ChatOllama — Local LLM Chatbot UI (Powered by Ollama + Streamlit)

ChatGemma is a simple and modular chatbot UI built with Streamlit and Ollama, intended for offline, educational, or personal experiments only.

It allows multiple users to:

• Sign up / log in
• Create and manage multiple chat conversations
• Chat with a local Ollama-compatible model
• Persist chats to disk
• Clear or delete chats easily

---

Example Screenshot

---

Features

• Multi-user login and signup
• Conversation history per user
• Streamed responses from a local Ollama model
• Memory of the last N messages per chat
• Clear and delete chat options
• Modular, maintainable architecture

⚠️ Security Notice
This implementation is for educational and local use only — not secure for production.
Authentication is stored in plain text in a local JSON file (data/users.json).
You should integrate proper password hashing (e.g., bcrypt) and session management before any deployment.

---

Project Structure

chatollama/
│
├── app.py                # Main Streamlit app
├── config.py             # Configurations (model name, constants)
│
├── utils/
│   ├── storage.py        # Load/save user data
│   ├── chat_engine.py    # Model chat logic (Ollama API)
│   └── ui_components.py  # Login/signup + sidebar UI
│
└── data/
└── users.json        # Auto-created user database

---

Requirements

You’ll need:

• Python 3.10+
• uv (modern Python environment manager)
• Ollama installed and running locally
  (with any Ollama-compatible model)

---

Installation (using uv)

# Clone the repository
git clone https://github.com/<your-username>/chatollama.git
cd chatollama

# Create and sync environment
uv sync

# Run Ollama in the background
ollama serve

# Pull the a Ollama-compatible model (if not already installed)
ollama pull gemma3:270m

# Launch Streamlit app
uv run streamlit run app.py

---

Configuration

You can customize app behavior in config.py:

USER_DB_PATH = "data/users.json"
MODEL_NAME = "gemma3:270m"
LAST_N_MESSAGES_MEMORY = 5
APP_TITLE = "ChatGemma-270m"

To change models, just modify MODEL_NAME (e.g., "llama3:8b" or "mistral:7b").

---

Data Storage & Authentication

All user data and chat history are stored locally in data/users.json. Passwords are currently stored in plain text — this is not secure.

This design is intended for offline, educational, or personal experiments only. If deploying publicly, you must:

• Use password hashing (e.g., bcrypt, passlib)
• Add CSRF/session handling
• Secure data with encryption or a proper database

---

Extend or Modify

• Add new UI elements → utils/ui_components.py
• Add new model integrations → utils/chat_engine.py
• Adjust configuration or model name → config.py
• Modify data schema → utils/storage.py

---

Future Ideas

• Add retrieval-augmented generation (RAG) support
• Deploy proper authentication
• Add model selector (Gemma, LLaMA, Mistral, etc.)

License

MIT License