Agentjw

<div align="center"> # Tool List ### Think • Plan • Execute • Reflect • Repair • Learn • Evolve • Scale []() []() []() [](https://python.org) [](LICENSE) </div> --- ## Overview **AgentJW** is an **AI Agent Platform** and **Autonomous Business Partner** built with a multi-agent architecture inspired by Claude Code. ### Core Capabilities | Component | Status | Description | |----------|--------|------------| | **12 Claude Code Patterns** | ✅ | The Loop, Planning, Sub-Agents, Agent Teams | | **Multi-LLM Fallback** | ✅ | OpenAI → OpenRouter → NVIDIA NIM → Ollama | | **Auto-Repair** | ✅ | Generalized Repair Engine | | **Planning Mode** | ✅ | Step-by-step execution | | **Sub-Agents** | ✅ | Fresh context per subtask | | **Agent Teams** | ✅ | Multi-agent collaboration | | **Worktree Isolation** | ✅ | Each agent has isolated workspace | | **Persistent Tasks** | ✅ | Disk-based task persistence | | **Context Compression** | ✅ | Auto-summarize long conversations | | **Knowledge on Demand** | ✅ | Load knowledge when needed | | **Background Tasks** | ✅ | Run slow ops in background | | **Semantic Router** | ✅ | LLM-based intent classification | --- ## Architecture — 12 Claude Code Patterns ``` ┌─────────────────────────────────────────────────────────────────────┐ │ AGENTJW │ ├─────────────────────────────────────────────────────────────────────┤ │ │ │ ┌─────────────────────────────────────────────────────────────────┐│ │ │ CORE LOOP ││ │ │ User → Brain (think_and_respond) → Action (execute_action) ││ │ └─────────────────────────────────────────────────────────────────┘│ │ │ │ │ ▼ │ │ ┌─────────────────────────────────────────────────────────────────┐│ │ │ PLANNING LAYER ││ │ │ Planning → Sub-Agents → Background Tasks → Persistent Tasks ││ │ └─────────────────────────────────────────────────────────────────┘│ │ │ │ │ ▼ │ │ ┌─────────────────────────────────────────────────────────────────┐│ │ │ TEAM LAYER ││ │ │ Agent Teams → Autonomous Agents → Worktree Isolation ││ │ └─────────────────────────────────────────────────────────────────┘│ │ │ │ │ ▼ │ │ ┌─────────────────────────────────────────────────────────────────┐│ │ │ LLM CHAIN ││ │ │ OpenAI → OpenRouter → NVIDIA NIM → Ollama (fallback) ││ │ └─────────────────────────────────────────────────────────────────┘│ │ │ └─────────────────────────────────────────────────────────────────────┘ ``` ### 12 Claude Code Patterns Implemented | # | Pattern | Status | File | |---|---------|--------|------| | 1 | The Loop | ✅ | `brain.py` | | 2 | Tool Dispatch | ✅ | `action_registry.py` | | 3 | Planning | ✅ | `planning.py` | | 4 | System Prompt Builder | ✅ | `system_prompt_builder.py` | | 5 | Sub-Agents | ✅ | `sub_agent.py` | | 6 | Persistent Tasks | ✅ | `persistent_tasks.py` | | 7 | Context Compression | ✅ | `context_compressor.py` | | 8 | Knowledge on Demand | ✅ | `knowledge_loader.py` | | 9 | Background Tasks | ✅ | `background_tasks.py` | | 10 | Agent Teams | ✅ | `agent_team.py` | | 11 | Autonomous Agents | ✅ | `agent_team.py` | | 12 | Worktree Isolation | ✅ | `agent_team.py` | --- ## Main Features ### Multi-LLM Fallback Chain | Layer | Provider | Status | |-------|----------|--------| | 1 | OpenAI | ✅ | | 2 | OpenRouter | ✅ | | 3 | NVIDIA NIM | ✅ | | 4 | Ollama (local) | ✅ | ### Agents | Agent | Function | Status | |-------|--------|--------| | **Planner** | Step-by-step planning | ✅ | | **Sub-Agent** | Fresh context per subtask | ✅ | | **Agent Teams** | Multi-agent collaboration | ✅ | | **Autonomous Agent** | Self-working without supervision | ✅ | | **Repair Agent** | Generalized repair engine | ✅ | ### Tools - `planning.py` — Planning mode with step tracking - `sub_agent.py` — Sub-agent with isolated context - `agent_team.py` — Agent teams + worktree isolation - `persistent_tasks.py` — Disk-based task persistence - `context_compressor.py` — Auto-summarize long conversations - `knowledge_loader.py` — Load knowledge when needed - `background_tasks.py` — Run slow ops in background --- ## Repository Structure ``` agentjw/ ├── core/ │ └── llm_client.py # Multi-LLM with fallback chain ├── sicuan/ │ ├── brain.py # Core agent loop │ ├── chat.py # Chat handler │ ├── core/ │ │ ├── planning.py # Planning mode │ │ ├── sub_agent.py # Sub-agent system │ │ ├── agent_team.py # Agent teams + worktree │ │ ├── persistent_tasks.py # Task persistence │ │ ├── context_compressor.py # Context compression │ │ ├── knowledge_loader.py # Knowledge on demand │ │ ├── background_tasks.py # Background tasks │ │ ├── system_prompt_builder.py # Dynamic prompt assembly │ │ ├── semantic_router.py # LLM-based routing │ │ └── generalized_repair.py # Auto-repair engine │ └── actions/ # 21 actions ├── agents/ │ └── orchestrator.py # Agent orchestrator ├── memory/ │ ├── workspaces/ # Per workspace data │ ├── users/ # User data │ └── worktrees/ # Worktree isolation └── projects/ └── godmeme_bot/ # Example project ``` --- ## Test Results | Test | Status | Description | |------|--------|------------| | Chat | ✅ | Successful response | | Auto-Repair | ✅ | Generalized repair engine | | Planning | ✅ | Step-by-step execution | | Sub-Agent | ✅ | Fresh context per subtask | | Agent Team | ✅ | Multi-agent collaboration | | Semantic Router | ✅ | LLM-based routing | | Memory Extraction | ✅ | JSON fallback | | Fallback Chain | ✅ | OpenAI → OpenRouter → NVIDIA NIM → Ollama | --- ## Installation ```bash git clone https://github.com/XdibsAI/Agentjw.git cd Agentjw python3 -m venv venv source venv/bin/activate pip install -r requirements.txt # Copy .env cp .env.example .env # Edit .env with your API keys ``` --- ▶ Run Telegram Bot (Production) ```bash sudo systemctl start sicuan-telegram.service ``` CLI Chat ```bash python3 main.py chat "hello" ``` Interactive CLI ```bash python3 main.py ``` --- 🛠️ Commands Telegram Commands Command Function /start Start bot, show user ID /status Status bot /metrics Metrics dashboard (owner only) User Commands Command Function @godmemeku_bot list project List project di workspace @godmemeku_bot auto-repair godmeme Auto-repair project @godmemeku_bot buat plan untuk ... Planning mode @godmemeku_bot status godmeme Cek status trading @godmemeku_bot analyze godmeme Analisis project --- 🔑 Environment Variables ```env # LLM OPENROUTER_API_KEY=sk-... OPENAI_API_KEY=sk-... OPENAI_BASE_URL=https://openrouter.ai/api/v1 # NVIDIA NIM (fallback) NVIDIA_NIM_API_KEY=... NVIDIA_NIM_BASE_URL=https://integrate.api.nvidia.com/v1 NVIDIA_NIM_MODEL=meta/llama-3.1-70b-instruct # Telegram TELEGRAM_BOT_TOKEN=... TELEGRAM_CHAT_ID=... OWNER_USER_ID=... # Platform MASTER_ENCRYPTION_KEY=... DEFAULT_PLAN=free ``` --- 📄 License MIT License --- <div align="center"> Built by XdibsAI AgentJW — Autonomous AI Agent Platform Production Ready 🚀 </div>