Show HN: Long term memory agent for "real world" data work

Category: ai-ml

Tags: data-pipeline, dbt, llm-agents, data-profiling, knowledge-graph

Score: 6.8/10 (Innovation: 6, Technical: 7, Documentation: 8, Utility: 6)

Portia is an AI copilot for data work that connects to Snowflake, BigQuery, PostgreSQL, or local files, profiles tables, answers questions, and generates dbt-shaped SQL pipelines via recorded YAML specs. Its key innovation is a deterministic architecture where the LLM never reads raw data—all numbers come from code—and a knowledge graph tracks column lineage and join overlaps, making it verifiable and safe for large tables.

Target audience: data engineers, analytics engineers, backend developers working with data warehouses

Repository: https://github.com/Jad1908/portia · Python · AGPL-3.0 · 3 stars

View on Hacker News