firstName

lastName

email

message

updatesConsent

Let’s talk... 💬

My Blog

Built a chatbot trained to explain what AI is to university-educated students.

# Introducing Ollama-ChatBot: A Multi-Agent AI Chatbot Built with FastAPI

Welcome to **Ollama-ChatBot**! This project is an open-source, FastAPI-based chatbot system that uses multiple intelligent agents to answer a wide range of questions. Whether you're looking for general knowledge, information about Concordia University's Computer Science admissions, or curious about artificial intelligence, this chatbot is designed to help.

Check out the demo video on YouTube:  
[![Ollama-ChatBot Demo](https://img.youtube.com/vi/H5o7eK-kLX0/0.jpg)](https://www.youtube.com/watch?v=H5o7eK-kLX0&t=10s&ab_channel=JaedenRotondo)

---

## What Is Ollama-ChatBot?

Ollama-ChatBot is a conversational AI assistant that intelligently routes user queries to specialized "agents." Each agent is tailored for a particular domain:

- **General Knowledge Agent:** Answers a broad range of factual and conversational questions.
- **Concordia CS Admissions Agent:** Provides answers specific to Concordia University's Computer Science program, leveraging a custom knowledge retriever and the university’s open data API.
- **AI Specialist Agent:** Explains AI concepts, neural networks, and machine learning topics in-depth.

The chatbot manages conversation context using vector embeddings and can retrieve knowledge from Wikipedia and other sources for up-to-date answers. Everything is accessible through a RESTful API and an interactive web frontend.

---

## Tools and Technologies Used

Here's a breakdown of the main tools and libraries that power Ollama-ChatBot:

- **[FastAPI](https://fastapi.tiangolo.com/):**  
  The web framework for serving the chatbot API. Fast, async, and easy to document.

- **[Ollama](https://ollama.ai/):**  
  For running local language models. Supports models like Mistral, which you can pull and serve locally.

- **[LangChain](https://python.langchain.com/):**  
  Used for agent orchestration and prompt management, specifically the `langchain_ollama` module for interfacing with Ollama models.

- **Python 3.8+:**  
  The backend is entirely built in Python for ease of development and cross-platform support.

- **Frontend:**  
  - **HTML/CSS/JavaScript** for the user interface.
  - Custom chat UI for a conversational experience.
  - Real-time communication with the FastAPI backend.

- **Vector Store & Embeddings:**  
  Used to store and retrieve conversation context, allowing the chatbot to maintain coherent, multi-turn conversations.

- **Knowledge Retrieval:**  
  - **Wikipedia API** for general facts.
  - **Concordia University Open Data API** for admissions and program-specific information.

---

## Project Structure

```
Ollama-ChatBot/
├── app/
│   ├── agents/        # Specialized agent logic
│   ├── core/          # Agent router and orchestration
│   ├── memory/        # Vector store and context manager
│   └── knowledge/     # Custom knowledge retrievers (Wikipedia, Concordia API)
├── frontend/          # Web UI (HTML/CSS/JS)
├── requirements.txt   # Python dependencies
└── run.py             # Entrypoint for FastAPI server
```

---

## How It Works

1. **Ask a Question:**  
   Use the web UI or send a POST request to `/chat` with your message.

2. **Router Agent:**  
   The system examines your query and routes it to the appropriate specialized agent.

3. **Context Management:**  
   The chatbot maintains context with vector embeddings, so it can handle multi-turn conversations and follow-ups.

4. **Knowledge Retrieval:**  
   Specialized agents pull information from Wikipedia, the Concordia API, or their internal knowledge.

5. **Response:**  
   The answer is returned via the API and displayed in the chat interface.

---

## Getting Started

1. **Clone the repository:**
   ```bash
   git clone https://github.com/JaedenRotondo/Ollama-ChatBot.git
   cd Ollama-ChatBot
   ```

2. **Set up your Python environment:**
   ```bash
   python -m venv .venv
   source .venv/bin/activate  # On Windows: .venv\Scripts\activate
   pip install -r requirements.txt
   ```

3. **Set up Ollama and pull the Mistral model:**
   ```bash
   ollama pull mistral
   ollama serve
   ```

4. **Run the FastAPI server:**
   ```bash
   python run.py
   ```

5. **Open the frontend (optional):**
   - Open `frontend/public/index.html` in your browser, or connect to `http://localhost:8000/docs` for API docs.

---

## Demo

Watch the full walkthrough and demo here:  
[YouTube: Ollama-ChatBot Demo by Jaeden Rotondo](https://www.youtube.com/watch?v=H5o7eK-kLX0&t=10s&ab_channel=JaedenRotondo)

---

## License

MIT License

---


Multi-Agent Ollama-ChatBot

An AI emergency assistant that analyzes text and image descriptions  developed at ConUHacks IX

## Inspiration

The inspiration behind **ResQLens** came from the need to improve emergency response times, especially in high-stress situations where users might struggle to call for help manually. By leveraging AI and hands-free technology, we sought to create a faster and more reliable way to communicate emergencies.

## What it does

**ResQLens** enables users to contact emergency services quickly through Metaglass. Users can send distress signals using voice commands or by taking pictures, which are then processed by AI to generate a clear and concise emergency report. This report is sent via WhatsApp to a backend server, which then relays the message to emergency responders.

## How we built it

*   **PHP** was used to develop the backend server that handles WhatsApp messages.

*   **WhatsApp API** was integrated to facilitate message relays.

*   **ChatGPT** was used for AI-powered summarization of emergency messages.

*   **Metaglass** served as the primary input device for capturing user commands and images.

*   The system was designed to work around Meta's private SDK by leveraging WhatsApp as a communication bridge.

## Challenges we ran into

*   **Lack of an official SDK for Metaglass**, which forced us to find a creative workaround using WhatsApp.

*   **Ensuring message delivery speed** and reliability to make the emergency system effective.

*   **Accurate AI summarization**, ensuring the messages sent to emergency responders were clear and useful.

*   **Integrating AI and WhatsApp** in a seamless and efficient way.

## Accomplishments that we're proud of

*   Successfully built a working prototype despite **Metaglass SDK limitations**.

*   Leveraged **AI to summarize emergencies**, reducing human error and improving response clarity.

*   Found a **practical way to use WhatsApp as a bridge** for emergency communication.

*   Developed a **scalable and adaptable** emergency response solution.

## What we learned

*   **Creative problem-solving** is key when working with hardware that lacks an official SDK.

*   **WhatsApp Business API** can be effectively used for emergency communication.

*   **AI can play a crucial role in emergency response**, making messages clearer and more actionable.

*   **Working as a team** helped us overcome multiple challenges and push forward to completion.

## What's next for ResQLens

*   **Expanding platform compatibility** beyond Metaglass.

*   **Direct integration with emergency services** for faster response times.

*   **Improving AI summarization** for more accurate and useful emergency reports.

*   **Developing an alternative SDK-independent solution** for better Metaglass integration.

*   **Exploring other AI-driven enhancements**, such as predictive emergency assistance.

## Built With

*   chatgpt

*   metaglass

*   openai

*   [php](https://devpost.com/software/built-with/php)

## Try it out

*   [ GitHub Repo](https://github.com/JaedenRotondo/ConUHacksIX)



ResQLens

React site for SuitedRanges.com You can check out Suited Ranges here: https://suitedranges.com/

Projects

Multi-Agent Ollama-ChatBot

ResQLens

Suited Web

Let’s talk... 💬