🤖 Code to Blog AI Agent

A Streamlit-based AI Agent built using LangGraph that transforms Python scripts into structured, human-readable blog posts. The agent uses LLMs to understand code, structure content, and iteratively draft each section, all with human-in-the-loop feedback.

---

Link to medium blog

---

🧠 What It Does

• 🔍 Understands uploaded Python code files
• 🧾 Generates a blog outline structure
• ✍️ Drafts each section individually using LLMs
• 🔁 Accepts human feedback on blog structure and drafts
• 📄 Outputs a final blog once all sections are approved

---

🔧 How It Works

This project uses LangGraph to model a multistep agent with interruptible nodes and memory:

Python Code → Code Summary → Blog Structure → Feedback Loop → Draft Sections → Feedback Loop → Final Blog

LangGraph Nodes:

• code_understanding: Summarizes uploaded Python code
• blog_structuring: Proposes an outline
• human_blog_feedback: Requests approval or revision of outline
• section_drafting: Generates individual section drafts
• section_drafting_feedback: Requests feedback on each section
• set_next_section: Tracks and selects the next section to draft

---

📁 File Structure

• main.py: Streamlit app interface
• blog_graph.py: LangGraph pipeline setup
• nodes.py: Core logic for summarization, structuring, drafting, and feedback handling
• blog_state.py: Typed state for LangGraph
• functions.py: Utilities (e.g., for loading code)
• requirements.txt: All required Python packages

---

📦 Installation

1. Clone the Repository

git clone https://github.com/nagi1995/blog-drafting-ai-agent.git
cd blog-drafting-ai-agent

2. Download and Install Miniconda

   Install Miniconda from the link.

3. Create a Conda Environment

conda create --name blog_drafting_ai_agent python=3.12.9 -y
conda activate blog_drafting_ai_agent  

4. Install Dependencies

pip install -r requirements.txt

5. Create a .env file with necessary credentials

GROQ_API_KEY=your_groq_api_key

---

🖥️ Interface Options

1. Streamlit UI

Launch the web interface with:

streamlit run main.py

Features:

• Upload .py files directly
• Review blog structure and draft sections
• Provide feedback via text input
• Automatically compiles final blog draft

2. CLI Agent

Run the agent in a terminal:

python blog_graph.py

---

▶️ Demo Video

demo.mp4

---

🧑‍💻 Author

Nagesh
GitHub | LinkedIn