1. Raw client data is stored in Object Storage
2. Data is processed utilizing custom tools or data preparation and transformation features available in watsonx.ai or watsonx Asssitant.
Machine learning models create vector indexes/embeddings and store 'in-memory' watsonx.ai or Elasticsearch vector db. Alternatively
create and store keyword indexes in Watson Discovery.
3. User asks natural language question (Query)
4. Gen AI app routes question through watsonx Assistant (for conversational flow) or through watsonx.ai
5. watsonx Assistant or watsonx.ai queries the vector or keyword indexes to get the relevant content (Retrieval)
6. The received content is sent to LLM to generate a summary (Generation)
7. Generated summary response is sent to the requesting Gen AI app
8. Generated summarized response is received by the user (Response)