Embedding Search Index

Steamship provides a built-in vector store and search interface. This can be used in combination with a Embedder to quickly create a search over text, images, or other content. To use the index, first create an instance of it with your preferred Embedder plugin (we recommend OpenAI):

index = steamship.use_plugin(
      "embedding-index",
      random_name(),
      config={"embedder":
          {
              "plugin_handle": "openai-embedder",
              "fetch_if_exists": True
          }
      }
  )

Once you’ve created the index and paired it with the embedder, Steamship will automatically vectorize content as it is inserted and queried.

Inserting Data

The unit of insertion in a search index is :ref:Tags. Embedding search is most useful on short-ish chunks of text like sentences or short paragraphs. If your text or content is in Blocks, you can create Tags over the spans that you wish to index and then insert them.

Querying Data

Once you’ve inserted content, you can query it using the query() method.