Contribution to Hugging Face 🤗 cookbook: Add a search agent cookbook

[ahnjj]

Hi @MartinEls,

I’m planning to create a new Hugging Face cookbook entry inspired by your excellent work. The topic would focus on building an AI search engine using DuckDB and smolagents.

I’d like to ask your permission to proceed kindly. I’ll make sure to credit you and reference your original notebook in the new content.
Would it be okay for me to move forward with this and open a PR?
Thank you in advance!

[MartinEls]

Sure go ahead! But I'm not really sure which notebook you're referencing.

[ahnjj]

Sure go ahead! But I'm not really sure which notebook you're referencing.

@MartinEls Thanks MartinEls! I've just uploaded this cookbook with reference :)

It's yet only available in Korean, but after it is merged, I will upload it in english too.

[ahnjj]

@MartinEls
I have referred to this notebook : Vector Search on Hugging Face with the Hub as Backend

This is the new cookbook in Korean version which has just been uploaded.
Thanks again for the inspiration!

[tonydzi]

Hi @ahnjj, @MartinEls — thanks for this thread, it's what pointed me at the gap.

Congrats on #332 landing! With that merged, agentic retrieval over Hub datasets is covered. I've opened #366 with a complementary English-language recipe on the layer above retrieval: a web search agent plus an adversarial verification step that checks the drafted answer against the retrieved snippets and fails safe ("insufficient evidence") when the support isn't clearly there. It runs fully locally on an open model (Qwen2.5-7B-Instruct, 4-bit GGUF, CPU, no API key) with smolagents' WebSearchTool, and is committed fully executed.

Both of you are credited in the notebook header — the recipe exists because of this thread.

@ahnjj, if you're still planning the English version of your vector-search agent, the two are designed not to overlap: yours is how to retrieve well, mine is whether to trust what came back. Happy to cross-link them, and equally happy for yours to land first.