UM-AI-BOT

The application simulate a "Digital Twin" that can mimic a user's conversational style and provide responses to incoming messages

System Design

Data Collection

• User profile data is stored via UserProfile
• Communication preferences are captured through questionnaire responses in UserResponse

In Context Learning

style_analysis = """
        You are responding as a chatbot that matches this user's style. Their questionnaire showed:
        - They are casual and friendly: "{intro_style}"
        - They like talking to: "{convo_pref}"
        - They start conversations with: "{conv_starter}"
        Respond to this message briefly and naturally, matching their style: {message}
        Only provide the direct answer, no explanations, previous context needed.
"""

Generate response using Gemini API

model = genai.GenerativeModel("gemini-1.5-flash")
response = model.generate_content(style_analysis)

Improvements

• Implement Federated Learing to improve response accuracy based on users personality
• securely fine-tune large language models with private data using federated learning

INSTALLATION

Backend

python -m venv .venv

Windows

   .venv\Scripts\activate

Macos

   source .venv/bin/activate

cd Backend

pip install -r requirements.txt

cp .env.example .env

python manage.py runserver

Frontend

flutter pub get

• try sticking to chrome browsers

flutter run

Create SuperUser for Admin page

python manage.py createsuperuser

• Username (ex= test)
• Email (optional)
• Password (ex= test@123)
• Password confirmation

http://localhost:8000/admin/login/

Postman-API

Import um-bot.postman_collection.json in postman and test the api

Output

the llm will automatical answer the text but you also can generate a response edit it and send

License

MIT License

Copyright (c) 2024 UMBOT