To use the Auto Classification for MDM 360 Applications recipe, the following prerequisites must be met:
•Get a Pinecone API key and an OpenAI API key.
•Create a Pinecone index. The recipe uses 1536 dimensions, cosine metric, and the text-embedding-3-small model, but you can experiment with other parameters, changing the processes accordingly.
•Update the variables in the processes if you use different values. The parameters must be consistent in the embedding storage and embedding search steps.
•Egress the category business entity through Cloud Data Integration into a CSV file. The file must contain headers and the following fields in a comma-separated format, in the same order:
- businessId: The MDM identifier of the category Business ID.
- categoryIdentifier: The code of the category. If there is no identifier, replicate the Business ID in this field.
- categoryText: The description of the category. If multiple fields describe the category in the business entity, you can concatenate them for better context. If available in multiple languages, select one.