LLM Reasoning Playground

Overview

LLM Reasoning Playground is a Streamlit web application that lets you benchmark large‑language‑model reasoning strategies in real time.
The app supports:

• Direct prompting
• Chain‑of‑Thought (CoT)
• Self‑Consistency voting
• Tree‑of‑Thought (ToT) path generation with automatic model voting
• Reflexion loops with self‑critique and retry

It works with both OpenAI GPT‑3.5 and Groq Gemma‑9B‑IT models.

Features

Module	Purpose
Direct / CoT	Compare terse answers with step‑by‑step reasoning
Self‑Consistency	Generate n CoT paths and return the majority answer
Tree‑of‑Thought	Create three independent reasoning paths and let the LLM rank them
Reflexion Agent	Critique its own answer, then retry with injected feedback

Quick Start (Local)

git clone https://github.com/pranav-here/llm‑reasoning‑playground.git
cd llm‑reasoning‑playground
cp .env.example .env            # add your API keys
pip install -r requirements.txt
streamlit run app.py

Quick Start (Docker)

docker build -t llm-reasoning-playground .
docker run -p 8501:8501 --env-file .env llm-reasoning-playground

Environment Variables

OPENAI_API_KEY=<your-openai-key>
GROQ_API_KEY=<your-groq-key>

Project Structure

├── app.py            # Streamlit application
├── requirements.txt
├── Dockerfile
└── README.md

License

MIT