全能自进化AI Agent - 基于Ralph Wiggum模式，永不放弃
Project description
OpenAkita
Your Loyal and Reliable AI Companion
Philosophy • Features • Quick Start • Architecture • Multi-Agent
What is OpenAkita?
OpenAkita is a self-evolving AI assistant — your loyal and reliable companion in the digital world.
Like the Akita dog it's named after, OpenAkita embodies:
- 🤝 Loyal Companion — Always by your side, ready to help whenever you need
- 🧠 Grows With You — Remembers your preferences and becomes more helpful over time
- 💪 Reliable Partner — Commits to completing tasks, never gives up easily
- 🛡️ Trustworthy — Keeps your data safe and respects your privacy
OpenAkita is more than a tool — it's a partner that remembers you, understands you, and stands by you through every challenge.
Philosophy
1. Human-Centered
OpenAkita's core is serving people, not showcasing technology. We focus on:
- Understanding Intent: Not just executing commands, but understanding what you really want
- Proactive Communication: Asks when encountering problems, rather than guessing or failing
- Privacy Respect: Your data belongs to you, never misused
2. Continuous Evolution
OpenAkita can learn and evolve:
- Memory System: Remembers your preferences, habits, common operations
- Skill Extension: Automatically searches or generates new capabilities for new needs
- Experience Accumulation: Learns from each task, becomes more efficient
3. Reliable Execution
Once a task is assigned to OpenAkita:
- Persistent Completion: Won't give up due to minor errors
- Smart Retry: Analyzes failure reasons, tries different approaches
- Progress Saving: Long tasks support checkpoint recovery
4. Multi-Platform Collaboration
Through Multi-Agent architecture for efficient parallelism:
- Master-Worker Architecture: Master coordinates, Workers execute
- Smart Scheduling: Allocates resources based on task complexity
- Fault Recovery: Automatic detection and restart of failed nodes
Features
Basic Capabilities
| Feature | Description |
|---|---|
| Smart Dialogue | Multi-turn contextual conversation |
| Task Execution | Shell commands, file operations, network requests |
| Code Abilities | Write, debug, explain code |
| Knowledge Retrieval | Search web, GitHub, local documents |
Advanced Capabilities
| Feature | Description |
|---|---|
| Plan Mode | Intelligent multi-step task planning and execution |
| Skill System | Extensible skill library, supports customization |
| MCP Integration | Connect browsers, databases, external services |
| Scheduled Tasks | Set reminders, periodic tasks |
| User Profile | Learn your preferences, personalized service |
Plan Mode (Multi-Step Task Management)
For complex tasks requiring multiple steps, OpenAkita automatically activates Plan Mode to ensure reliable execution:
How It Works
┌─────────────────────────────────────────────────────────────┐
│ Plan Mode │
├─────────────────────────────────────────────────────────────┤
│ │
│ User Request │
│ ↓ │
│ ┌─────────────────────────────────────────────────────┐ │
│ │ Multi-Step Detection │ │
│ │ • 5+ action words (complex task) │ │
│ │ • 3+ actions + connectors ("then", "after") │ │
│ │ • Compound task type │ │
│ └─────────────────────────────────────────────────────┘ │
│ ↓ │
│ ┌─────────────────────────────────────────────────────┐ │
│ │ create_plan → Execute Steps → update_plan_step │ │
│ │ ↓ ↓ ↓ │ │
│ │ Plan Created Tool Calls Progress Tracked │ │
│ │ ↓ ↓ ↓ │ │
│ │ User Notified Real-time Status Updates │ │
│ └─────────────────────────────────────────────────────┘ │
│ ↓ │
│ complete_plan → Summary Report │
│ │
└─────────────────────────────────────────────────────────────┘
Features
| Feature | Description |
|---|---|
| Auto Detection | Automatically identifies multi-step tasks |
| Plan First | Forces plan creation before execution |
| Progress Tracking | Real-time status updates for each step |
| User Notification | Sends progress to IM channels |
| Persistent Plans | Saves plans as Markdown files |
| Step Dependencies | Supports step dependency chains |
Example
User: "Open Baidu, search for weather, and take a screenshot"
📋 Task Plan: Search weather on Baidu and screenshot
├─ 1. Launch browser and navigate to Baidu
├─ 2. Enter search keyword
├─ 3. Click search button
└─ 4. Take screenshot and send
Execution:
✅ [1/4] Launch browser - Done
✅ [2/4] Enter keyword - Done
✅ [3/4] Click search - Done
✅ [4/4] Screenshot - Done
🎉 Task completed! 4/4 steps successful.
Plan Tools
| Tool | Description |
|---|---|
create_plan |
Create execution plan with steps |
update_plan_step |
Update step status (pending/in_progress/completed/failed) |
get_plan_status |
Get current plan progress |
complete_plan |
Mark plan as completed, generate summary |
Self-Maintenance System
OpenAkita's "sleep cycle" — automatic maintenance during quiet hours:
| Feature | Schedule | Description |
|---|---|---|
| Memory Consolidation | 03:00 AM | Consolidate daily conversations into long-term memory |
| Self-Check & Repair | 04:00 AM | Analyze errors, auto-fix tool issues, generate reports |
| Task Retrospection | After long tasks | Review execution, extract lessons learned |
| Loop Detection | Real-time | Prevent infinite loops |
How It Works
┌─────────────────────────────────────────────────────────────┐
│ Self-Maintenance System │
├─────────────────────────────────────────────────────────────┤
│ │
│ ┌─────────────────┐ ┌─────────────────┐ │
│ │ Log System │ │ Memory System │ │
│ │ • File rotation│ │ • Consolidation│ │
│ │ • Auto cleanup │ │ • Deduplication│ │
│ │ • Error filter │ │ • Vector search│ │
│ └────────┬────────┘ └────────┬────────┘ │
│ │ │ │
│ └──────────┬───────────┘ │
│ ▼ │
│ ┌─────────────────────┐ │
│ │ Daily Self-Check │ │
│ │ • Analyze errors │ │
│ │ • LLM diagnosis │ │
│ │ • Auto-fix tools │ │
│ │ • Generate report │ │
│ └─────────────────────┘ │
│ │
│ Error Types: │
│ • Core (Brain/Memory/Scheduler) → Report only, no auto-fix │
│ • Tool (Shell/File/Web/MCP) → Auto-fix, self-test │
│ │
└─────────────────────────────────────────────────────────────┘
Configuration
# Log settings
LOG_LEVEL=INFO
LOG_DIR=logs
LOG_RETENTION_DAYS=30
# Self-check (system tasks, cannot be deleted, can be disabled)
# Runs at 03:00 (memory) and 04:00 (self-check) daily
Multi-Platform Support
| Platform | Status | Supported Message Types |
|---|---|---|
| CLI | ✅ Full Support | Text |
| Telegram | ✅ Full Support | Text, Voice, Image, File |
| Feishu | ✅ Full Support | Text, Voice, Image, File |
| WeCom | ⚠️ Implemented, Not Tested | - |
| DingTalk | ⚠️ Implemented, Not Tested | - |
| ⚠️ Implemented, Not Tested | - |
Quick Start
⚠️ Early Stage Project: This project is in early development. For faster deployment, we recommend using AI coding assistants like Cursor, Claude, or GitHub Copilot to help with setup and troubleshooting.
Requirements
- Python 3.11+
- LLM API Key (Anthropic, OpenAI-compatible, or other providers)
Recommended Models
| Model | Provider | Notes |
|---|---|---|
claude-sonnet-4-5-* |
Anthropic | Default, balanced |
claude-opus-4-5-* |
Anthropic | Most capable |
qwen3-max |
Alibaba | Strong Chinese support |
minimax-2.1 |
MiniMax | Good for dialogue |
kimi-2.5 |
Moonshot | Long-context capability |
💡 Tip: Enable "extended thinking" mode for complex tasks. Set model to
*-thinkingvariant (e.g.,claude-opus-4-5-20251101-thinking) for better reasoning.
Quick Install (Recommended)
One-line install with interactive setup wizard:
# Install from PyPI
pip install openakita
# Run setup wizard (interactive configuration)
openakita init
The setup wizard will guide you through:
- LLM API configuration (API key, model selection)
- IM channel setup (optional: Telegram, Feishu, etc.)
- Memory system configuration
- Directory structure creation
Or use the quick start script:
# Linux/macOS
curl -fsSL https://raw.githubusercontent.com/openakita/openakita/main/scripts/quickstart.sh | bash
# Windows (PowerShell)
irm https://raw.githubusercontent.com/openakita/openakita/main/scripts/quickstart.ps1 | iex
Manual Installation
# Clone repository
git clone https://github.com/openakita/openakita.git
cd openakita
# Create virtual environment
python -m venv venv
source venv/bin/activate # Windows: venv\Scripts\activate
# Install
pip install -e .
# Run setup wizard
openakita init
# Or configure manually
cp .env.example .env
# Edit .env, fill in your API keys
Run
# Interactive CLI
openakita
# Execute single task
openakita run "Write a Python calculator"
# Service mode (IM channels only)
openakita serve
# Check status
openakita status
Basic Configuration
# .env file
# Required
ANTHROPIC_API_KEY=your-api-key
# Optional: Custom API endpoint
ANTHROPIC_BASE_URL=https://api.anthropic.com
# Optional: Enable Telegram
TELEGRAM_ENABLED=true
TELEGRAM_BOT_TOKEN=your-bot-token
# Optional: Enable Multi-Agent
ORCHESTRATION_ENABLED=true
Architecture
Overall Architecture
┌─────────────────────────────────────────────────────────────────────────┐
│ OpenAkita │
├─────────────────────────────────────────────────────────────────────────┤
│ │
│ ┌──────────────────────── Identity Layer ──────────────────────┐ │
│ │ │ │
│ │ SOUL.md AGENT.md USER.md MEMORY.md │ │
│ │ (Values) (Behavior) (Profile) (Memory) │ │
│ │ │ │
│ └───────────────────────────────────────────────────────────────┘ │
│ │ │
│ ▼ │
│ ┌──────────────────────── Core Layer ──────────────────────────┐ │
│ │ │ │
│ │ ┌─────────┐ ┌──────────┐ ┌───────────────┐ │ │
│ │ │ Brain │ │ Identity │ │ Memory │ │ │
│ │ │ (LLM) │ │ (Self) │ │ (System) │ │ │
│ │ └─────────┘ └──────────┘ └───────────────┘ │ │
│ │ │ │
│ └───────────────────────────────────────────────────────────────┘ │
│ │ │
│ ▼ │
│ ┌──────────────────────── Tool Layer ──────────────────────────┐ │
│ │ │ │
│ │ ┌───────┐ ┌───────┐ ┌───────┐ ┌───────┐ │ │
│ │ │ Shell │ │ File │ │ Web │ │ MCP │ │ │
│ │ └───────┘ └───────┘ └───────┘ └───────┘ │ │
│ │ │ │
│ │ ┌───────────┐ ┌────────────┐ ┌─────────────┐ │ │
│ │ │ Skills │ │ Scheduler │ │ Evolution │ │ │
│ │ └───────────┘ └────────────┘ └─────────────┘ │ │
│ │ │ │
│ └───────────────────────────────────────────────────────────────┘ │
│ │ │
│ ▼ │
│ ┌──────────────────────── Channel Layer ───────────────────────┐ │
│ │ │ │
│ │ ┌─────┐ ┌──────────┐ ┌────────┐ ┌──────────┐ ┌────┐ │ │
│ │ │ CLI │ │ Telegram │ │ Feishu │ │ DingTalk │ │ QQ │ │ │
│ │ └─────┘ └──────────┘ └────────┘ └──────────┘ └────┘ │ │
│ │ │ │
│ └───────────────────────────────────────────────────────────────┘ │
│ │
└─────────────────────────────────────────────────────────────────────────┘
Core Components
| Component | Description |
|---|---|
| Brain | LLM interaction layer, supports multi-endpoint failover |
| Identity | Identity system, loads SOUL/AGENT/USER/MEMORY |
| Memory | Vector memory system, supports semantic retrieval |
| Skills | Skill system, supports dynamic loading and extension |
| Scheduler | Scheduled task scheduler |
| Channels | Multi-platform message channels |
Multi-Agent
When ORCHESTRATION_ENABLED=true is set, OpenAkita enters Multi-Agent collaboration mode:
Architecture
┌────────────────────────────────────────────────────────────────┐
│ Main Process │
│ │
│ ┌─────────┐ ┌──────────┐ ┌───────────┐ │
│ │ CLI │ │ Gateway │ │ Scheduler │ │
│ │(Command)│ │(IM Chan.)│ │ (Tasks) │ │
│ └────┬────┘ └────┬─────┘ └─────┬─────┘ │
│ │ │ │ │
│ └──────────────┼────────────────┘ │
│ ▼ │
│ ┌────────────────┐ │
│ │ MasterAgent │ │
│ │ (Coordinator) │ │
│ │ │ │
│ │ • Task Routing │ │
│ │ • Worker Mgmt │ │
│ │ • Health Check │ │
│ │ • Fault Recov. │ │
│ └───────┬────────┘ │
│ │ │
│ ┌───────┴────────┐ │
│ │ AgentBus │ │
│ │ (ZMQ Comm.) │ │
│ └───────┬────────┘ │
│ │ │
│ ┌───────┴────────┐ │
│ │ AgentRegistry │ │
│ │ (Registry) │ │
│ └────────────────┘ │
│ │
└────────────────────────────────────────────────────────────────┘
│
┌────────────┼────────────┐
▼ ▼ ▼
┌────────────┐ ┌────────────┐ ┌────────────┐
│ Worker 1 │ │ Worker 2 │ │ Worker N │
│ (Process) │ │ (Process) │ │ (Process) │
│ │ │ │ │ │
│ • Execute │ │ • Execute │ │ • Execute │
│ • Heartbeat│ │ • Heartbeat│ │ • Heartbeat│
│ • Return │ │ • Return │ │ • Return │
└────────────┘ └────────────┘ └────────────┘
Features
| Feature | Description |
|---|---|
| Smart Routing | Simple tasks local, complex tasks to Workers |
| Stateless Workers | Session history via messages, flexible scheduling |
| Shared Memory | All Workers use same memory storage |
| Fault Recovery | Heartbeat + automatic Worker restart |
| Dynamic Scaling | Auto-adjust Worker count based on load |
Configuration
# Enable Multi-Agent
ORCHESTRATION_ENABLED=true
# Worker count
ORCHESTRATION_MIN_WORKERS=1
ORCHESTRATION_MAX_WORKERS=5
# Heartbeat interval (seconds)
ORCHESTRATION_HEARTBEAT_INTERVAL=5
# ZMQ addresses
ORCHESTRATION_BUS_ADDRESS=tcp://127.0.0.1:5555
ORCHESTRATION_PUB_ADDRESS=tcp://127.0.0.1:5556
CLI Commands
# View Agent status
/agents
# View collaboration stats
/status
Project Structure
openakita/
├── identity/ # Identity configs
│ ├── SOUL.md # Values
│ ├── AGENT.md # Behavior rules
│ ├── USER.md # User profile
│ └── MEMORY.md # Working memory
├── src/openakita/
│ ├── core/ # Core modules
│ │ ├── agent.py # Agent main class
│ │ ├── brain.py # LLM interaction
│ │ ├── identity.py # Identity system
│ │ └── ralph.py # Task loop
│ ├── orchestration/ # Multi-Agent
│ │ ├── master.py # MasterAgent
│ │ ├── worker.py # WorkerAgent
│ │ ├── registry.py # Registry
│ │ ├── bus.py # ZMQ communication
│ │ └── monitor.py # Monitoring
│ ├── tools/ # Tool layer
│ ├── skills/ # Skill system
│ ├── channels/ # Message channels
│ ├── memory/ # Memory system
│ └── scheduler/ # Scheduled tasks
├── skills/ # Skills directory
├── data/ # Data storage
└── docs/ # Documentation
Documentation
| Document | Description |
|---|---|
| Quick Start | Installation and basic usage |
| Configuration | All configuration options |
| Skill System | Creating and using skills |
| MCP Integration | Connecting external services |
| IM Channels | Telegram/Feishu/DingTalk setup |
| Deployment | Production deployment |
Contributing
Contributions welcome! See Contributing Guide.
# Development environment
pip install -e ".[dev]"
# Run tests
pytest tests/ -v
# Code check
ruff check src/
mypy src/
Community
Join our community for help, discussions, and updates!
Instant Communication
 WeChat Group Scan to join (Chinese) |
💬 WeChat - Chinese community chat 🎮 Discord - Join Discord 🐦 X (Twitter) - @openakita 📧 Email - zacon365@gmail.com |
Project Resources
- 📖 Documentation - Complete guides
- 🐛 Issues - Bug reports & feature requests
- 💬 Discussions - Q&A and ideas
- ⭐ Star us - Show your support
Acknowledgments
- Anthropic Claude — LLM Engine
- browser-use — AI browser automation
- AGENTS.md Standard — Agent behavior specification
- ZeroMQ — Inter-process communication
License
MIT License - See LICENSE
OpenAkita — Your Loyal and Reliable AI Companion
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
openakita-1.2.6.tar.gz
(5.2 MB
view details)
Built Distribution
Filter files by name, interpreter, ABI, and platform.
If you're not sure about the file name format, learn more about wheel file names.
Copy a direct link to the current filters
openakita-1.2.6-py3-none-any.whl
(574.9 kB
view details)
File details
Details for the file openakita-1.2.6.tar.gz.
File metadata
- Download URL: openakita-1.2.6.tar.gz
- Upload date:
- Size: 5.2 MB
- Tags: Source
- Uploaded using Trusted Publishing? No
- Uploaded via: twine/6.2.0 CPython/3.11.14
File hashes
| Algorithm | Hash digest | |
|---|---|---|
| SHA256 |
0c60eaa760e08361b5ebfc653b5dc0f897b12a55dedd1715350aa3eb069be75e
|
|
| MD5 |
05e651adf58e878e23f5db815b30186a
|
|
| BLAKE2b-256 |
3d4f8fd8b0548d03102ac2db4562dfd70265a612e4767e13bbff7505d2f9f0e9
|
File details
Details for the file openakita-1.2.6-py3-none-any.whl.
File metadata
- Download URL: openakita-1.2.6-py3-none-any.whl
- Upload date:
- Size: 574.9 kB
- Tags: Python 3
- Uploaded using Trusted Publishing? No
- Uploaded via: twine/6.2.0 CPython/3.11.14
File hashes
| Algorithm | Hash digest | |
|---|---|---|
| SHA256 |
e6b804d6b4254f63c09908b8c6993db992fe34edc5979a31873b16eed5b1a9fe
|
|
| MD5 |
6ea20b5fed6cf7eeeb94264e8ca7101a
|
|
| BLAKE2b-256 |
7fda04a3af52d33bc2e84a2f480fe546f3f595014e71c931549a3410a3368106
|
