This project demonstrates applications showcasing interaction with a local Ollama instance 🦙 for various AI tasks using Python and LangChain.

This project implements a Retrieval-Augmented Generation (RAG) system using:

• A local Large Language Model (LLM) served by Ollama 🦙
• A Chroma vector database 💾 for document storage and retrieval
• A Gradio web interface 🌐 for interaction.

It allows you to upload your documents (PDF, TXT, MD), ask questions about their content, and receive answers generated by a local LLM, ensuring your data stays private 🔒. The system also keeps track of your chat history 🕒.

• 🔒 Local Processing: Uses Ollama to run LLMs locally, keeping your data private.
• 📤 Document Upload: Supports uploading PDF, TXT, and Markdown (.md) files.
• 💾 Vector Storage: Uses ChromaDB to store document embeddings for efficient retrieval.
• 💬 Chat Interface: Provides a user-friendly Gradio interface to chat with your documents.
• 📚 Source Citation: Displays the source documents used to generate the answer.
• 🔢 Token Count: Shows the number of tokens used in the user query and the assistant's response.
• 🕒 Chat History: Saves and loads conversation history (also stored in ChromaDB).
• ⚙️ Management: Allows listing stored documents, deleting all documents, and deleting chat history.

• Python 3.8+ 🐍
• pip (Python package installer) 📦
• Git (for cloning the repository) 🐙
• Ollama installed and running 🦙💨.

1. Clone the Repository: 💻

   git clone github.com:maurizioorani/python_ai.git

2. Create a Virtual Environment (Recommended): 🌱

   # Linux/macOS
   python3 -m venv venv
   source venv/bin/activate
   
   # Windows
   python -m venv venv
   .\venv\Scripts\activate

3. Create the Requirements File (RAG_requirements.txt): 📄 Create a file named RAG_requirements.txt in the project directory with the following content:

   gradio
   langchain
   langchain-chroma
   langchain-huggingface
   langchain-openai # Used for Ollama's OpenAI-compatible API
   tiktoken
   pypdf
   unstructured[md] # Installs markdown parsing extras for unstructured
   sentence-transformers # For HuggingFaceEmbeddings
   chromadb >= 0.4.20 # Ensure a recent version for compatibility
   ollama # Optional, but useful for managing Ollama via command line

   Note: You might need to adjust versions based on compatibility.

4. Install Dependencies: ⬇️ Ensure your virtual environment is active, then run:

   pip install -r RAG_requirements.txt

   Alternatively, if an environment.yml is provided for Conda:

   conda env create -f environment.yml # Or conda install --file environment.yml

5. Setup Ollama: 🦙

   • Install Ollama: Follow instructions on the Ollama website.
   • Run Ollama: Ensure the Ollama service is running (e.g., run the installed application or use ollama serve in the terminal).
   • Verify Ollama: Check if Ollama is running by opening a terminal and typing:

     ollama list

     This command should list your installed models if Ollama is running and accessible.

6. Pull the LLM Model: 🧠 The RAG script is configured by default to use llama3.2. Pull this model using the Ollama CLI:

   ollama pull llama3.2

   Wait for the download to complete. You can verify the model is installed using ollama list again. Note: If you wish to use a different model, update the model value within the Python script (likely in an LLM_CONFIG dictionary) and pull the corresponding model using ollama pull <new_model_name>.

Once the setup is complete and Ollama is running with the required model:

python RAG_example.py

Then, open your web browser and navigate to http://localhost:7860 🌐. You should see the Gradio interface where you can upload documents and interact with the RAG system.

Please note: For optimal performance in languages other than English, consider using larger models or those specifically trained for multilingual tasks, as smaller CPU-only models may produce unexpected results due to limited non-English training.

This project provides a more sophisticated Retrieval-Augmented Generation (RAG) system, enhanced with document management capabilities using metadata for precise file deletion. See its dedicated Readme for details.

A minimalist command-line interface for interacting with locally hosted Ollama models. Perfect for:
✅ Learning LLM interaction fundamentals
✅ Testing model responses
✅ Prototyping without UI complexity

Advanced web analysis tool featuring:

1. Website Scraping - Automatically harvest all hyperlinks
2. AI Analysis - Local Ollama-powered relevance scoring 🧠
3. Business Intelligence - Identifies core business sections:
   • About Us
   • Services/Products
   • Contact Information
   • Company Values

Perfect for competitive analysis and website mapping!

🦙 Ollama Running:

ollama serve  # Keep running in background