⚠️ Work in ProgressThis project is currently under active development.
- A step-by-step tutorial and a Docker container for easy setup are coming soon.
- Currently, I'm still testing and refining the system to mitigate errors and improve performance speed.
An agentic AI chatbot that autonomously transforms Hugging Face daily papers into summaries and LinkedIn posts. Built with LangGraph for advanced conversational AI workflows.
| 📄 Automated Paper Summaries | 💼 LinkedIn Post Generation | 💬 General Chat Capabilities |
|---|---|---|
| - Fetch papers from HuggingFace Daily Papers by date | - Select any paper from your fetched list by number | - Ask questions to better understand papers and concepts |
| - Generates concise summaries for each paper | - Automatically crafts LinkedIn post for the paper | - Natural conversation flow for clarifications and discussions |
| - Saves summaries locally as Markdown files | - Context-aware responses throughout the conversation | |
| - Returns a numbered list of processed papers in the chat |
During this phase, the agent fetches the papers from HuggingFaceDailyPapers, creates a markdown summary (saved in a summary folder) for each fetched paper, and returns the list of the summarized papers to the user.
The time of completion of this task might be long. It depends on the number of summaries fetched, on their length, and on your GPU. Each summary is a call to the LLM.
Note: if no year is specified, the current year is taken into account
User: "Make me the summaries of the papers from the 25th of September"
Here the agent takes the summary generated for the specified paper requested, and creates a linkedin post from it.
User: "Can you make me a Linkedin post for the second one?"
You can as well have a general conversation with the ChatBot. This might help you understand difficult or new topics better while you are reading the summaries or the papers themselves.
User: "I don't know much about the Mixture of experts. Can you tell me more?"
Hug2Post is an agentic AI system that demonstrates autonomous behavior:
- 🎯 Goal-Oriented: Understands high-level user goals and breaks them down into executable steps
- 🔄 Multi-Step Reasoning: Autonomously orchestrates complex workflows (fetch → download → process → summarize → save)
- 🛠️ Tool Usage: Independently decides when and how to use tools (web scraping, file systems, databases)
- 🧠 Context Awareness: Maintains state across conversation turns to reference previous actions
- 🚦 Dynamic Routing: Uses LangGraph to intelligently route through different processing nodes based on intent
- ⚡ Autonomous Execution: Completes entire workflows without step-by-step human guidance
The project uses a graph-based conversational AI architecture powered by LangGraph:
- Intent Recognition: Identifies user intent (summarize papers, create LinkedIn post, general chat)
- State Management: Maintains conversation context across interactions
- Workflow Orchestration: Routes tasks through appropriate processing nodes
This project is licensed under the MIT License - see the LICENSE file for details.
- Hugging Face Daily Papers for daily paper feeds
- LangGraph for the graph-based LLM framework
- Ollama for local LLM inference