JADE EcoSystem: Multi-Agent Chat with Local LLM

A distributed multi-agent system implemented in Java (JADE), integrated with an Artificial Intelligence model (Ollama/Python) for intelligent assistance, translations and moderation.

Main Features

1. ** P2P Chat:** Direct communication between agents (Alice, Bob) with GUI.
2. ** Artificial Intelligence (Local LLM):**

• Bridge Agent: Interface between JADE (ACL) and Python (HTTP).
• Capabilities: General Chatbot, Multilingual Translation, Content Moderation.

3. ** Administrator (AuditAgent):**

• Centralized monitoring of network activity.
• Kill Switch: Emergency shutdown of the entire platform (using AMSService).

4. ** Organized History:** Saving conversations.
5. ** Smart Launcher:** MainLauncher class automatically starts Python environment and JADE Containers.

Architecture and Protocols

The system complies with FIPA standards and uses the following flows:

1. Service Discovery

Agents use Directory Facilitator (DF) to find partners (chat-peer) and AI service (llm-service).

2. Chat Protocol (P2P)

• Sender: ChatUserAgent -> Receiver: ChatUserAgent
• Message Type: ACLMessage.INFORM
• Content: Plain text.

3. AI Protocol (Request-Response)

• Step 1: User -> LlmAgent (Type: REQUEST, Content: JSON {task, content, extra_param})
• Step 2: LlmAgent -> Python API (HTTP POST)
• Step 3: Python -> LlmAgent (HTTP Response JSON)
• Step 4: LlmAgent -> User (Type: INFORM, Content: Result Text)

4. Shutdown Protocol (Broadcast)

• Initiator: AuditAgent (Admin)
• Method: Query AMSService for the list of active agents.
• Message: ACLMessage.PROPAGATE with the content "SHUTDOWN".
• Effect: Agents close their GUI and call doDelete(). The platform is force-shut down.

---

Installation and Run Guide

Requirements:

• Java JDK 17+
• Python 3.10+
• Ollama installed and model downloaded (ollama pull gemma3:12b)

Step 1: Prepare Python

If you are running for the first time, install the dependencies in the python_microservice folder:

cd python_microservice
python -m venv venv
# Windows:
venv\Scripts\activate
# Linux/Mac:
source venv/bin/activate

pip install -r requirements.txt

Step 2: Launch the Application

No complex commands required. The project includes a Java Launcher.

• Open/Import the project in your IDE (IntelliJ / Eclipse).

• Run the MainLauncher.java class.

What happens:

1. The Python server starts in the background.

2. The JADE Main Container is launched.

3. The Admin (Audit), Alice and Bob windows appear.

---

Project Structure

• src/agents/ - JADE agent logic.

• src/gui/ - Graphical interfaces (Swing).

• src/utils/ - History and file manager.

• python_microservice/ - FastAPI + Pydantic microservice.

• Chat_Logs/ - (Auto-generated) Chat history.