Research / Private and observable AI
anySQL
Query LLM responses, agent traces and RAG pipelines with standard SQL.
Version 0.1.1Released 2026-11-05Python
The problem
Evidence of how an AI system behaves is split across model calls, agent traces, evaluations and retrieval logs, each in its own tool.
What's new
Logs every AI API call into six canonical tables so they can be joined with standard SQL: cost, quality and attack evidence in one query.
My role
Sole author: designed, built and maintain it.
Cite this
Sadayamuthu, K. (2026). anySQL: SQL Analytics for AI Systems [Computer software]. Zenodo. https://doi.org/10.5281/zenodo.23980548
