Semantic Cache
We're excited to unveil a significant update to our system with the launch of version 1.15.0: The all-new Semantic Cache. This feature significantly accelerates SQL generation while reducing token consumption, making your interactions quicker and more cost-effective.
How It Works:
Semantic Cache optimizes SQL query generation by intelligently reusing components from similar past queries. Whether it's finding tables, selecting columns, or constructing complex join graphs, this feature efficiently repurposes previously processed elements. This not only speeds up the process but also enhances accuracy.
Levels of Caching:
- Exact Match: Reuse results for identical questions. The questions might be phrased differently, but the intent is the same.
- Semantic Similarity: Apply components from semantically similar queries. Different filters, projections, limits, orders, but similar in structure.
- Partial Relevance: Integrate parts from previous queries that are applicable. Might be a different analysis, but the join graph / model is the same.
Benefits:
- Speed: Faster query processing by avoiding redundant steps.
- Efficiency: Uses fewer resources, saving on token costs.
- Accuracy: Improved precision in SQL generation through refined query handling.
User-Centric Optimization:
Semantic Cache works seamlessly—automatically analyzing your query history to populate and refine the cache. No setup or configuration needed. As more users engage with the system, it learns and adapts, enhancing its effectiveness. We pay close attention to users feedback to make sure the cache prioritizes the patterns users find the most useful.