Hey Guys,
Right now i am doing a testcase with a chatbot:
I have a Vectorstore with dosen of manuals of cars.
When i start a chat, i select the car brand and type for context.
Is there way to say to the assistant that file X in the vectordb is more related than the other manuals(other then the content brand/type similarities)? Or what should is the best practice for questions like this?
Yes! Similar to that, it would be a really nice feature to be able to segment the vector store using tags (or whatever) and dynamically (or even statically) adjust the priority of certain tags before/during retrieval.
For a practical example in our vector store we have a content A and content B and we want to boost content A since it’s overall more accurate. It would be nice to tag documents that belong to content A and then have the priorities of the tags adjusted to fine-tune the quality of the answers.
To me that sounds trivial to implement just a case of adjusting the ranking factor when sorting the results from the vector store.
How do I tag a document as mentioned here? Is this an option in the vector store files API?