Apppublic

burcuuu/AIAgents

sourceHugging Faceupdated 1y agoView on Hugging Face

0likes

App README

Alfred Goes RAG

This project uses a lightweight, local 7B LLM (Mistral) to act as an intelligent, tool-using agent. Alfred can:

—Retrieve information from a provided guest list.
—Handle follow-up questions dynamically (even pronoun-based ones like "What's her email?").
—Search the web if the guest isn't known locally.
—Generate natural conversation responses.
—Propose conversation starters based on the guest's background.

All done locally — no paid APIs, no cloud magic, just your machine, some code, and pure willpower.

How to Run

1.Clone the repo:

bash

   git clone https://huggingface.co/spaces/burcuuu/AIAgents
   cd alfred-goes-rag
   
2. **Install dependencies**

pip install -r requirements.txt

1.Download a local LLM model
2.I used quantized mistral-7B. If you have a more powerful hardware, just download the model and set the modelpath inside alfredgoes_rag.py

1.Run Alfred

bash

   python alfred_goes_rag.py
   
---

## Notes

- Requires decent hardware for smooth local inference. For comparison, I used M4 Pro with 24 GB RAM and achieved around 20 tokens per second. 
- HuggingFace 🤗 spaces will need CPU settings unless you connect a GPU.
- Yes, Ada Lovelace is on the guest list 🤬. No, you can't email her.

---

## Credits

- Inspired by HuggingFace 🤗 AI Agent's Course: https://huggingface.co/learn/agents-course/unit3/agentic-rag/invitees
- Built with some love 💕, huge amounts of frustration 👹 and large amounts of coffee ☕️.
- Please do reach me at: bselcuk@cse.yeditepe.edu.tr. Maybe we can gossip about Alfred 😈. 

---