AI Blog Writing Agent

An end-to-end AI-powered blog generation system built using LangGraph + Ollama + Streamlit. It automatically plans, researches (optional), writes, and formats high-quality technical blogs.

---

Project Strucutre

.
├── backend.py          # Agent pipeline (LangGraph)
├── frontend.py         # Streamlit frontend
├── blog_outputs/       # Generated Blog .md files
│   ├── blog1.pdf
│   ├── blog2.pdf
│   └── blog3.pdf
├── .env
└── README.md

---

Features

• Smart Routing

  • Decides if research is needed (closed_book, hybrid, open_book)

• Optional Web Research

  • Uses Tavily API for fresh and relevant content

• Structured Blog Planning

  • Generates a full outline (5–7 sections) with goals & word targets

• Parallel Section Writing

  • Each section is generated independently using LangGraph workers

• Markdown + HTML Output

  • Clean blog export with preview UI

• Persistence

  • Saves blogs locally and allows reloading past content

---

Architecture

User Input (Topic)
        ↓
     Router
        ↓
   Research (optional)
        ↓
   Orchestrator (Plan)
        ↓
   Fanout → Workers (parallel writing)
        ↓
     Reducer (merge)
        ↓
   Final Blog Output

---

Tech Stack

• LangGraph – workflow orchestration
• Ollama – local LLM
• Streamlit – UI
• Tavily API – web search

---

Setup

1. Install dependencies

pip install -r requirements.txt

2. Run Ollama

ollama run llama3.2:1b

3. Set environment variables (optional)

export OLLAMA_MODEL=llama3.2:1b
export TAVILY_API_KEY=your_key        

4. Run app

streamlit run frontend.py

---

TODO

• Add Image generation logic