🎓 Educational AI Agent System

This project demonstrates a multi-agent educational platform built using the Fetch.ai uAgents framework. It features a modular architecture with a Student, Tutor, and Knowledge agent. A key feature is the AI Assessment agent, which leverages Google's Gemini AI to provide personalized learning recommendations. The system also uses the SingularityNET MeTTa knowledge graph for structured curriculum retrieval and the Chat Protocol for user interaction, ensuring compatibility with the ASI:One interface.

Project Overview

Agent Name	Role	Technology Used
Student Agent	User Interface (CLI): Handles session initiation, user input (subject/level choices, answers, AI diagnostics), and displays the Tutor's questions and feedback.	uagents, Chat Protocol
Tutor Agent	Central Coordinator: Manages the user's session state (history), routes queries to the Knowledge Agent, grades answers, and forwards diagnostic requests to the AI Assessment Agent.	uAgents, Chat Protocol, State Management
Knowledge Agent	Knowledge Base: Serves questions and answers based on specific Subject/Level queries by querying the structured curriculum.metta file using MeTTa.	uAgents, MeTTa (hyperon), Knowledge Graph
AI Assessment Agent	(Optional) AI Diagnostician: Receives user challenges from the Tutor, analyzes them, and provides a personalized subject/level recommendation.	uAgents, AI Integration (Google Gemini)

Required Submission Badges

The agents are categorized under Innovation Lab and are part of the Hackathon:

---

Setup and Running Instructions

This guide assumes you have Python 3.9+ and have initialized a virtual environment (venv).

1. Clone the Repository

git clone https://github.com/ayomideadeniran/edu-agent-platform
cd edu-agent-platform/src 

2. Install Dependencies

This project requires uagents for agent communication and hyperon for MeTTa integration:

pip install -r requirements.txt

3. Agent Addresses

The three agents communicate using pre-determined addresses, which are written to local files on startup. Please ensure the following files exist in the src directory with the most recently generated addresses:

• Tutor Agent Address: tutor_address.txt
• Knowledge Agent Address: knowledge_address.txt
• Student Agent Address: student_address.txt

For the judges' reference, the expected addresses for the agents are:

• Tutor Agent: agent1qfz0z6wc0ep8ser24qssdf3dtkln4lyzvvutr0zwj0kv0h74w6r8yhe7kyt
• Knowledge Agent: agent1q0n0gf3nm2mevkj6mm45cmjvm3sx23glx38sdmn4kjmw8xm4stn2q600dnq
• Student Agent: agent1qddgfypwrksthknutu7pxxk4mkp86uma7g0lrxp8ae3wdx472qjxyxhzgp5
• Knowledge Agent: agent1qtmnjcac90xddj67av0j3jumtrj8y6y69xtxslpt9e0uuszufnz3xxu0pky

4. Run Agents (In Order)

Run each agent in a separate terminal window, ensuring the venv is activated for all sessions:

Terminal	Command	Role
Terminal 1	python knowledge_agent.py	Starts the MeTTa knowledge base.
Terminal 2	python ai_assessment_agent.py	(Optional) Starts the AI diagnostician.
Terminal 3	python tutor_agent.py	Starts the central coordinator.
Terminal 4	python student_agent.py	Starts the user interface (CLI).

5. Interaction

Once all three agents are running, interact with the platform via Terminal 3 (student_agent.py). Follow the menu prompts to:

1. Select a subject (Math, History, Science).
2. Select a difficulty level (Beginner, Intermediate).
3. Answer the question.
4. Use the [0] Check My History option to view the Tutor Agent's grading and session history.

Use of ASI Alliance Technology

| Technology | Implementation in Project | | :--- | :--- | :--- | | uAgents Framework | Forms the foundation of the entire architecture. Used for all inter-agent communication, agent state management, and defining clear protocols. | | MeTTa Knowledge Graph | Integrated into the knowledge_agent.py to query the curriculum.metta file. We use a robust (question ...) fact structure for reliable, subject/level-based retrieval. | | Chat Protocol | Implemented in the student_agent.py and tutor_agent.py to handle all user interactions, ensuring the application is compatible with the Agentverse and ASI:One interface for seamless human-agent interaction. |

Quick demo (how to run a minimal end-to-end)

This quick demo shows the minimal steps to run the Student, Tutor and Knowledge agents and the Flask UI locally so a reviewer can verify a round-trip (send input → student ack → tutor/question → answer).

1. Open a terminal and activate the project virtual environment (from src/):

cd ~/Documents/Project/ASI-Agents-Track/edu-agent-platform/src
source venv/bin/activate

2. Start the Knowledge Agent (terminal 1):

# Terminal 1
python knowledge_agent.py

3. (Optional) Start the AI Assessment Agent (terminal 2) if you use the diagnostic flow:

# Terminal 2
python ai_assessment_agent.py

4. Start the Tutor Agent (terminal 3):

# Terminal 3
python tutor_agent.py

5. Start the Student Agent (terminal 4):

# Terminal 4
python student_agent.py

6. Start the Flask UI (separate terminal) and open the browser UI:

# Terminal 5
python src/app.py
# then open http://127.0.0.1:5000 in your browser

What to expect (happy path)

• The UI will accept your input and show an immediate acknowledgement ("[Student Ack] Received input: ...").
• The Student Agent terminal will display the tutor's question (the agent writes the question to its console and also stores it in a buffer exposed to the UI).
• The UI will show the tutor question (blue box) and, once you submit an answer, will show your answer (gray box) and the tutor's feedback (green for correct, red for incorrect).

Quick troubleshooting

• If the UI shows "Input cannot be empty": ensure the browser request is posting JSON and that Flask is running on port 5000. Use the browser DevTools Network tab to inspect the POST /submit_input payload.
• If Flask returns a 503 saying the Student Agent is not running, check that student_agent.py started without errors and is listening on port 8000.
• If you see {"error":"contents do not match envelope schema"} when posting to /submit, make sure the Flask app is configured to POST to http://127.0.0.1:8000/ui (the Student Agent exposes /ui and /submit compatibility handlers).
• If tutor responses are delayed, wait a few seconds — the UI polls for recent outputs for ~5–6s after submission. You can also re-submit or click Refresh to force another fetch.

Notes for reviewers

• The agent addresses are written to student_address.txt, tutor_address.txt and knowledge_address.txt on startup — include these in your review notes if needed.
• The code includes agent.include(..., publish_manifest=True) which prepares manifests for Agentverse publishing.

You typically do this step only once per project.

1. Navigate to your project directory:

   cd ~/Documents/Project/ASI-Agents-Track/edu-agent-platform/src

2. Create the environment: This command tells Python to create a virtual environment named venv inside your current directory.

   python -m venv venv

   • Note: If you are on an older system, you might need to use python3 instead of python.

---

2. Activating the Virtual Environment

You must do this step every time you open a new terminal session for your project.

To activate the environment:

source venv/bin/activate

How to Confirm Activation

After running the command above, your terminal prompt will change to show the name of the environment in parentheses. In your case, it will look like this:

(venv) knights@knights-VivoBook-ASUSLaptop-X420UA:~/Documents/Project/ASI-Agents-Track/edu-agent-platform/src$

Why Use venv?

A virtual environment ensures that the Python packages you install for your project (like uagents and hyperon) are isolated from the rest of your system's Python packages. This prevents conflicts and keeps your project dependencies clean.

---

3. Deactivating the Virtual Environment

When you are finished working on this project in your current terminal window, you can exit the virtual environment using the command:

deactivate

Environment variables and .env file

For local development you can store sensitive keys in a .env file at the project root (same directory that contains src/). The agents will load these values at startup when python-dotenv is installed.

Example .env content:

# Gemini API Key for the AI Assessment Agent
GEMINI_API_KEY="your_gemini_api_key_here"

# (Optional) Legacy name accepted by the code
GOOGLE_API_KEY="your_gemini_api_key_here"

Installation and usage tips:

- Install the loader library:

```bash
pip install python-dotenv

• Add .env to your .gitignore to avoid committing secrets:

.env

• The AI Assessment Agent prefers GEMINI_API_KEY but will fall back to GOOGLE_API_KEY if present. If no key is available, the agent uses a local mock fallback.

Your terminal prompt will return to its normal state, and any subsequent pip install commands will affect your global Python environment (which is usually what you want to avoid).