Native macOS app

AiSystemmacOS application

A native macOS app that brings skills and documents together, highlights differences across projects, and controls distribution to Claude and Codex.

View on GitHub
Illustrated preview of AiSystem: overview and resource synchronization for Claude and Codex.

Overview

One source of truth for AI development resources

The problemWhen projects maintain separate rules and AI skills, copies drift and changes become difficult to track across Claude and Codex.

The solutionAiSystem centralizes reference sources, detects differences, and lets users choose, preview, and distribute versions without silently overwriting local edits.

User workflow

Shared reference library

Canonical files are organized and linked to their target projects.

1 / 4: Collect shared resources. Previous step / Next step

Technical architecture

SwiftUI invokes Bash routes through Process; Python modules use YAML sources to handle checks, versions, and project copies.

macOS app

SwiftUI

Navigation, library and operation feedback
Typed commands

CommandRunner

Swift / Process

Runs local actions using separate arguments
Process + argv

Action router

Bash

Dispatches backend actions
Local routes

Domain services

Python

Inventory, versions, distribution and checks
  • macOS app: SwiftUI. Navigation, library and operation feedback Typed commands.
  • CommandRunner: Swift / Process. Runs local actions using separate arguments Process + argv.
  • Action router: Bash. Dispatches backend actions Local routes.
  • Domain services: Python. Inventory, versions, distribution and checks

Technical challenges

01

Spot drift without competing sources

Constraint

The same resources are copied across projects and Claude/Codex formats.

The solution

A registry and manifest define canonical sources; Inventory and Doctor classify missing, drifting, or conflicting copies.

02

Prevent accidental overwrites during distribution

Constraint

Synchronization can meet local edits or an outdated operation plan.

The solution

Preview, plan IDs, explicit blockers, per-project atomic writes, and conditional undo.

03

Keep business state separate from process logs

Constraint

The app needs clear operation feedback while business actions run as local command-line processes.

The solution

Versioned JSON contracts, Swift decoding, cached reads, and a session-scoped operation history.

Main technical stack

SwiftSwiftUIPythonBashXcodeJSON

Outcomes and learnings

Native macOS control interface

Overview, Projects, Library and Activity destinations, a command palette, and contextual operation feedback.

Reusable resource library

Shared resource management, version selection, project targeting, imports and redistribution.

Checks and recovery workflows

Inventory, diagnostics, change previews, guarded writes, and conditional undo support.

Previous projectSmartFlow
Next project

Aimoto

A local lab to experiment with, backtest and compare daily Bitcoin forecasts.