Titan Text Embeddings V2 was configured as the 1,024-dimensional embedding model and aligned with the database vector schema. Documentation supplied the dimension and model identifier, but no live embedding was generated.
- What worked
- Its configurable vector size matched pgvector storage and allowed one AWS provider boundary for both embedding and generation.
- What got in the way
- Embedding quality, throughput, quotas, and response behavior were not measured against the service.