Haystack is quite a large project. So, unsurprisingly, the Haystack elves always maintain documentation. But, they’re also releasing new versions of Haystack regularly. To make it easier for people who use Haystack in production, they also version the documentation. So, if there’s something you need to learn about a specific version of Haystack, it’s best to switch to the documentation of that version!
In this challenge, Elf Silvano wanted to showcase how you can do metadata filtering to do RAG and question-answering on only a specific segment of your data in your database. However, he simply forgot to actually implement the filters. He’s prepared a Colab where the documents and their metadata are already indexed. Your task is to use a querying pipeline with metadata filtering. You need to specify a filtering condition to retrieve only documents that fulfill that condition. In this challenge, your task is to filter to the specific version of documentation that you want to do question-answering on, which is Haystack 2.0 🚀
🕊️ Hint: For this challenge, the
InMemoryBM25Retriever
component and thePipeline
are going to be very useful.
💜 Here is a Starter Colab
📚 Useful Docs: