Mako Networks RAG: ChromaDB + Ollama

This project is a minimal Retrieval-Augmented Generation (RAG) pipeline for querying content from makonetworks.com using ChromaDB for vector search and Ollama for LLM-powered answers.

---

What does the script do?

The main script (main.py) performs the following steps:

1. Scrapes makonetworks.com using Playwright to collect up-to-date website content.
2. Splits the scraped text into manageable chunks for processing.
3. Embeds the content using Ollama embedding models (e.g., nomic-embed-text).
4. Stores and indexes embeddings in a local ChromaDB database for fast vector search.
5. Answers user queries by retrieving relevant context from ChromaDB and generating answers with an Ollama LLM (e.g., llama3 (llama3.2:1b required for k8) ).

---

Setup Instructions

Requirements

• Python 3.9+
• Ollama (for local embeddings/LLM)
  • Models: nomic-embed-text and llama3
• Playwright (for dynamic site scraping)

Installation

1. Clone the repository:

   git clone https://github.com/eflaatten/cxai-backend.git
   cd cxai-backend

2. Install dependencies:

   pip install -r requirements.txt
   python -m playwright install
   # Or manually:
   pip install langchain langchain-ollama chromadb playwright requests beautifulsoup4
   python -m playwright install

3. [Local Ollama Only] Pull required models:

   ollama pull nomic-embed-text
   ollama pull llama3/Preferred Model

---

The script will:

• Scrape makonetworks.com for content (using a headless browser)
• Chunk, embed, and index content with ChromaDB

---

Using a Remote Ollama Server

If your Ollama model is running remotely (e.g., in a Kubernetes cluster or on a remote server):

1. Edit the rag_query function in main.py:

   def rag_query(question, db, llm_model="llama3"):
       ...
       llm = OllamaLLM(
           model=llm_model,
           base_url="https://YOUR-OLLAMA-ENDPOINT"
       )
       return llm.invoke(prompt)

---

Testing Locally

1. Run

   uvicorn main:app --host 0.0.0.0 --port 8000

2. In Postman, make a POST request to http://0.0.0.0:8000/api/rag with the following JSON body:

{
    "question": "What is Mako Networks?"
}

---

You should receive a response similar to:

{
"choices": [
        {
            "message": {
                "content": "Mako Networks is another name for the Mako System."
            }
        }
    ]
}