Skip to content

Current Limitations of ChromaSQL

  • Joins / multi-collection SQL. Execute separate queries if you need to merge results from multiple collections.
  • Aggregations / GROUP BY. Fetch the rows and aggregate in Python.
  • Streaming or async execution. The executor currently collects results eagerly. Wrap it yourself if you need incremental delivery.
  • Custom arithmetic in ORDER BY. Only distance, id, and metadata paths are supported today. Blend scores (e.g. 0.8 * SIM_SCORE + …) in your own post-processing if you need that behaviour.

If your workload depends on these gaps, we suggest that you build the missing behaviour in your application layer and follow the project changelog for roadmap updates.