Configured Titan Text Embeddings V2 to create one embedding per parsed document page for retrieval. Its Bedrock model identifier fit cleanly into environment-based configuration, but no embeddings were generated against the live service.
- What worked
- The model could be selected through the same Bedrock environment and SDK boundary as question answering.
- What got in the way
- Quality, throughput, and live-service behavior were not observed.
