diff --git a/README.md b/README.md index 34dbe48..746d329 100644 --- a/README.md +++ b/README.md @@ -1,478 +1,224 @@ # AgentForge -> Orchestrate AI coding agents from your Mac — schedule, monitor, and chain Claude Code tasks on a kanban board. +**A local-first macOS command center for running Claude Code and OpenAI Codex CLI as schedulable, observable tasks.** [![License: MIT](https://img.shields.io/badge/License-MIT-blue.svg)](LICENSE) -[![Platform: macOS](https://img.shields.io/badge/Platform-macOS%2012%2B-lightgrey?logo=apple)](https://github.com/releases) -[![Bun 1.3+](https://img.shields.io/badge/Bun-1.3%2B-black?logo=bun)](https://bun.sh) [![TypeScript](https://img.shields.io/badge/TypeScript-5-blue?logo=typescript)](https://www.typescriptlang.org) -[![Claude Code](https://img.shields.io/badge/Powered%20by-Claude%20Code-orange)](https://docs.anthropic.com/en/docs/claude-code) +[![Platform: macOS](https://img.shields.io/badge/Platform-macOS%2012%2B-lightgrey?logo=apple)](../../releases) +[![Bun 1.3+](https://img.shields.io/badge/Bun-1.3%2B-black?logo=bun)](https://bun.sh) +[![TypeScript](https://img.shields.io/badge/TypeScript-5-blue?logo=typescript)](https://www.typescriptlang.org) -**Website**: https://agentforge-landing-weld.vercel.app/ +AgentForge turns agent work into a visible local workflow: queue tasks, watch live runs, schedule recurring checks, route tasks from chat, and distill repeated patterns into reusable skills. -![AgentForge Screenshot](assets/screenshot.png) +**Website:** https://agentforge-landing-weld.vercel.app/ ---- +![AgentForge task board preview](assets/readme-preview.svg) -## Table of Contents +## Why AgentForge -- [Features](#features) -- [Skill Library](#skill-library) -- [Requirements](#requirements) -- [Installation](#installation) -- [Troubleshooting](#troubleshooting) -- [Quick Start](#quick-start) -- [Chat Channels](#chat-channels) -- [Multi-Agent Pipelines (DAG)](#multi-agent-pipelines-dag) -- [API Reference](#api-reference) -- [Background Service](#background-service) -- [Architecture](#architecture) -- [Contributing](#contributing) -- [License](#license) +- **Stop babysitting terminals.** Create tasks once, then monitor queue, running, and done states from a kanban board. +- **Run the right agent per job.** Use Claude Code or Codex CLI on each task, with a configurable default. +- **Make agent work repeatable.** Schedule immediate, delayed, one-time, or cron tasks, then reuse successful patterns through the Skill Library. +- **Keep execution local.** The backend listens on `127.0.0.1`, stores state in SQLite, and runs agent CLIs on your Mac. ---- - -## Features - -- **🧬 Compounding Skill Library** — AgentForge detects recurring patterns across your completed tasks and distills them — using Anthropic's [skill-creator](https://github.com/anthropics/skills/tree/main/skills/skill-creator) — into reusable Claude Code skills. Review, approve, and they're symlinked into both Claude Code and Codex. Your agents get better at *your* work over time. [Learn more →](#skill-library) -- **Agents that spawn agents** — Any running task can create sub-tasks, build DAG pipelines, and collect results via the bundled Claude Code skill -- **Kanban Task Board** — Visual Queue / Running / Done columns with live output streaming -- **DAG Pipelines** — Define task dependencies with automatic cascade execution and failure propagation -- **Flexible Scheduling** — Immediate, delayed, one-time datetime, and cron-based recurring tasks -- **Dual Agent Backends** — Run tasks with Claude Code CLI or OpenAI Codex CLI; choose per-task or set a default -- **Chat Control** — Create tasks and receive notifications from Telegram, Slack, Feishu/Lark, or WeChat -- **Persistent Storage** — SQLite-backed task history, run logs, and streaming output -- **Native macOS App** — Electron shell with one-click DMG install - ---- - -## Skill Library - -**Turn the work you repeat into skills your agents reuse — automatically.** The more you use AgentForge, the better it gets at *your* workflows. - -1. **Detect** — A background sweep (or the manual **Scan** button) analyzes your completed task runs and tracks recurring patterns in a cross-run ledger — both *recipes* (workflows you repeat) and *pitfalls* (failures you diagnosed and fixed). Counting is idempotent per run, so re-scanning never inflates it. -2. **Distill** — For a pattern worth codifying, AgentForge uses Anthropic's **[skill-creator](https://github.com/anthropics/skills/tree/main/skills/skill-creator)** skill to author a proper `SKILL.md` — and first judges whether it's even *worth* a skill, so you don't drown in noise. -3. **Approve** — Review the draft: the worthiness verdict, the contributing tasks, and a fully editable `SKILL.md`. Approve or dismiss — you stay in control. -4. **Deliver** — Approved skills are written once to a canonical store and symlinked into **both** `~/.claude/skills` (Claude Code) and `~/.agents/skills` (Codex). Both agents load them natively via progressive disclosure — zero prompt injection. -5. **Manage** — Enable/disable each skill (toggles the symlinks, keeps the file), view its `SKILL.md`, or delete it — all from the **Skills** tab. - -> The automatic daily sweep is **off by default** (it spends tokens). Flip it on in **Settings → Skill Library**, choose the agent and cadence, or just hit the manual **Scan** button whenever you like — that always works, threshold or not. - -Distilled skills are standard Claude Code `SKILL.md` files (canonical copy in `~/.agentforge/skills`, symlinked to each agent's native dir) — portable, inspectable, no lock-in. - ---- - -## Requirements - -- macOS 12.0+ (Apple Silicon or Intel) -- [Bun](https://bun.sh) 1.3+ -- [Claude Code CLI](https://docs.anthropic.com/en/docs/claude-code) installed and on `PATH` (default agent) -- [OpenAI Codex CLI](https://github.com/openai/codex) on `PATH` — optional, required only if using Codex as agent backend (`npm install -g @openai/codex`) - ---- - -## Installation +## Quick Start -### Option 1: Desktop App (Recommended) +### Install The App -1. Download the latest `AgentForge-*.dmg` from the [Releases](../../releases) page. -2. Open the DMG and drag **AgentForge** into `/Applications`. -3. Launch from Launchpad or the Applications folder. +1. Download the latest `AgentForge-*.dmg` from [Releases](../../releases). +2. Drag **AgentForge** into `/Applications`. +3. Launch it, choose a working directory, and create your first task. -### Option 2: Build from Source +### Run From Source ```bash -git clone https://github.com/your-org/agentforge.git -cd agentforge +git clone https://github.com/hetaoBackend/agentforge.git +cd agentforge/taskboard-electron -# Install all dependencies -make install-deps - -# Build and package a DMG -make package-dmg -# Output: taskboard-electron/out/make/AgentForge-1.0.0-arm64.dmg +bun install --ignore-scripts +bun run start ``` -### Option 3: Development Mode +The Electron dev command builds the renderer, starts the Bun backend, and reloads the app while you work. -```bash -git clone https://github.com/your-org/agentforge.git -cd agentforge +### Build A DMG -make install-deps # bun install in backend/ and taskboard-electron/ - -# Start Electron + backend (Bun builds, watches, and spawns the backend) -cd taskboard-electron && bun run start +```bash +cd taskboard-electron +bun run make ``` ---- - -## Troubleshooting +The packaged app bundles the compiled Bun backend at `taskboard-electron/out/make/AgentForge-1.0.0-arm64.dmg`. -### `bun install` hangs or freezes +## What You Can Do -This is the most common setup issue. The Electron binary (~100 MB) is downloaded from GitHub and may stall on slow connections or in China. +| Area | What it gives you | +| --- | --- | +| **Tasks** | Kanban queue for pending, scheduled, blocked, running, completed, failed, and cancelled work. | +| **Scheduling** | Immediate, delayed, `scheduled_at`, and cron schedules with max-run limits. | +| **Live output** | Structured stream output from Claude Code or Codex CLI, persisted per run. | +| **Heartbeats** | Recurring checks that can decide whether to trigger, resume, or notify. | +| **Skill Library** | Pattern detection across completed runs, editable `SKILL.md` drafts, and native skill installation. | +| **Chat channels** | Telegram, Slack, Feishu/Lark, and WeChat adapters for creating or tracking tasks from chat. | +| **DAG pipelines** | Task dependencies, failure propagation, and upstream result injection for multi-agent workflows. | -**Quick fix — use mirrors:** +## Requirements -```bash -export ELECTRON_MIRROR=https://npmmirror.com/mirrors/electron/ -cd taskboard-electron && bun install --registry https://registry.npmmirror.com -``` +- macOS 12+ +- [Bun](https://bun.sh) 1.3+ +- [Claude Code CLI](https://docs.anthropic.com/en/docs/claude-code) or [OpenAI Codex CLI](https://github.com/openai/codex) on `PATH` -**Full guide:** [docs/installation-troubleshooting.md](docs/installation-troubleshooting.md) covers: -- How to diagnose whether the hang is network or native compilation -- Mirror setup for China and slow connections -- Installing build tools (macOS / Linux / Windows) -- Node.js version requirements -- Full clean-install procedure +## Skill Library ---- +The Skill Library turns repeated work into reusable agent skills. -## Quick Start +1. **Detect** recurring recipes and pitfalls across completed task runs. +2. **Distill** useful patterns into standard `SKILL.md` drafts. +3. **Approve** the draft before anything is installed. +4. **Deliver** approved skills to `~/.agentforge/skills`, symlinked into `~/.claude/skills` and `~/.agents/skills`. -Once the app is running, the backend listens on `http://127.0.0.1:9712`. +The automatic sweep is off by default because it spends tokens. Enable it in Settings, or run a manual scan from the Skills tab. -**Create a task via the UI:** Open AgentForge, click **New Task**, fill in the title, prompt, and working directory, choose a schedule type, and click **Create**. +## Architecture -**Create a task via curl:** +```text +┌─────────────────────┐ HTTP/JSON ┌─────────────────────┐ +│ Electron + React UI │ <--------------------> │ Bun TypeScript API │ +│ Task board renderer │ 127.0.0.1:9712 │ Scheduler + runner │ +└─────────────────────┘ └──────────┬──────────┘ + │ + ┌─────────────────────┼─────────────────────┐ + │ │ │ + SQLite TaskScheduler AgentExecutor + ~/.agentforge cron + delayed claude / codex +``` -```bash -# Run immediately (uses default agent — claude) -curl -X POST http://localhost:9712/api/tasks \ - -H "Content-Type: application/json" \ - -d '{ - "title": "Code review", - "prompt": "Review the code changes in this directory", - "working_dir": "~/projects/myapp", - "schedule_type": "immediate" - }' +- `taskboard-electron/src/main.ts` starts and stops the backend with the desktop app. +- `taskboard-electron/src/renderer/App.tsx` renders the task board, heartbeats, settings, and skills. +- `backend/taskboard.ts` serves the REST API through `Bun.serve`. +- `backend/src/db.ts` stores tasks, runs, output events, settings, heartbeats, and skills in SQLite. +- `backend/src/executor.ts` runs Claude Code or Codex CLI and persists streaming output. +- `backend/src/scheduler.ts` polls for due tasks and handles schedules, heartbeats, and skill sweeps. -# Run after a delay (seconds) -curl -X POST http://localhost:9712/api/tasks \ - -H "Content-Type: application/json" \ - -d '{ - "title": "Delayed review", - "prompt": "Review recent commits", - "working_dir": "~/projects/myapp", - "schedule_type": "delayed", - "delay_seconds": 300 - }' +## API Quick Example -# Recurring cron schedule (daily at 9 AM) -curl -X POST http://localhost:9712/api/tasks \ - -H "Content-Type: application/json" \ - -d '{ - "title": "Daily digest", - "prompt": "Summarize TODO comments across the codebase", - "working_dir": "~/projects/myapp", - "schedule_type": "cron", - "cron_expr": "0 9 * * *", - "max_runs": 30 - }' +All API routes are served at `http://127.0.0.1:9712/api`. -# Use Codex CLI instead of Claude Code -curl -X POST http://localhost:9712/api/tasks \ +```bash +curl -X POST http://127.0.0.1:9712/api/tasks \ -H "Content-Type: application/json" \ -d '{ - "title": "Refactor auth module", - "prompt": "Refactor the authentication module for clarity", + "title": "Review auth changes", + "prompt": "Review the latest auth module diff for regressions and missing tests.", "working_dir": "~/projects/myapp", "schedule_type": "immediate", "agent": "codex" }' ``` -To set Codex as the default agent for all tasks: +Common task endpoints: -```bash -curl -X PUT http://localhost:9712/api/settings \ - -H "Content-Type: application/json" \ - -d '{"default_agent": "codex"}' -``` - ---- +| Method | Endpoint | Description | +| --- | --- | --- | +| `GET` | `/api/tasks` | List tasks | +| `POST` | `/api/tasks` | Create a task | +| `GET` | `/api/tasks/:id` | Read one task | +| `POST` | `/api/tasks/:id/cancel` | Cancel a task | +| `POST` | `/api/tasks/:id/retry` | Retry a failed task | +| `GET` | `/api/tasks/:id/events` | Read structured output events | +| `GET` | `/api/health` | Backend health check | ## Chat Channels -Control AgentForge from your favorite messaging app. Channels auto-start when the corresponding environment variables are detected. +Channels are optional. They let teammates create tasks or receive updates from the tools where they already talk. -| Channel | Transport | Difficulty | -|---------|-----------|------------| -| Telegram | Bot API (polling) | Easy | -| Slack | Socket Mode | Moderate | -| Feishu / Lark | WebSocket long-connection | Moderate | -| WeChat | Node bridge (experimental) | Moderate | +| Channel | Transport | Notes | +| --- | --- | --- | +| Telegram | Bot API polling | Simple bot token setup. | +| Slack | Socket Mode | Uses bot and app-level tokens. | +| Feishu / Lark | WebSocket long connection | No public IP required. | +| WeChat | Local bridge | Experimental text-only bridge. | -
-Telegram setup +Channel adapters live in `backend/src/channels/` and can be configured from the app settings page or REST endpoints. -### 1. Create a Bot +## Multi-Agent Pipelines -1. Chat with [@BotFather](https://t.me/BotFather) on Telegram. -2. Send `/newbot` and follow the prompts. -3. Copy the **HTTP API token**. +AgentForge ships with `skills/agentforge/`, a Claude Code skill that lets one task create and manage other AgentForge tasks. -### 2. Configure +That enables fan-out research, fan-in summaries, dependency gates, upstream result injection, and scheduled sub-workflows. -```bash -export TELEGRAM_BOT_TOKEN="123456789:ABCdefGHIjklMNOpqrSTUvwxYZ" -export TELEGRAM_ALLOWED_USERS="123456789" # optional — restrict access by user ID +```text +User + | + v +Task A --creates--> Task B + | Task C + | | + `------depends------v + Task D ``` -### 3. Commands - -| Command | Description | -|---------|-------------| -| `/newtask \| <prompt>` | Create a task | -| `/list` | List all tasks | -| `/status <id>` | Task details | -| `/cancel <id>` | Cancel a task | -| `/help` | Show help | - -</details> - -<details> -<summary><b>Slack setup</b></summary> - -### 1. Create a Slack App - -Go to [api.slack.com/apps](https://api.slack.com/apps) → **Create New App → From scratch**. - -### 2. Bot Permissions - -Under **OAuth & Permissions → Bot Token Scopes**, add: -`app_mentions:read`, `chat:write`, `im:history`, `im:read`, `im:write`, `reactions:write` - -### 3. Enable Socket Mode - -**Socket Mode** → toggle on → generate an **App-Level Token** (`xapp-…`) with `connections:write`. - -### 4. Subscribe to Events - -**Event Subscriptions → Subscribe to bot events**: `app_mention`, `message.im` - -### 5. Configure +Install the bundled skill: ```bash -export SLACK_BOT_TOKEN="xoxb-..." -export SLACK_APP_TOKEN="xapp-..." -export SLACK_ALLOWED_USERS="U012AB3CD" # optional +ln -s /path/to/agentforge/skills/agentforge ~/.claude/skills/agentforge ``` -### 6. Commands - -Send as a DM or @mention: +## Development -| Command | Description | -|---------|-------------| -| `newtask <title> \| <prompt>` | Create a task | -| `list` | List all tasks | -| `status <id>` | Task details | -| `cancel <id>` | Cancel a task | - -</details> - -<details> -<summary><b>Feishu / Lark setup</b></summary> - -Feishu uses the settings API (no environment variables needed). It connects via WebSocket — no public IP required. - -### 1. Create a Feishu App - -- [Feishu Open Platform](https://open.feishu.cn/app) → new app → enable **Bot** -- Permissions: `im:message`, `im:resource` -- Events: `im.message.receive_v1` with **Long Connection** mode -- Copy **App ID** and **App Secret** - -### 2. Configure via API +Backend: ```bash -curl -X POST http://127.0.0.1:9712/api/feishu/settings \ - -H "Content-Type: application/json" \ - -d '{ - "feishu_app_id": "cli_xxxx", - "feishu_app_secret": "your_app_secret", - "feishu_default_working_dir": "~/projects", - "feishu_enabled": "true" - }' +cd backend +bun taskboard.ts ``` -Or configure from the desktop app's settings page. - -</details> - -<details> -<summary><b>WeChat setup (experimental)</b></summary> - -WeChat uses a Bun/TypeScript sidecar bridge — no environment variables needed. Configure and enable it via the API or the desktop app's settings page. - -### 1. Enable via API +Frontend and Electron app: ```bash -curl -X POST http://127.0.0.1:9712/api/channels/settings \ - -H "Content-Type: application/json" \ - -d '{ - "weixin_enabled": "true", - "weixin_default_working_dir": "~/projects", - "weixin_base_url": "https://ilinkai.weixin.qq.com", - "weixin_account_id": "" - }' +cd taskboard-electron +bun run start ``` -### 2. Scan the QR code - -On first launch the bridge will request a QR code login. Scan it with your WeChat mobile app to authenticate. The session is saved for subsequent restarts. - -### Notes - -- Text-only MVP — rich media is not supported. -- The bridge uses the `getupdates` / `sendmessage` HTTP protocol. -- Single-account per AgentForge instance. - -</details> - -Channel adapters live in [`backend/src/channels/`](backend/src/channels/). Configure them from the desktop app settings page or the REST endpoints shown above. - ---- - -## Multi-Agent Pipelines (DAG) - -AgentForge ships with a **Claude Code skill** (`skills/agentforge/`) that lets any agent running inside Claude Code create and manage other AgentForge tasks — enabling recursive, multi-agent workflows. - -``` - User - | - v - [ Task A ] ──creates──> [ Task B ] [ Task C ] - | - creates - v - [ Task D ] (depends on B) -``` - -### Install the skill +Quality gates: ```bash -# Symlink into your Claude Code skills directory -ln -s /path/to/agentforge/skills/agentforge ~/.claude/skills/agentforge -``` - -### What the skill enables - -- **Fan-out / fan-in** — Spawn N sub-tasks in parallel, poll until complete, synthesize results -- **DAG dependencies** — `--depends-on <ids>` makes downstream tasks wait for upstream to finish; failures cascade-cancel the rest -- **Result injection** — `--inject-result` prepends upstream output into the downstream prompt automatically -- **Scheduled sub-workflows** — Combine cron, delayed, and immediate schedules within a pipeline +# Backend CI gate +make check -### Example — parallel research pipeline - -``` -Research the top 3 competitors of Acme Corp. -For each competitor, create a separate AgentForge task that: - 1. Gathers recent news and financials - 2. Writes a one-page summary -Then create a final task (depends-on the above 3) that -synthesizes everything into a comparative report. +# Frontend CI gate +cd taskboard-electron +bun run typecheck +bun run lint +bun run format:check +bun run test +bun run build:check ``` -AgentForge handles scheduling, dependency tracking, and result passing automatically. - ---- - -## API Reference +## Troubleshooting -All endpoints are served at `http://127.0.0.1:9712/api`. - -| Method | Endpoint | Description | -|--------|----------|-------------| -| GET | `/api/tasks` | List all tasks | -| GET | `/api/tasks/:id` | Get a single task | -| POST | `/api/tasks` | Create a task | -| POST | `/api/tasks/:id/cancel` | Cancel a task | -| POST | `/api/tasks/:id/retry` | Retry a failed task | -| DELETE | `/api/tasks/:id` | Delete a task | -| GET | `/api/tasks/:id/runs` | Get run history | -| GET | `/api/tasks/:id/output` | Get accumulated output | -| GET | `/api/tasks/:id/events` | Get structured output events | -| GET | `/api/health` | Health check | - ---- - -## Background Service - -To keep the backend running persistently without the desktop app: - -```xml -<!-- ~/Library/LaunchAgents/com.agentforge.taskboard.plist --> -<?xml version="1.0" encoding="UTF-8"?> -<!DOCTYPE plist PUBLIC "-//Apple//DTD PLIST 1.0//EN" - "http://www.apple.com/DTDs/PropertyList-1.0.dtd"> -<plist version="1.0"> -<dict> - <key>Label</key> - <string>com.agentforge.taskboard</string> - <key>ProgramArguments</key> - <array> - <string>/usr/local/bin/bun</string> - <string>/path/to/agentforge/backend/taskboard.ts</string> - </array> - <key>WorkingDirectory</key> - <string>/path/to/agentforge</string> - <key>RunAtLoad</key> - <true/> - <key>KeepAlive</key> - <true/> - <key>StandardOutPath</key> - <string>/tmp/agentforge.log</string> - <key>StandardErrorPath</key> - <string>/tmp/agentforge.err</string> -</dict> -</plist> -``` +If `bun install` stalls while downloading Electron, use a mirror: ```bash -launchctl load ~/Library/LaunchAgents/com.agentforge.taskboard.plist -``` - ---- - -## Architecture - -``` -┌──────────────────┐ HTTP/JSON ┌──────────────────┐ -│ React Frontend │ <────────────────> │ Bun/TS Backend │ -│ (Kanban UI) │ localhost:9712 │ (Scheduler+API) │ -└──────────────────┘ └───────┬──────────┘ - | - ┌───────────────┼───────────────┐ - v v v - [ SQLite DB ] [ Scheduler ] [ Claude CLI ] +export ELECTRON_MIRROR=https://npmmirror.com/mirrors/electron/ +cd taskboard-electron +bun install --registry https://registry.npmmirror.com --ignore-scripts ``` -- **Bun/TypeScript backend** (`backend/`) — `Bun.serve` HTTP server. Manages tasks in SQLite via `bun:sqlite` (`~/.agentforge/tasks.db`), runs `claude` or `codex` CLI via `AgentExecutor`, and schedules work with `TaskScheduler` (polls every 2 s, supports cron via `cron-parser`). -- **Electron shell** (`taskboard-electron/`) — spawns the backend on start, kills it on quit. Renderer/main/preload are bundled with `Bun.build` (`scripts/build.ts`); in dev `bun run start` watches and reloads. -- **React frontend** (`App.tsx`) — single-component kanban board that polls the REST API and renders colorized streaming output. - ---- +More setup notes live in [docs/installation-troubleshooting.md](docs/installation-troubleshooting.md). ## Contributing -Contributions are welcome! Here's how to get started: - 1. Fork the repository and create a feature branch. -2. Start the app in development mode (see [Option 3](#option-3-development-mode) above). -3. Make your changes and run the relevant Bun quality gate (`make check` for backend changes; the frontend gate from `taskboard-electron/` for renderer/Electron changes). -4. Open a pull request with a clear description of the change. +2. Run the app in development mode with `cd taskboard-electron && bun run start`. +3. Keep changes scoped and run the relevant quality gate before opening a PR. -**Key files:** -- `backend/` — entire TypeScript backend (DB, scheduler, executor, HTTP API, channels) -- `taskboard-electron/src/main.ts` — Electron main process -- `taskboard-electron/src/renderer/App.tsx` — React frontend -- `backend/src/channels/` — pluggable chat channel adapters (Telegram, Slack, Feishu, WeChat) -- `skills/agentforge/` — Claude Code skill for agent-to-agent delegation +Key files: ---- +- `backend/` - Bun/TypeScript backend, scheduler, executor, API, and channels. +- `taskboard-electron/src/main.ts` - Electron main process. +- `taskboard-electron/src/renderer/App.tsx` - React renderer. +- `skills/agentforge/` - bundled skill for agent-to-agent delegation. ## License -MIT — see [LICENSE](LICENSE) for details. +MIT - see [LICENSE](LICENSE). diff --git a/README.zh.md b/README.zh.md index f9e1b06..8f1b749 100644 --- a/README.zh.md +++ b/README.zh.md @@ -1,426 +1,224 @@ # AgentForge -> 在 Mac 上编排 AI 编程智能体 —— 通过看板界面调度、监控和串联 Claude Code 任务。 +**一个本地优先的 macOS 智能体工作台，用看板方式调度、观察和复用 Claude Code / OpenAI Codex CLI 任务。** [![License: MIT](https://img.shields.io/badge/License-MIT-blue.svg)](LICENSE) -[![Platform: macOS](https://img.shields.io/badge/Platform-macOS%2012%2B-lightgrey?logo=apple)](https://github.com/releases) -[![Bun 1.3+](https://img.shields.io/badge/Bun-1.3%2B-black?logo=bun)](https://bun.sh) [![TypeScript](https://img.shields.io/badge/TypeScript-5-blue?logo=typescript)](https://www.typescriptlang.org) -[![Claude Code](https://img.shields.io/badge/Powered%20by-Claude%20Code-orange)](https://docs.anthropic.com/en/docs/claude-code) +[![Platform: macOS](https://img.shields.io/badge/Platform-macOS%2012%2B-lightgrey?logo=apple)](../../releases) +[![Bun 1.3+](https://img.shields.io/badge/Bun-1.3%2B-black?logo=bun)](https://bun.sh) +[![TypeScript](https://img.shields.io/badge/TypeScript-5-blue?logo=typescript)](https://www.typescriptlang.org) -**官网**: https://agentforge-landing-weld.vercel.app/ +AgentForge 把 AI 编程智能体变成可见、可排队、可调度的本地工作流。你可以创建任务、查看实时输出、设置周期检查、从聊天工具触发任务，并把重复工作沉淀成可复用技能。 -![AgentForge 截图](assets/screenshot.png) +**官网:** https://agentforge-landing-weld.vercel.app/ ---- +![AgentForge 任务看板预览](assets/readme-preview.svg) -## 目录 +## 为什么用 AgentForge -- [功能特性](#功能特性) -- [技能库（Skill Library）](#技能库skill-library) -- [环境要求](#环境要求) -- [安装](#安装) -- [常见问题](#常见问题) -- [快速开始](#快速开始) -- [消息频道](#消息频道) -- [多智能体流水线（DAG）](#多智能体流水线dag) -- [API 参考](#api-参考) -- [后台服务](#后台服务) -- [架构说明](#架构说明) -- [参与贡献](#参与贡献) -- [许可证](#许可证) +- **不用盯着终端等结果。** 把任务放进队列，从 Queue / Running / Done 看板里查看状态。 +- **每个任务选择合适智能体。** 任务级支持 Claude Code 或 Codex CLI，也可以设置默认智能体。 +- **把重复工作产品化。** 支持立即、延迟、定时和 cron 调度，成功模式还可以进入 Skill Library。 +- **执行留在本机。** 后端只监听 `127.0.0.1`，状态存进 SQLite，智能体 CLI 在你的 Mac 上运行。 ---- - -## 功能特性 - -- **🧬 自复利技能库（Skill Library）** —— AgentForge 自动发现你反复跑的任务模式，借助 Anthropic [skill-creator](https://github.com/anthropics/skills/tree/main/skills/skill-creator) 蒸馏成可复用的 Claude Code Skill。审核、批准后自动 symlink 进 Claude Code 与 Codex —— 用得越多，agent 越懂你的活。[了解更多 →](#技能库skill-library) -- **智能体生成智能体** —— 任何运行中的任务都可通过内置的 Claude Code Skill 创建子任务、构建 DAG 流水线并汇总结果 -- **看板任务视图** —— 可视化的「待执行 / 运行中 / 已完成」三列布局，支持实时输出流 -- **DAG 流水线** —— 定义任务依赖关系，支持自动级联执行和失败传播 -- **灵活调度** —— 即刻执行、延迟执行、指定时间一次性执行、cron 周期任务 -- **消息频道控制** —— 通过 Telegram、Slack 或飞书/Lark 创建任务并接收完成通知 -- **持久化存储** —— SQLite 存储任务历史、运行记录和流式输出 -- **原生 macOS 应用** —— Electron 封装，一键 DMG 安装 - ---- - -## 技能库（Skill Library） - -**把你反复做的活，自动沉淀成 agent 可复用的技能。** 用得越多，AgentForge 越懂你的工作流。 - -1. **检测** —— 后台定时扫描（或手动点「扫一遍」）分析你已完成的任务，把跨任务复发的模式记进账本：既包括*成功配方*（你反复跑的工作流），也包括*避坑*（你诊断并修好的失败）。按 run 幂等计数，重复扫描不会灌水。 -2. **蒸馏** —— 对值得固化的模式，AgentForge 用 Anthropic 的 **[skill-creator](https://github.com/anthropics/skills/tree/main/skills/skill-creator)** 技能起草规范的 `SKILL.md` —— 而且会**先判断到底值不值得**做成 skill，免得噪声泛滥。 -3. **审批** —— 审阅草稿：价值判定、贡献的任务、以及完全可编辑的 `SKILL.md`，批准或驳回，决定权在你。 -4. **投递** —— 批准后写一份 canonical 文件，并 symlink 进 **Claude Code**（`~/.claude/skills`）和 **Codex**（`~/.agents/skills`）两个原生目录，两个 agent 都按渐进式披露原生加载 —— 零 prompt 注入。 -5. **管理** —— 在 **Skills** 标签里逐个启用/停用（增删 symlink，保留文件）、查看 `SKILL.md`、或删除。 - -> 自动每日扫描**默认关闭**（会消耗 token）。在 **设置 → Skill Library** 里打开、选 agent 和节奏，或随时点手动「扫一遍」—— 手动按钮不受阈值限制，永远可用。 - -蒸馏出的都是标准 Claude Code `SKILL.md` 文件（canonical 存于 `~/.agentforge/skills`，symlink 到各 agent 原生目录）—— 可移植、可检视、零锁定。 - ---- - -## 环境要求 - -- macOS 12.0+（Apple Silicon 或 Intel） -- [Bun](https://bun.sh) 1.3+ -- [Claude Code CLI](https://docs.anthropic.com/en/docs/claude-code)（已安装并在 `PATH` 中） - ---- - -## 安装 +## 快速开始 -### 方式一：桌面应用（推荐） +### 安装桌面 App -1. 从 [Releases](../../releases) 页面下载最新的 `AgentForge-*.dmg`。 -2. 打开 DMG，将 **AgentForge** 拖入 `/Applications`。 -3. 从启动台或应用程序文件夹打开。 +1. 从 [Releases](../../releases) 下载最新 `AgentForge-*.dmg`。 +2. 把 **AgentForge** 拖进 `/Applications`。 +3. 启动 App，选择工作目录，创建第一个任务。 -### 方式二：从源码构建 +### 从源码运行 ```bash -git clone https://github.com/your-org/agentforge.git -cd agentforge - -# 安装全部依赖 -make install-deps +git clone https://github.com/hetaoBackend/agentforge.git +cd agentforge/taskboard-electron -# 构建并打包 DMG -make package-dmg -# 输出路径：taskboard-electron/out/make/AgentForge-1.0.0-arm64.dmg +bun install --ignore-scripts +bun run start ``` -### 方式三：开发模式 +开发命令会构建 Electron/React 前端、启动 Bun 后端，并在改动后自动刷新。 -```bash -git clone https://github.com/your-org/agentforge.git -cd agentforge - -make install-deps # bun install in backend/ and taskboard-electron/ +### 构建 DMG -# 启动 Electron + 后端（Bun 负责构建、监听并自动拉起后端） -cd taskboard-electron && bun run start +```bash +cd taskboard-electron +bun run make ``` ---- +产物位于 `taskboard-electron/out/make/AgentForge-1.0.0-arm64.dmg`。 -## 常见问题 +## 核心能力 -### `bun install` 卡住不动 +| 模块 | 作用 | +| --- | --- | +| **Tasks** | Pending、scheduled、blocked、running、completed、failed、cancelled 任务看板。 | +| **Scheduling** | 支持 immediate、delayed、`scheduled_at`、cron 和最大运行次数。 | +| **Live output** | 持久化 Claude Code / Codex CLI 的结构化流式输出。 | +| **Heartbeats** | 周期检查，可以决定 trigger、resume 或 notify。 | +| **Skill Library** | 从完成任务中发现模式，生成可编辑 `SKILL.md`，再安装到智能体技能目录。 | +| **Chat channels** | Telegram、Slack、飞书/Lark、微信渠道可创建或跟踪任务。 | +| **DAG pipelines** | 支持任务依赖、失败传播和上游结果注入。 | -这是最常见的安装问题。Electron 二进制包约 100MB，从 GitHub 下载，在国内网络环境下极易卡住。 +## 环境要求 -**快速解决——使用国内镜像：** +- macOS 12+ +- [Bun](https://bun.sh) 1.3+ +- [Claude Code CLI](https://docs.anthropic.com/en/docs/claude-code) 或 [OpenAI Codex CLI](https://github.com/openai/codex) 在 `PATH` 中 -```bash -export ELECTRON_MIRROR=https://npmmirror.com/mirrors/electron/ -cd taskboard-electron && bun install --registry https://registry.npmmirror.com -``` +## Skill Library -**完整排查指南：** [docs/installation-troubleshooting.md](docs/installation-troubleshooting.md)，涵盖： -- 如何判断是网络卡住还是编译卡住 -- 国内镜像及永久配置方法 -- macOS / Linux / Windows 编译工具安装 -- Node.js 版本要求 -- 完整清理重装流程 +Skill Library 会把你反复做的工作沉淀成智能体可复用技能。 ---- +1. **Detect**: 从已完成任务中发现重复 recipe 和常见 pitfall。 +2. **Distill**: 把值得复用的模式蒸馏成标准 `SKILL.md` 草稿。 +3. **Approve**: 你先审核和编辑草稿，再决定是否安装。 +4. **Deliver**: 批准后的技能写入 `~/.agentforge/skills`，并 symlink 到 `~/.claude/skills` 与 `~/.agents/skills`。 -## 快速开始 +自动 sweep 默认关闭，因为它会消耗 token。你可以在 Settings 里开启，也可以在 Skills tab 手动扫描。 -应用启动后，后端监听在 `http://127.0.0.1:9712`。 +## 架构 -**通过 UI 创建任务：** 打开 AgentForge，点击 **New Task**，填写标题、提示词和工作目录，选择调度类型后点击 **Create**。 +```text +┌─────────────────────┐ HTTP/JSON ┌─────────────────────┐ +│ Electron + React UI │ <--------------------> │ Bun TypeScript API │ +│ Task board renderer │ 127.0.0.1:9712 │ Scheduler + runner │ +└─────────────────────┘ └──────────┬──────────┘ + │ + ┌─────────────────────┼─────────────────────┐ + │ │ │ + SQLite TaskScheduler AgentExecutor + ~/.agentforge cron + delayed claude / codex +``` -**通过命令行创建任务：** +- `taskboard-electron/src/main.ts` 随桌面 App 启停后端。 +- `taskboard-electron/src/renderer/App.tsx` 渲染任务看板、心跳、设置和技能库。 +- `backend/taskboard.ts` 通过 `Bun.serve` 暴露 REST API。 +- `backend/src/db.ts` 用 SQLite 存储任务、运行记录、输出事件、设置、心跳和技能。 +- `backend/src/executor.ts` 运行 Claude Code 或 Codex CLI，并持久化流式输出。 +- `backend/src/scheduler.ts` 轮询 due tasks，并处理调度、心跳和技能扫描。 -```bash -# 立即执行 -curl -X POST http://localhost:9712/api/tasks \ - -H "Content-Type: application/json" \ - -d '{ - "title": "代码审查", - "prompt": "审查当前目录的代码变更", - "working_dir": "~/projects/myapp", - "schedule_type": "immediate" - }' +## API 示例 -# 延迟执行（300 秒后） -curl -X POST http://localhost:9712/api/tasks \ - -H "Content-Type: application/json" \ - -d '{ - "title": "延迟审查", - "prompt": "审查最近的提交", - "working_dir": "~/projects/myapp", - "schedule_type": "delayed", - "delay_seconds": 300 - }' +所有 API 都在 `http://127.0.0.1:9712/api` 下。 -# cron 周期任务（每天早上 9 点） -curl -X POST http://localhost:9712/api/tasks \ +```bash +curl -X POST http://127.0.0.1:9712/api/tasks \ -H "Content-Type: application/json" \ -d '{ - "title": "每日摘要", - "prompt": "汇总代码库中的所有 TODO 注释", + "title": "Review auth changes", + "prompt": "Review the latest auth module diff for regressions and missing tests.", "working_dir": "~/projects/myapp", - "schedule_type": "cron", - "cron_expr": "0 9 * * *", - "max_runs": 30 + "schedule_type": "immediate", + "agent": "codex" }' ``` ---- - -## 消息频道 +常用任务接口: -通过消息应用控制 AgentForge。检测到对应环境变量后，频道会自动启动。 +| Method | Endpoint | Description | +| --- | --- | --- | +| `GET` | `/api/tasks` | 列出任务 | +| `POST` | `/api/tasks` | 创建任务 | +| `GET` | `/api/tasks/:id` | 读取单个任务 | +| `POST` | `/api/tasks/:id/cancel` | 取消任务 | +| `POST` | `/api/tasks/:id/retry` | 重试失败任务 | +| `GET` | `/api/tasks/:id/events` | 读取结构化输出事件 | +| `GET` | `/api/health` | 后端健康检查 | -| 频道 | 传输方式 | 配置难度 | -|------|----------|----------| -| Telegram | Bot API（轮询） | 简单 | -| Slack | Socket Mode | 中等 | -| 飞书 / Lark | WebSocket 长连接 | 中等 | +## 消息渠道 -<details> -<summary><b>Telegram 配置</b></summary> +消息渠道是可选能力，用来从常用聊天工具创建任务或接收状态更新。 -### 1. 创建机器人 - -1. 在 Telegram 中与 [@BotFather](https://t.me/BotFather) 对话。 -2. 发送 `/newbot` 并按提示操作。 -3. 复制 **HTTP API Token**。 - -### 2. 配置环境变量 - -```bash -export TELEGRAM_BOT_TOKEN="123456789:ABCdefGHIjklMNOpqrSTUvwxYZ" -export TELEGRAM_ALLOWED_USERS="123456789" # 可选——限制允许访问的用户 ID -``` +| 渠道 | 传输方式 | 说明 | +| --- | --- | --- | +| Telegram | Bot API polling | 配置 bot token 即可使用。 | +| Slack | Socket Mode | 使用 bot token 和 app-level token。 | +| 飞书 / Lark | WebSocket long connection | 不需要公网 IP。 | +| 微信 | Local bridge | 实验性文本渠道。 | -### 3. 可用命令 +渠道适配器位于 `backend/src/channels/`，可以在 App 设置页或 REST API 中配置。 -| 命令 | 说明 | -|------|------| -| `/newtask <标题> \| <提示词>` | 创建任务 | -| `/list` | 列出所有任务 | -| `/status <id>` | 查看任务详情 | -| `/cancel <id>` | 取消任务 | -| `/help` | 显示帮助 | +## 多智能体流水线 -</details> +AgentForge 自带 `skills/agentforge/`，让 Claude Code 中运行的任务可以创建和管理其他 AgentForge 任务。 -<details> -<summary><b>Slack 配置</b></summary> +这可以用于并行研究、结果汇总、依赖卡点、上游结果注入和周期性子工作流。 -### 1. 创建 Slack App - -前往 [api.slack.com/apps](https://api.slack.com/apps) → **Create New App → From scratch**。 - -### 2. Bot 权限 - -在 **OAuth & Permissions → Bot Token Scopes** 中添加： -`app_mentions:read`、`chat:write`、`im:history`、`im:read`、`im:write`、`reactions:write` - -### 3. 启用 Socket Mode - -**Socket Mode** → 开启 → 生成 **App-Level Token**（`xapp-…`），scope 选 `connections:write`。 - -### 4. 订阅事件 - -**Event Subscriptions → Subscribe to bot events**：`app_mention`、`message.im` - -### 5. 配置环境变量 - -```bash -export SLACK_BOT_TOKEN="xoxb-..." -export SLACK_APP_TOKEN="xapp-..." -export SLACK_ALLOWED_USERS="U012AB3CD" # 可选 +```text +User + | + v +Task A --creates--> Task B + | Task C + | | + `------depends------v + Task D ``` -### 6. 可用命令 - -以私信或 @mention 方式发送： - -| 命令 | 说明 | -|------|------| -| `newtask <标题> \| <提示词>` | 创建任务 | -| `list` | 列出所有任务 | -| `status <id>` | 查看任务详情 | -| `cancel <id>` | 取消任务 | - -</details> - -<details> -<summary><b>飞书 / Lark 配置</b></summary> - -飞书通过设置 API 配置（无需环境变量），通过 WebSocket 长连接通信，无需公网 IP。 - -### 1. 创建飞书应用 - -- 前往 [飞书开放平台](https://open.feishu.cn/app) → 新建应用 → 开启**机器人**能力 -- 权限：`im:message`、`im:resource` -- 事件：`im.message.receive_v1`，选择**长连接**模式 -- 复制 **App ID** 和 **App Secret** - -### 2. 通过 API 配置 +安装内置技能: ```bash -curl -X POST http://127.0.0.1:9712/api/feishu/settings \ - -H "Content-Type: application/json" \ - -d '{ - "feishu_app_id": "cli_xxxx", - "feishu_app_secret": "your_app_secret", - "feishu_default_working_dir": "~/projects", - "feishu_enabled": "true" - }' +ln -s /path/to/agentforge/skills/agentforge ~/.claude/skills/agentforge ``` -也可以在桌面应用的设置页面中配置。 +## 开发 -</details> - -频道适配器位于 [`backend/src/channels/`](backend/src/channels/)，可在桌面应用设置页或上面的 REST 接口中配置。 - ---- - -## 多智能体流水线（DAG） - -AgentForge 内置了一个 **Claude Code Skill**（`skills/agentforge/`），让任何在 Claude Code 中运行的智能体都能创建和管理其他 AgentForge 任务，从而实现递归的多智能体工作流。 - -``` - 用户 - | - v - [ 任务 A ] ──创建──> [ 任务 B ] [ 任务 C ] - | - 创建 - v - [ 任务 D ] (依赖 B) -``` - -### 安装 Skill +后端: ```bash -# 将 skill 目录符号链接到 Claude Code skills 目录 -ln -s /path/to/agentforge/skills/agentforge ~/.claude/skills/agentforge +cd backend +bun taskboard.ts ``` -### Skill 的能力 - -- **扇出 / 扇入** —— 并行创建 N 个子任务，轮询等待全部完成后汇总结果 -- **DAG 依赖** —— `--depends-on <ids>` 使下游任务等待上游完成；上游失败会级联取消下游 -- **结果注入** —— `--inject-result` 自动将上游输出追加到下游任务的提示词中 -- **调度子工作流** —— 在流水线中混合使用 cron、延迟和即刻调度 - -### 示例——并行研究流水线 +前端和 Electron App: -``` -研究 Acme Corp 的前 3 名竞争对手。 -针对每个竞争对手，分别创建一个 AgentForge 任务： - 1. 收集近期新闻和财务数据 - 2. 撰写一页摘要报告 -然后创建一个汇总任务（依赖上面 3 个任务）， -将所有内容整合成一份对比分析报告。 +```bash +cd taskboard-electron +bun run start ``` -AgentForge 自动处理调度、依赖追踪和结果传递。 - ---- - -## API 参考 - -所有接口均在 `http://127.0.0.1:9712/api` 下提供。 - -| 方法 | 路径 | 说明 | -|------|------|------| -| GET | `/api/tasks` | 获取所有任务 | -| GET | `/api/tasks/:id` | 获取单个任务 | -| POST | `/api/tasks` | 创建任务 | -| POST | `/api/tasks/:id/cancel` | 取消任务 | -| POST | `/api/tasks/:id/retry` | 重试失败任务 | -| DELETE | `/api/tasks/:id` | 删除任务 | -| GET | `/api/tasks/:id/runs` | 获取运行历史 | -| GET | `/api/tasks/:id/output` | 获取累计输出 | -| GET | `/api/tasks/:id/events` | 获取结构化输出事件 | -| GET | `/api/health` | 健康检查 | - ---- - -## 后台服务 - -如需在不打开桌面应用的情况下持续运行后端，可用 launchd 托管： - -```xml -<!-- ~/Library/LaunchAgents/com.agentforge.taskboard.plist --> -<?xml version="1.0" encoding="UTF-8"?> -<!DOCTYPE plist PUBLIC "-//Apple//DTD PLIST 1.0//EN" - "http://www.apple.com/DTDs/PropertyList-1.0.dtd"> -<plist version="1.0"> -<dict> - <key>Label</key> - <string>com.agentforge.taskboard</string> - <key>ProgramArguments</key> - <array> - <string>/usr/local/bin/bun</string> - <string>/path/to/agentforge/backend/taskboard.ts</string> - </array> - <key>WorkingDirectory</key> - <string>/path/to/agentforge</string> - <key>RunAtLoad</key> - <true/> - <key>KeepAlive</key> - <true/> - <key>StandardOutPath</key> - <string>/tmp/agentforge.log</string> - <key>StandardErrorPath</key> - <string>/tmp/agentforge.err</string> -</dict> -</plist> -``` +质量门禁: ```bash -launchctl load ~/Library/LaunchAgents/com.agentforge.taskboard.plist +# Backend CI gate +make check + +# Frontend CI gate +cd taskboard-electron +bun run typecheck +bun run lint +bun run format:check +bun run test +bun run build:check ``` ---- +## 常见问题 -## 架构说明 +如果 `bun install` 下载 Electron 时卡住，可以使用镜像: -``` -┌──────────────────┐ HTTP/JSON ┌──────────────────┐ -│ React 前端 │ <────────────────> │ Bun/TS 后端 │ -│ （看板 UI） │ localhost:9712 │ （调度器+API） │ -└──────────────────┘ └───────┬──────────┘ - | - ┌───────────────┼───────────────┐ - v v v - [ SQLite DB ] [ 调度器 ] [ Claude CLI ] +```bash +export ELECTRON_MIRROR=https://npmmirror.com/mirrors/electron/ +cd taskboard-electron +bun install --registry https://registry.npmmirror.com --ignore-scripts ``` -- **Bun/TypeScript 后端**（`backend/`）—— `Bun.serve` HTTP 服务。通过 `bun:sqlite` 在 SQLite（`~/.agentforge/tasks.db`）中管理任务，通过 `AgentExecutor` 运行 `claude` / `codex` CLI，通过 `TaskScheduler` 调度任务（每 2 秒轮询，支持通过 `cron-parser` 解析 cron 表达式）。 -- **Electron 外壳**（`taskboard-electron/`）—— 启动时创建后端进程，退出时终止它。主进程/preload/渲染器均由 `Bun.build` 打包（`scripts/build.ts`）；开发模式 `bun run start` 自动监听并热重载。 -- **React 前端**（`App.tsx`）—— 单组件看板，轮询 REST API 并渲染带颜色的流式输出。 - ---- +更多说明见 [docs/installation-troubleshooting.md](docs/installation-troubleshooting.md)。 ## 参与贡献 -欢迎提交 Issue 和 Pull Request！ - -1. Fork 本仓库并创建功能分支。 -2. 以开发模式启动应用（参见[方式三](#方式三开发模式)）。 -3. 修改代码并运行对应的 Bun 质量门禁（后端改动运行 `make check`；前端/Electron 改动在 `taskboard-electron/` 下运行前端门禁）。 -4. 提交 PR，并清晰描述改动内容。 +1. Fork 仓库并创建功能分支。 +2. 使用 `cd taskboard-electron && bun run start` 启动开发模式。 +3. 保持改动聚焦，提交 PR 前运行相关质量门禁。 -**关键文件：** -- `backend/` —— 整个 TypeScript 后端（数据库、调度器、执行器、HTTP API、频道） -- `taskboard-electron/src/main.ts` —— Electron 主进程 -- `taskboard-electron/src/renderer/App.tsx` —— React 前端 -- `backend/src/channels/` —— 可插拔消息频道适配器 -- `skills/agentforge/` —— 用于智能体间委托的 Claude Code Skill +关键文件: ---- +- `backend/` - Bun/TypeScript 后端、调度器、执行器、API 和渠道。 +- `taskboard-electron/src/main.ts` - Electron 主进程。 +- `taskboard-electron/src/renderer/App.tsx` - React 渲染进程。 +- `skills/agentforge/` - 智能体委派任务的内置技能。 ## 许可证 -MIT —— 详见 [LICENSE](LICENSE)。 +MIT - 详见 [LICENSE](LICENSE)。 diff --git a/assets/readme-preview.svg b/assets/readme-preview.svg new file mode 100644 index 0000000..61d6997 --- /dev/null +++ b/assets/readme-preview.svg @@ -0,0 +1,170 @@ +<svg xmlns="http://www.w3.org/2000/svg" width="1600" height="930" viewBox="0 0 1600 930" role="img" aria-labelledby="title desc"> + <title id="title">AgentForge task board preview + A light macOS-style AgentForge interface with task, heartbeat, and skill tabs plus queue, running, and done columns. + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + AgentForge + Local-first agent orchestration board + + + + + 1 running / 3 queued + + + + + + Tasks + Heartbeats + Skills + + + + + Search + + + + New + + + + + + Total + 6 + + + + Queue + 3 + + + + Running + 1 + + + + Done + 2 + + + + + + + + + Queue + 3 + + + + Review scheduler drift + Check delayed and cron tasks after app sleep. + + scheduled + + codex + + + + Draft channel smoke test + Create a compact startup checklist. + + qa + + + + + + + + Running + 1 + + + + Refine trace output + Group stream-json events into checkpoints. + + trace + + renderer + + + + + + + + Done + 2 + + + + Ship skill sweep summary + Summarize promoted patterns and skills. + + completed + + skills + + + + + diff --git a/assets/screenshot.png b/assets/screenshot.png deleted file mode 100644 index 779e647..0000000 Binary files a/assets/screenshot.png and /dev/null differ