Bender Enterprise v6.1 - Local RAG UI

Bender Enterprise is a lightweight, high-performance web interface designed for Local RAG (Retrieval-Augmented Generation). It allows you to chat with local LLMs via Ollama while providing the ability to upload documents for context-aware responses—all running 100% locally in your browser.

🚀 Key Features

• Multi-Format RAG Support: Upload and "shred" documents to use as context. Supported formats: .pdf, .docx, .xlsx, .xls, .txt, .csv.
• Multiple Chats Management: Create, rename, and switch between different conversation threads.
• Smart Message Handling:
  • Dynamic Controls: Messages cannot be sent unless text is typed.
  • Delete Individual Messages: Remove specific user queries or AI responses to fine-tune the conversation history.
  • Copy Chat History: Export the entire conversation to your clipboard with one click.
  • Context Window: Automatically remembers the last 10 messages to maintain coherent dialogue.
• Visual & UX Enhancements:
  • Light/Dark Mode: Smooth theme transitions saved to your browser preferences.
  • Generation State: UI intelligently locks (Sidebar and New Chat) while Bender is thinking to prevent state conflicts.
  • Bender Quotes: Randomized witty greetings upon startup.
• Privacy & Storage: Everything is stored in your browser's localStorage. No data ever leaves your machine (except to your local Ollama instance).

🛠 Configuration

You can easily customize the behavior in the JavaScript section of the file. Here is the default configuration:

const CONFIG = {
    OLLAMA_URL: 'http://localhost:11434/api/chat', 
    MODEL_NAME: 'llama3.1:8b',
    CONTEXT_WINDOW: 10,
    SYSTEM_PROMPT_DEFAULT: "You are Bender, a helpful robot assistant. Be precise and witty.",
    SYSTEM_PROMPT_DOCS: "You are Bender. FOCUS STRICTLY ON THE PROVIDED DOCUMENTS.",
    QUOTES: [
        "System ready. What are we solving today, meatbag?",
        "I'm faster than a vending machine coffee, and likely smarter.",
        "I'm here to do the thinking so you don't have to.",
        "Local LLM at your service. Just ask!",
        "I'm bored. Give me a complicated PDF to shred."
    ]
};

📂 Document Handling & Constraints

• Multiple Uploads: You can attach several documents to a single chat session.
• Storage Note: Extracted text is stored in localStorage.
• Recommended Limit: To ensure smooth browser performance and stay within standard browser storage quotas, it is recommended to keep the total combined text content under 5MB.

🏁 Quick Start

For the best and lowest-latency experience, it is highly recommended to run Ollama on the same machine as this UI.

1. Start Ollama Ensure Ollama is running. If you encounter CORS issues, set the environment variable: export OLLAMA_ORIGINS="*" and restart Ollama.

2. Run the Web Server Bender is a zero-dependency frontend. Run it using a simple one-liner:

Linux / macOS / Windows (with Python):

python3 -m http.server 8000

Then open your browser at http://localhost:8000.

---

Created with wit and metal. "Bite my shiny metal interface!"