🚀 AI-Powered LinkedIn Post Generator

Multi-Agent Content Automation System built with LangChain & LLMs

An intelligent AI agent that reads a topic, decides its category, and writes a publish-ready LinkedIn post — automatically.

Overview • How It Works • Features • Tech Stack • Quick Start • Demo • Roadmap • Contact

---

📌 Overview

Manually writing engaging, on-brand LinkedIn content every day doesn't scale. This project solves that with a multi-agent AI pipeline — instead of a single prompt-to-output call, it uses conditional agent routing to decide how a post should be written before it's written, resulting in more relevant, higher-quality content every time.

Give it a topic and a language, and the system:

1. Classifies the topic (Tech vs. General)
2. Routes it to a specialized writer agent trained for that voice
3. Returns a publish-ready post — formatted, on-tone, and CTA-driven

This project demonstrates practical, production-style agentic AI architecture — the same pattern used in real-world content, support, and workflow-automation tools.

---

🧠 How It Works

                     ┌────────────────────────────┐
                     │   User Input (Topic + Lang) │
                     └──────────────┬─────────────┘
                                    │
                                    ▼
                     ┌────────────────────────────┐
                     │     Agent 1 — Router        │
                     │  Classifies topic as        │
                     │     TECH or GENERAL         │
                     └──────────────┬─────────────┘
                                    │
                     ┌──────────────┴──────────────┐
                     ▼                              ▼
        ┌─────────────────────┐        ┌─────────────────────┐
        │   Agent 2            │        │   Agent 3            │
        │   Tech Writer         │        │   General Writer      │
        │   Professional,       │        │   Warm, mentor-like,  │
        │   industry-aware      │        │   relatable            │
        └──────────┬───────────┘        └──────────┬───────────┘
                     │                              │
                     └──────────────┬───────────────┘
                                    ▼
                     ┌────────────────────────────┐
                     │     Final LinkedIn Post      │
                     └────────────────────────────┘

This is a classic conditional routing pattern — the same architecture used in real enterprise AI agents that triage and dispatch tasks to specialized sub-agents.

---

✨ Features

• 🤖 Multi-Agent Architecture — Router + specialized writer agents, not a single generic prompt
• 🌐 Multi-Language Support — Generates posts in any language the user specifies (including Bangla)
• 🎯 Smart Topic Classification — Automatically detects Tech vs. General content and adapts tone
• ✍️ Publish-Ready Output — 2–4 paragraphs, natural emoji use, and a closing question/CTA for engagement
• 💾 Auto-Save to File — Exports generated posts to a UTF-8 text file for easy reuse
• 🔌 Pluggable LLM Backend — Powered by GitHub Models API, easily swappable with any OpenAI-compatible provider
• 🧩 Clean, Incremental Codebase — Versioned build (v1 → v4) showing clear architectural progression from a basic LLM call to a full routed agent system

---

🛠 Tech Stack

Layer	Technology
Agent Framework	LangChain
LLM Integration	LangChain OpenAI
Model Provider	GitHub Models API (openai/gpt-4.1-nano)
Config Management	python-dotenv
Language	Python 3.10+

---

📁 Project Structure

File	Description
v1_basic_agent.py	Baseline LangChain LLM call
v2_router_agent.py	Agent 1 — Classifies topic as Tech or General
v3_writer_agents.py	Agent 2 & 3 — Tech Writer and General Writer agents
v4_linkedin_generator.py	Final version — full conditional routing pipeline
save_output.py	Saves generated post to linkedin_output.txt (UTF-8)
requirements.txt	Project dependencies

---

⚡ Quick Start

1. Clone the repository

git clone https://github.com/delowarhossaincse63/linkedin-post-generator.git
cd linkedin-post-generator

2. Install dependencies

pip install -r requirements.txt

3. Configure environment variables

Create a .env file in the project root:

BASE_URL=https://models.github.ai/inference/v1
API_KEY=your-github-token-here
MODEL_NAME=openai/gpt-4.1-nano

4. Run the generator

python v4_linkedin_generator.py

5. (Optional) Save output to file

python save_output.py

---

🎬 Demo

Example 1 — Tech Topic (English)

Topic    : AI in Healthcare
Language : English
Category : TECH → routed to Tech Writer Agent

Example 2 — General Topic (Bangla)

Topic    : Remote Work Productivity
Language : Bangla
Category : GENERAL → routed to General Writer Agent

---

🗺 Roadmap

• Streamlit / web UI for non-technical users
• Direct LinkedIn API publishing integration
• Post scheduling & content calendar
• Hashtag & image suggestion agent
• Analytics agent (predict engagement before posting)

---

🤝 Let's Connect

I build practical, production-style AI agents and LLM-powered automation systems — from multi-agent pipelines like this one to full-stack AI integrations for businesses.

If you're looking to automate content, workflows, or customer interactions with AI, I'd love to talk.

📧 Email: your-email@example.com 💼 LinkedIn: linkedin.com/in/your-profile 🐙 GitHub: @delowarhossaincse63

---

📄 License

This project is licensed under the MIT License.

⭐ If you find this project useful, consider giving it a star — it helps a lot!