Forkit AI Footprints 发布：本地监测 AI 工具活跃时长与本地模型

Forkit AI Footprints shows when supported AI tools are active on a device, how long they stay active, and which local AI models are available. Monitoring is local by default. Joining the community Preview is optional.

Product · npm · Releases · Support · Contributing · Security

Platform support: the current release supports Apple Silicon macOS. Other platforms are not yet supported.

Install

npx --yes forkit-ai-footprints@latest

The bootstrap installs the persistent app in the current user's Applications folder and opens it. Later launches use Applications, Spotlight, the Dock, or the Forkit menu-bar icon. Run the same command again to update.

The bootstrap requires Node.js 20, 22, or 24. A signed direct download and Homebrew formula are not currently available. See the distribution policy.

What it shows

AI activity timeline — working, ready, paused, and excluded intervals for supported AI applications.
Observation summary — valid observed time, AI-active time, longest active block, and supported tools observed.
Per-app evidence — measured runtime plus current and bounded local aggregate CPU and memory summaries.
Local AI footprint — recognized local model records, logical storage, and supported runtime loaded-state evidence.
Share image — a local 1080×1080 summary containing measured totals and supported tool names only.
Community Preview — an optional, consented comparison using an allowlisted anonymous aggregate.

Forkit does not infer prompts, tasks, tokens, cost, energy, or model ownership from process activity. GPU and per-app storage attribution are not measured.

Measurement boundary

Signal	What it means	Important limit
AI-active	Sustained recent CPU-time change in a supported AI-app process tree	Not prompt, task, or token attribution
Ready	A supported app is present without sustained measured activity	Presence alone is never counted as active
Valid observed time	Time while the device is awake, unlocked, and eligible for observation	Locked, stopped, sleep-gap, and excluded time do not count
CPU and memory	Current process-tree readings and bounded local aggregate summaries	Not energy, cost, or exclusive workload ownership
Local model storage	Logical bytes for recognized files inside supported roots	Not whole-disk or physical APFS usage
Loaded local model	Provider-confirmed loaded state from a supported local runtime	Availability alone does not mean a model is running

Detector rules are conservative and reviewable. A curated conformance corpus is not a representative field-accuracy study. See the measurement foundation and validation record.

Privacy

The app is metadata-only and private by default.

It does not inspect prompts, responses, or model contents; collect credentials or account identity; or retain or upload raw commands, full paths, or the local device name. Runtime inspection is restricted to approved loopback endpoints.

The protected local app may display supported tool names and limited local context. That context is excluded from share images, saved reports, and the community payload. Read the complete privacy boundary before changing any collection, persistence, or network behavior.

Community Preview

Nothing is sent before the user accepts the separate v3 contribution notice. After consent and ten valid observed minutes, the app may send one signed, allowlisted aggregate and refresh that installation's latest row no more than once per hour when internet is available.

An active contributor is an opted-in installation that has reached the ten-minute gate and successfully synced. Downloads are reported separately and do not represent people, installations, or contributors.

The public AI-active rate is duration-weighted:

sum(AI-active seconds) / sum(valid observed seconds)

Personal percentile placement remains unavailable until there are at least 30 eligible contributors with one valid observed hour each. Community signatures prove payload integrity and continuity, not unique hardware or truthful measurement. The public result is therefore labelled Preview and community-reported.

See PRIVACY.md for the exact payload allowlist and exclusions.

Common commands

forkit-ai-footprints serve
forkit-ai-footprints monitor
forkit-ai-footprints scan
forkit-ai-footprints doctor
forkit-ai-footprints --version

Advanced local evaluation and field-testing commands are documented in the macOS tester guide. Truth files and evaluation results remain local unless a tester explicitly chooses to share an aggregate.

Development

Requirements: Apple Silicon macOS, Node.js 20/22/24, and npm.

git clone https://github.com/arpitasarker01/forkit-ai-footprints.git
cd forkit-ai-footprints
npm ci
npm test
npm run smoke:package

Detector changes require positive, negative, deduplication, and privacy tests. Changes to collection, persistence, network destinations, consent, or packaging require explicit product-boundary and maintainer review.

Read CONTRIBUTING.md before opening a pull request. For installation help or bug reports, see SUPPORT.md. Report security or privacy vulnerabilities privately as described in SECURITY.md.

Project documentation

License

Forkit AI Footprints is available under the MIT License.