VectorStore Reader
Last updated
Last updated
A Vector Store is a database designed for storing vector embeddings, which are numerical representations of human readable text. These numerical representations aim to capture the meaning of the corresponding texts. By comparing vectors, computers are able to compare texts based on their underlying meanings.
The FireKit vector store provides an interface for storing the text, its corresponding vector embedding, and the mapping between the two. You can configure a project's vector stores within the Databases page.
You can populate a vector store by uploading files through our interface. Each file will automatically be encoded into a vector embedding and both the file and the vector will be stored.
Once you have setup a vector store, you can incorporate it into any application within a project by dragging a VectorStore Reader block into your application.
To find files within a vector store that are relevant to a certain text, pass the text into the query of a VectorStore Reader block. The Top K most relevant results will be returned as the output in this format from most relevant to least relevant:
You can also optionally limit the number of results by a Token Limit. This is particularly useful for ensuring that the result of a reader does not exceed the token limit of a language model. You can read about limits .
