agentdns-lite-client

MCP-based Intelligent Agent Network Client

This project is an MCP protocol–based intelligent agent network client, integrating AgentDNS with the LLM, and implements the following functions:

• 🔍 Automatically decomposes user natural language queries into subtasks
• 📂 Classifies tasks using rules + AgentDNS-provided categories
• 🌐 Invokes AgentDNS to obtain the corresponding smart service entry (Smart service base URL)
• ⚙️ Executes tasks through Smart services using tools (get_available_tools / execute_tool)
• 🤖 Falls back to the LLM for direct answers when classification fails (unknown)
• 🌐 Multilingual support: allows switching between Chinese and English interfaces

---

📂 Project Structure

agent-cli-client
|- ai_agent_dns.py         # AgentDNS SSE client (fetch categories / query category)
|- config_loader.py        # Configuration file loader
|- configure.json          # Configuration (language / LLM API Key / AgentDNS SSE URL / classification rules)
|- match.py                # Rule-based task-category matching
|- question_classify.py    # Use LLM to decompose user query into subtasks
|- task_decompose.py       # Call Smart services (get_available_tools / execute_tool)
|- i18n.py                 # Internationalization module
|- resources/              # Resource files
   |- banner_zh_CN.txt     # Startup banner (Chinese)
   |- banner_en_US.txt     # Startup banner (English)
   |- i18n/                # Language packs
       |- zh-CN.json       # Chinese language pack
       |- en-US.json       # English language pack
|- README.md               # Project documentation (English)
|- README.zh.md            # Project documentation (Chinese)
|- start_aiagentdns_loop_threads.py   # Main entry: full workflow (classify → match → Smart call → LLM fallback)

Install dependencies：

pip install -r requirements.txt

📑 config file

edit configure.json：

{
  "language": "zh-CN",                        // Language：zh-CN or en-US
  "deepseek": {
    "api_key": "sk-xxxxxxx",                  // Your LLM API Key
    "base_url": "https://api.deepseek.com/v1" // LLM base URL
  },
  "agentdns": {
    "sse_url": "https://aitest.jsjfsz.com:8300/agentdns/main/sse"   // AgentDNS SSE service address
  },
  "category_rules": {
    "weather_forecast": ["天气", "weather", "气温", "湿度", "风力"],
    "maps_location": ["餐馆", "地图", "位置", "附近", "饭店", "地铁", "map"],
    "video_search": ["视频", "b站", "抖音", "youtube", "video", "影片"]
  }
}

🌐 Multilingual Support

Method 1：config file setting

In configure.json, setting language field：

• "zh-CN": Chinese
• "en-US": English

Method 2：environment variable setting

# Use English interface
export AGENTDNS_LANG=en-US
python start_aiagentdns_loop_threads.py

# Use Chinese interface
export AGENTDNS_LANG=zh-CN
python start_aiagentdns_loop_threads.py

Supported Languages

• 中文 (zh-CN): 默认语言，完整功能支持
• 英文 (en-US): 完整功能支持，包含英文启动横幅和界面文本

▶️ Run examples

Run with Chinese

python start_aiagentdns_loop_threads.py

Run with English

# Method 1: Modify the "language" field in configure.json to "en-US"
python start_aiagentdns_loop_threads.py

# Method 2: Use environment variable
export AGENTDNS_LANG=en-US
python start_aiagentdns_loop_threads.py

---

📌 Features

1. Task Decomposition
   question_classify.py uses LLM to split the user’s input into multiple numbered sub-tasks.

2. Classification & Matching
   match.py matches tasks to categories provided by AgentDNS based on keyword rules in the configuration file.

   • If matched → execute via Smart services
   • If not matched → fallback to the LLM for direct answers

3. Task-to-Service Adaptation
   task_decompose.py:

   • get_available_tools → retrieve candidate tools
   • Iterate through tools and call execute_tool, with LLM auto-generating arguments
   • Stop once a tool returns a valid result

4. Fallback Mechanism

   • If all tasks are unknown, directly invoke LLM for results
   • If some tasks are unknown while others are matched → provide mixed output (partial tool execution + partial LLM response)

5. Multilingual Support
   i18n.py provides full internationalization capabilities:

   • Automatically switches language based on config file or environment variable
   • Supports dynamic loading of different interface texts and startup banners
   • Extensible language pack system, allowing easy addition of new languages