For the past two months, I’ve been building a new open-source project called Support Mail Bot.
It’s a highly configurable auto-mail service that experiments with the latest LLM tooling:
- Llama and Faiss for vector search
- Retrieval-Augmented Generation (RAG) and embeddings
- Chain-of-Thought style reasoning
- Models in GGUF format to keep things lightweight and portable
The idea is simple: I wanted to explore how modern LLMs can be combined into practical agents while keeping the project fun and educational.
In the repository, you’ll find examples running on Mistral, DeepSeek, GPT, and several other open-source models.
🔗 Repo: github.com/jorgemartinezpizarro/smb
Resources explored during development
This project is part of my sabbatical learning journey—just a “for fun” exploration of LLMs in practice. If you’re curious, check out the repo and feel free to play around with it.