Show HN: KelpMesh – pure SQL data transformation, no Jinja required¶
KelpMesh is an open-source alternative to dbt that uses pure SQL instead of Jinja-templated SQL.
No Jinja. No macros. Just SQL.
-- This is valid KelpMesh. Dependencies resolved by parsing the AST.
SELECT customer_id, SUM(amount) AS total FROM orders GROUP BY 1
Why this matters: dbt forced analysts to learn Jinja templating embedded inside SQL. This breaks IDE intellisense, AI coding assistants (Copilot/Cursor can't parse Jinja-templated SQL), SQL linters, and code review. Every {{ ref('model') }} and {{ config(...) }} block makes the SQL harder to read, review, and maintain.
KelpMesh's approach: dependency resolution via AST parsing. Your SQL is valid, portable SQL. Nothing more.
What ships in 0.2:
- Column-level lineage (free, not enterprise-locked)
- 6 warehouses: DuckDB, Snowflake, BigQuery, Postgres, Databricks, Microsoft Fabric
- Security suite: audit logging, RLS, column masking, data classification, PII erasure, secrets scanning, transparent encryption
- dbt migration:
KelpMesh import ./dbt-project --output ./KelpMesh-project - No telemetry (zero phone-home, enforced at code level)
- 27 CLI commands
License: Apache 2.0
https://github.com/RoyPulseAI/kelpmesh