Is Reusability using OpenAI Assistant Vector Stores possible?

While researching regarding OpenAI Assistant Vector Stores, ChatGPT gave me this:

Reusability
Reusability refers to the ability to use the same piece of information or knowledge across multiple interactions or threads. It’s about efficiently using stored information to handle various queries without needing to duplicate effort or data.

Example:
Consider the same e-commerce customer support assistant. If a customer asks about shipping options in one thread, and later in a different thread asks about the same topic, the assistant can reuse the information from the vector store without needing to reprocess or fetch new data. The assistant has reusable knowledge on shipping options, which can be applied to multiple threads and contexts.

But I didn’t find any documentation around this. Is this really a feature of Embedding and Vector Stores in general? Or ChatGPT just hallucinated?

Unlike langchain, indeed there is no documentation present under OpenAI for “Vector Stores”. But in general, yes, the concept of reusability is there in Embedding and Vector stores as well.
Recently I was working in some project regarding text summarisation where the embedding and vector store part is kept so free that if any other related prospective LLM model need to be developed then just attributes of these parts need to be changed and minor tweaks will do the charm.

Hope it gives you some more insights on your research. Cheers!!

For embeddings, you only need to send the large search corpus once in chunks as a preliminary operation.

That knowledge can then be re-used in multiple ways, besides a single original embedding that searches against the knowledge.

I can download another’s $500 spent on embedding wiki data for research and do my own research on similarities and groupings without making another AI call at all.

The current vector stores are based on embeddings. The cost is paid upfront by OpenAI. To get search results, they don’t have to send any of the language data being searched across back to an AI, and you can have a thousand users receiving chat from an assistant informed by that, so it does embody the concept of reusability (while being locked within Assistants).

Thank you @_j and @shitlesh. Yes, that answers my question.

1 Like