CookbooksLLM

Scenario:

• You have a collection of cookbooks, each one being a PDF or epub file several hundred pages long. Let’s say a couple hundreds of them. That is your knowledge base.
• You are cooking or planning dinner and you you want to prepare something interesting using what you have available in your fridge and pantry

The problem:

You want to be able to query exclusively and exhaustively this knowledge base with question that may be as simple as:

List all the recipes using kale in the knowledge base providing the source title, author, and page number

to more complex ones such as, for instance,

Provide a list of all recipes suitable as a main course that include a green vegetable similar to kale as one of the main ingredients, providing the source title, author, and page number

The constraints, in short:

You have a corpus of documents that are semantically fairly homogeneous and therefore all more or less relevant to the possible queries and you need the answers to be exhaustive, without resorting to extensive preprocessing of the knowledge base into a structured database

While LLMs are possibly optimally suited for the interface, the local nature of the knowledge base requires the query be carried out through other tools. The goal of this project is simply to assess the potential relevance of some tool categories for the problem at hand, such as RAGs, Semantic search, etc.