🤖 QueryMate2

A local AI chatbot for company documents

Developed as part of a practical bachelor thesis with a focus on privacy-friendly document retrieval using RAG (Retrieval-Augmented Generation).

---

🔍 Features

• 📄 PDF Upload & Processing
  Extracts and splits content from uploaded PDFs

• 🧠 Vector-based Knowledge Retrieval
  Document search using ChromaDB

• 🔌 LLM Support (Local)
  Uses models from Ollama
  → e.g. mistral, llama3, gemma, nomic-embed-text, deepseek-r1

• 🧬 Embedding Backend Selection

  • HuggingFace (all-MiniLM-L6-v2, BAAI/bge-base-en-v1.5)
  • Ollama (nomic-embed-text, mxbai-embed-large)

• 💬 User-Friendly Streamlit Interface

  • LLM model selection
  • Uploaded files list + delete option
  • Manual index refresh
  • Contextual Q&A interface

• 📦 Privacy-First & Fully Local

  • Runs entirely offline, no cloud dependency

📊 Prerequisites

• Ollama must be installed and running locally

---

🧱 Project Structure

QueryMate2/
├── backend/
│   ├── chroma_index.py         # Chroma index & vector search
│   ├── ollama_client.py        # LLM requests via Ollama (API or library)
│   ├── embedding.py            # Select and initialize embedding method
│   ├── config.py               # Central configuration
│   └── logger.py               # Logging setup
├── frontend/
│   ├── ui.py                   # Main UI logic (Streamlit)
│   ├── sidebar.py              # Model selector, index actions
│   └── faq.py                  # Help/FAQ sidebar section
├── data/                       # Uploaded PDF documents
├── models/chroma_index/        # Persistent vector index
└── requirements.txt            # Python dependencies
              

---

🛠 Developer Notes

• Code is organized into backend/, frontend/, and tests/
• Logging is configured via backend/logger.py
• Embedding options (HuggingFace vs. Nomic) are set in backend/config.py or via CLI
• Use python backend/chroma_index.py --reset to rebuild the document index
• Run validation scripts from tests/ or backend/ollama_validation.py

---

⚙️ Installation

Running the Application

1. Clone the Repository:

git clone https://github.com/emacs45/querymate2.git
cd querymate2

2. Create a virtual environment

python3 -m venv venv
source venv/bin/activate # for Linux and macOS systems
venv/Scripts/activate # for Windows systems

3. Install dependencies

pip install -r requirements.txt

⚠️ If you're installing this project on Windows and encounter errors when running the command above, especially this error:

error: Microsoft Visual C++ 14.0 or greater is required.

It means your system is missing required C++ build tools for packages like chroma-hnswlib

✅ Solution

1. Install Microsoft C++ Build Tools:

   • Download here

2. Open the Installer and choose "Modify" on your current installation or start a new one

   • Go to the "Individual Components" tab and select:
   • MSVC v143 - C++ Build Tools
   • Windows 11 SDK (10.0.22621.0)
   • C++ CMake Tools for Windows

3. Alternatively, you can select the whole "Desktop development with C++" workload for simplicity

4. Reboot your system (recommended)

   • (Optional) Install virtualenv manually: If you encounter issues with python -m venv venv, you might need to install virtualenv first:

python -m pip install virtualenv

---

🚀 Getting Started

Launch the Web UI:

python3 run-streamlit.py

⚠️ After changing the embedding model, you must reset and reindex your PDFs manually:

python backend/chroma_index.py --reset

⚙️ Configuration (via config.py)

Variable	Description	Default value
OLLAMA_METHOD	Choose requests or library mode	library
OLLAMA_URL	API URL if using requests method	http://127.0.0.1:11434/api/generate
EMBEDDING_TYPE	Embedding backend: huggingface or nomic	huggingface
OLLAMA_MODEL	Default LLM model	mistral:latest

---

🧪 Sample Use Case

Q: “What changes are introduced in the latest software release?” QueryMate scans your internal documentation and provides a concise summary based on the extracted context.

📘 License

MIT License — Free to use for learning, research, or internal company purposes.

---

👨‍🎓 About the Project

This chatbot was developed as part of a Bachelor of Science in Business Informatics. The goal was to prototype an AI assistant for SMEs that runs locally, protects sensitive data, and helps support agents or employees retrieve knowledge from internal documents quickly.

---

🙌 Contributions Welcome

Found a bug? Have an idea? PRs and Issues are welcome!