[Example]: Multimodal RAG with LanceDB

[AyushExel]

Multimodal RAG with LanceDB.

What this does:

• Fetches a product catalog from a live API (fakestoreapi.com).
• Embeds product descriptions using a CLIP model (clip-ViT-B-32) and stores them in LanceDB.
• Implements a RAG agent with a find_products tool that can:
  1. Perform semantic search (e.g., "something for the cold weather").
  2. Perform metadata filtering (e.g., "show me all electronics").
     (e.g., "a cool t-shirt" from the "men's clothing" category under $20).
• Generates and displays an image collage of the results for instant visual feedback.
• Creates logfire tracing dashboard if api key is set

Install dependencies:

    pip install lancedb sentence-transformers torch httpx pandas Pillow logfire[httpx]

Set your Google API key (for the agent's text generation):

    export GOOGLE_API_KEY=your_api_key_here

Usage:
First, build the product database from the live API

    python lancedb_multimodal.py build

# Then, ask for a recommendation:

    python lancedb_multimodal.py search "a cool t-shirt in men's clothing under 20 dollars"

[hyperlint-ai]

PR Change Summary

Added a new example for using LanceDB in a multimodal e-commerce context, enhancing the documentation for product cataloging and search functionalities.

• Introduced a new example for LanceDB multimodal e-commerce RAG
• Included installation instructions and usage examples for building and searching the product database
• Demonstrated vector search and object storage capabilities with LanceDB

Modified Files

• docs/dependencies.md

Added Files

• docs/examples/lancedb-multimodal.md

---

How can I customize these reviews?

Check out the Hyperlint AI Reviewer docs for more information on how to customize the review.

If you just want to ignore it on this PR, you can add the hyperlint-ignore label to the PR. Future changes won't trigger a Hyperlint review.

Note specifically for link checks, we only check the first 30 links in a file and we cache the results for several hours (for instance, if you just added a page, you might experience this). Our recommendation is to add hyperlint-ignore to the PR to ignore the link check for this PR.

[AyushExel]

I need some help with managing deps. It seems like lancedb is not being detected during test time even though I've added it to examples deps list.

[Kludex]

I need some help with managing deps. It seems like lancedb is not being detected during test time even though I've added it to examples deps list.

You need to add the dependency you need for the docs on the dev = in the root pyproject.toml.

[Kludex]

Why so many?

[samuelcolvin]

to be clear, we're happy to accept an example from you, but this is just a big middle finger to how we operate.

it's either saying:

• I don't care about any of your typing, linting, DX work
• or, we're special, we don't need to do this

Either lanceDB is typesafe, in which case remove these. or LanceDB is not type-safe in which case the example probably doesn't belong in our docs!

[AyushExel]

Added these to figure out the deps part while fighting CI.. Will revert back

[samuelcolvin]

to be clear, we're happy to accept an example from you, but this is just a big middle finger to how we operate.

it's either saying:

• I don't care about any of your typing, linting, DX work
• or, we're special, we don't need to do this

Either lanceDB is typesafe, in which case remove these. or LanceDB is not type-safe in which case the example probably doesn't belong in our docs!

[AyushExel]

Addressed specific feedbacks:

1. Moved lancedb out of separate dependency group to dev
2. Removed hotfixes for pyright - except a couple of places involving dynamic types, similar to https://github.com/pydantic/pydantic-ai/blob/main/examples/pydantic_ai_examples/rag.py#L189 . Not sure if there's a better practice. Happy to be educated.
3. Removed top-level docstring guide from code file - Might want to double check this, as the other examples have them.