Codeninja: Code Assistant

Codeninja is a local code assistant powered by a custom model named (codeninja) hosted using ollama and interfaced with a Gradio web application. It enables you to interact with a locally running AI model for coding assistance.

Demo

Below is a preview of the Codeninja interface in action:

Features

• Simple interface for coding queries.
• Powered by codellama under the hood, named as a custom model (codeninja), created and hosted locally.
• Maintains a history of prompts for context-aware responses.
• Uses Gradio for a user-friendly web interface.

Prerequisites

Before setting up Codellama, ensure you have the following installed:

• Conda (for virtual environment setup)
• Ollama (for running the backend model)
• Python (compatible with the dependencies listed in requirements.txt)

Installation

1. Clone the repository:

   git clone https://github.com/SurajBhar/codeninja.git
   cd codeninja

2. Set up a virtual environment:

   conda create -n codellama python
   conda activate codellama

3. Install Dependencies: The application uses the following Python libraries:

• requests: For interacting with the backend API.
• json: For handling JSON data.
• gradio: For creating the web interface.

   pip install -r requirements.txt

4. Download the codellama:

• Make sure you have downloaded the Ollama.
• Go to terminal and run following command:

ollama run codellama

5. Set up the model: Go to terminal, change directory to codellama where modelfile is present and run:

   ollama create codeninja -f modelfile

6. Run the application: Make sure you are in the right directory codeninja and run:

   python app.py

Directory Structure

codeninja/
├── app.py               # Main application file
├── modelfile            # Model configuration file
├── requirements.txt     # Dependencies
└── README.md            # Project documentation
└── demo.png             # Project demonstration

Code Explanation:

app.py

• Backend URL: The application interacts with a locally hosted API (http://localhost:11434/api/generate) using requests.
• Custom Model: Requests are sent to the codeninja model hosted on Ollama.
• Prompt History: Maintains a history of prompts to provide context-aware responses.
• Gradio Interface: Creates a simple web interface with a text input and a text output for user interaction.

Key Functionality

generate_response(prompt):

• Takes a user prompt, appends it to the history for context, and sends a request to the backend API.
• Returns the AI-generated response or logs an error if the API request fails.

Gradio Interface:

• Input: A multi-line text box for user prompts.
• Output: A text display for the generated responses.
• Launch: Opens the interface on a local server.

Usage

• Enter your coding-related queries into the input box.
• View the AI-generated responses in real-time.
• The assistant retains a history of previous prompts for better contextual understanding.

Troubleshooting

• Model not running: Ensure the codeninja model is created and running via Ollama.
• API connection issues: Verify the backend URL (http://localhost:11434) is correct and accessible.
• Dependency issues: Check that all dependencies in requirements.txt are installed.

Contributing

Contributions are welcome! Feel free to fork this repository and submit pull requests for enhancements or bug fixes.

Acknowledgments

• Ollama for enabling local model hosting.
• Gradio for the interactive web interface.