I have used index.query , index.query_namespaces with all the metric (“cosine”, “euclidean”, “dotproduct”) but did not get any chunks related to order number 10779 .
chunks related to order number 10779 are present in pinecone database.
If i add some more content related to that query then it will retrieve that chunk
example: tell me about order number 10779 of white sand or i just type order detail of white sand it will then fetch the chunks.
it is not working with numeric data and i am using openai embeddings (model = text-embedding-3-large) for making chunks and for similarity search also
As you noticed, adding more context to your query returns results, because they are semantically related. Since you are expecting queries to contain a specific keywords, in this case an order ID, it sounds like what you are doing may be a better fit for a lexical/keyword search with a sparse index. You can read more about how to implement that here.
Another alternative would be to parse the order IDs into a metadata field and use metadata filtering as described here. This would give you records that are most semantically similar to a query vector AND that have an order ID metadata field with the value 10779.