CodeGenBot: AI-Powered Coding Assistant

CodeGenBot is a Streamlit-based AI chatbot that helps developers generate, explain, and refine code. It features:

• Multi-step Prompt Engineering: Enter your coding request → choose a technique → generate redrafts → finalize code.
• Multiple Prompting Techniques: Zero Shot, Few Shot, Chain-of-Thought, Least-to-Most, Self-Consistency, Reflective, Backtracking, and Verbalizer.
• Dark/Light Mode: Toggle a moon button in the top-right corner for dark theme.
• Follow-up Conversation: Continue chatting with the bot to refine or extend solutions—all responses are structured as valid JSON with "code" and "explanation" keys.

---

Table of Contents

1. Project Overview
2. Features
3. Installation & Setup
4. Usage
5. Prompting Techniques
6. Repository Structure
7. License
8. Contributing

---

Project Overview

This project was originally built as part of an Intro to LLM lab assignment. It uses the LangChain Google Generative AI integration (langchain_google_genai) to interact with Gemini-1.5-flash. The Streamlit app provides a user-friendly interface for prompt engineering, code generation, and iterative refinement.

---

Features

1. Multi-step Prompting Flow

   • Users provide an initial coding request, choose a prompting technique, see multiple redrafted prompts, and pick one to finalize the solution.

2. Dark/Light Mode

   • A “🌙 Dark Mode” toggle in the top-right corner lets you switch between dark mode (black background with white text) and light mode.

3. Technique Descriptions

   • An expandable section explains each prompting technique in plain language, making it accessible to non-technical users.

4. Follow-up Conversation

   • After the initial solution is generated, users can ask further questions or refine their requests. All responses are returned as valid JSON with "code" and "explanation".

5. Strict JSON Enforcement

   • Every response from CodeGenBot is enforced to follow a strict JSON structure for consistency and ease of parsing.

---

🚀 Installation & Setup

1. Clone the GitHub repository

git clone https://github.com/your-username/codegenbot.git
cd codegenbot

2. Create a virtual environment (recommended)

python -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate

3. Install dependencies

pip install -r requirements.txt

4. Set your Google API key

Option A: Export it in your shell

export GOOGLE_API_KEY="YOUR_API_KEY_HERE"

Option B: The app will prompt you for your API key if it is not already set.

5. Run the Streamlit app

streamlit run streamlit_app.py

Open http://localhost:8501 in your browser to use CodeGenBot.

---

💡 Usage

1. Enter your coding request:
   For example: “How do I write a function to reverse a list in Python?”

2. Choose a prompting technique:
   Select one (e.g., Zero Shot, Few Shot, Chain-of-Thought, etc.) from the dropdown.

3. Generate Redrafts:
   Click “Generate Redrafts” to view 2–3 rephrasings of your request with the selected technique.

4. Pick a Redraft:
   Choose your preferred redrafted prompt using radio buttons.

5. Generate Final Code:
   Click “Generate Final Code” to receive a JSON response containing:

   • "code": displayed in a Python code block.
   • "explanation": shown in plain text.

6. Follow-up Conversation:
   Use this section to ask further questions or make refinements. CodeGenBot uses the context to generate responses.

7. Dark/Light Mode:
   Use the “🌙 Dark Mode” checkbox at the top-right to toggle themes.

---

🧠 Prompting Techniques

Each technique adds a unique flavor to the model's response:

• Zero Shot: No examples provided—the model answers directly.
• Few Shot: A few examples guide the response format/style.
• Chain-of-Thought: The model "thinks aloud" step by step.
• Least-to-Most: Breaks the problem into smaller sub-problems.
• Self-Consistency: Generates multiple solutions and picks the most consistent one.
• Reflective: Generates an initial solution, then refines it.
• Backtracking: Corrects its initial answer by backtracking reasoning.
• Verbalizer: Structures the answer by explicitly verbalizing the reasoning.

---

📁 Repository Structure

codegenbot/
├── LICENSE
├── README.md
├── requirements.txt
├── .gitignore
├── streamlit_app.py         # Main entry point for the Streamlit app
└── notebooks/
    └── llm_lab_1.ipynb
└── src/
    ├── config.py            # LLM configuration, base prompts, and technique instructions/descriptions
    ├── prompts.py           # Functions for prompt redrafting and final code generation
    ├── utils.py             # Utility functions (e.g., conversation history management)
    ├── theme.py             # Functions to inject custom CSS for dark/light mode
    └── conversation.py      # Functions for handling follow-up conversation responses

• streamlit_app.py: Main Streamlit app with multi-step prompt flow.
• requirements.txt: Lists required Python packages.
• LICENSE: MIT License.
• notebooks/llm_lab_1.ipynb: Original lab notebook.

---

📜 License

This project is licensed under the MIT License. You are free to modify and distribute it as long as you include the original license.

---

🤝 Contributing

Contributions are welcome! Here's how:

1. Fork this repository.
2. Create a new branch for your feature or fix.
3. Make your changes and test them.
4. Submit a Pull Request with a clear explanation.

---

Contact

Created with ❤️ by Ojasva Goyal - feel free to contact me at ojasvagoyal9@gmail.com for any questions or feedback.

---

Happy Coding! 💻✨