LLM-Exceptions for IPython

In addition to the error and stack trace, get explanations and actionable steps to solve your errors directly in Jupyter Notebook/Google Colab. This python package gives the input and the error stack to an LLM to give you detailed explanations for debugging, right inside the output of the cell.

Spreed.chat was conceptualized

Alternate reality chatbot, where you can ask multiple questions to the same LLM response. Each question branches out into its own conversation history. You end up with a mind-map of the entire topic you wish to explore.

Frameworks used:

• Node.js
  Used as the runtime environment for the server-side application and building the RESTful API and handling backend operations.
• React
  Used to develop the frontend of the application and branching logic.
• MongoDB
  For storing user conversations and managing user data.
• Redis
  For storing user sessions.

Check out the beta version

Towards End-to-end Generation with Multi-agent Adversarial LLMs

Created the idea of having multiple LLMs going back and forth on a topic to improve each other.

Gradient Descent Visual Demo

Visual gradient descent algorithm for \((x-5)^2\), \(sin(x) \times 10 + (x-5)^2\), and a Saddle function \(cos(x) \times 10 + (x-5)^2\).