CLI that produces an offline snapshot of a database (DDL, column- and relation-level statistics, values with frequencies), combines it with annotations written by you or an agent, and exposes it as MCP and as a web UI. Works with PG, MySQL, Snowflake, ClickHouse, Databricks, BigQuery.
How to try it:
uvx dbprint context \
arboretum.seedbank.germination_trial \
--purpose query \
--project https://github.com/jakubro/dbprint/blob/main/docs/format/v1/examples/production/.dbprint.yaml
Features:
-
Philosophy: snapshots are meant to sit next to your code and versioned so that you have a history of changes and agents can hit them when searching codebase.
-
Offline: no need to expose live credentials to agents.
-
CI: "dbprint check" can verify schema/data drift.
-
Annotations: complement statistics by documenting the meaning behind data. An agent can prepare them for you ("dbprint annotate").
-
Redaction: values can be masked/excluded/hashed either explicitly with FQNs or automatically based on inferred sensitivity and value shape.
-
Format: Snapshots are YAML and markdown, so greppable and usable by other tools.
Why I built this: SQL queries produced by agents were inaccurate, and the overall time to result was slow and costly because of finding the answers in codebase or in live database.
Comments URL: https://news.ycombinator.com/item?id=50305171
Points: 1