Ability to do semantic caching with Azure Managed Redis

[robertopc1]

Add semantic caching for SQL query execution to Data API Builder. This feature stores results for previously executed queries and reuses them when future requests are the same—or semantically equivalent—by indexing queries using vector embeddings (Azure OpenAI) and performing similarity search in a vector-capable cache (Azure Managed Redis). This goes beyond traditional “exact text” query caching by enabling reuse across minor query variations (e.g., whitespace changes, reordered predicates, or equivalent filter phrasing), improving cache hit rate and reducing repeated work.

This is important for agents because agent-driven applications tend to generate many repetitive or near-duplicate queries while planning, retrying, or refining responses. Semantic caching reduces end-to-end latency, stabilizes performance under iterative agent loops, and lowers database and token/embedding costs by avoiding unnecessary query execution. The result is faster, cheaper, and more scalable agent workflows, especially when agents repeatedly explore the same entities, filters, and joins over short time windows.