about your WIP, finally visible.

--json output

Machine-readable everything. Pipe it into your own scripts, dashboards, or morning routine.

fails soft

No Ollama running? Everything except the brief works fine. It just tells you and moves on.

# commands

One binary, two names: wip and wip-ai.

command	what it does
`wip`	scan repos, show recent activity + AI brief
`wip --all`	show every repo, not just the top 10
`wip --no-ai`	skip the Ollama brief (faster)
`wip --json`	machine-readable output
`wip --days 7`	only repos active in the last n days
`wip --model llama3.2`	pick a different Ollama model
`wip --path ~/work`	one-off extra directory to scan

# config

Optional. Lives at ~/.config/wip-ai/config.json.

{
  "roots": ["~/Projects", "~/sites"], // folders of repos (default: auto-detected)
  "extraRepos": ["~/dotfiles"],    // repos outside your roots
  "ignore": ["node_modules"],      // directory names to skip
  "model": "qwen2.5:7b",           // Ollama model for the brief
  "ai": true,                      // master switch for the brief
  "max": 10                        // rows shown by default
}

# why

Three honest reasons.

Context switching is expensive

Especially with ADD, especially with many side projects. Reloading "where was I" costs real time and energy. wip makes it one command.

Local-first means private

Your repo names and commit messages never leave your machine. The brief runs on your own Ollama model, offline if you want.

Uncommitted work is invisible until it hurts

Dirty repos, unpushed branches, forgotten stashes: wip surfaces all of it before you lose an evening of work to a clean install.