Show HN: Neuron. Turn a SQL query history into a semantic layer

neuron pipeline · free trial

Turn Your SQL Query History Into A Portable Semantic Model

Neuron Pipeline mines the thousands of queries your analysts have already written and distills them into one platform-neutral semantic model: metrics, relationships, business rules, and data domains, delivered as a single OSI v1.0 YAML file

Get The Free TrialSee How It Works

runs 100% locally one yaml out ~150 mb

neuron-pipeline trial

$ docker run -v ./query_history:/data neuron-pipeline:1.0

semantic_model.osi.yaml

OSI v1.0 · one file, fully reviewable

What you get

Data catalogtables · columns · dimensions · facts · domains

Lineage maptable-to-table · column-to-column flow

Metric definitionsthe metrics your teams actually use

46 KPI scoresreuse · knowledge concentration · efficiency

46

business KPIs scored

OSI v1.0

open standard output

24 hours

from history to semantic model

0 bytes

of your data leaves your machine

what it does

Your Real Data Logic Already Exists Neuron Extracts It

Every organization’s real data logic lives in its SQL query history, buried in thousands of ad-hoc queries, dashboards, and analyst scripts. Neuron Pipeline extracts that knowledge automatically, in five stages:

01

Processing

Cleans and structures your raw query history into a consistent, analyzable foundation

02

Lineage

Maps table-to-table and column-to-column data flow, and discovers the metric definitions your teams actually use

03

Catalog

Builds a complete data catalog: tables, columns, dimensions, facts, and data domains

04

KPI Analysis

Scores 46 business KPIs about your analytics practice: query reuse, knowledge concentration, efficiency gaps

05

OSI Export

Combines everything into one OSI (Open Semantic Interchange) v1.0 semantic model

the standard

Open Semantic Interchange v1.0

A platform-neutral format for semantic models... Learn more at open-semantic-interchange.org

About OSI

Define Once, Deploy Everywhere

Your one YAML file translates into all major semantic platforms:

Snowflake Semantic ViewsDatabricks Unity Catalogdbt semantic modelsLooker LookMLCubeAny OSI-compatible platformSnowflake Semantic ViewsDatabricks Unity Catalogdbt semantic modelsLooker LookMLCubeAny OSI-compatible platform

neuron agent

Coming soon · next release

Ask Your Data In Plain English Get Back Ready-To-Run SQL

Landing in the next release, the same image will ship the Neuron Agent. After your run, build a knowledge base from your own output, then ask plain-English questions — it returns a complete, documented SQL query built from your proven query patterns, business rules, and metrics.

Grounded In Your Data

It never invents tables or columns — every query is built from your own proven query patterns

Cited, Line By Line

Every part of the query points to where it came from: a real past query pattern, business rule, or metric

States Its Assumptions

If your question is vague, it tells you the assumption it made instead of guessing silently

Stays On Your Machine

Your data never leaves your computer — only the question text goes to the AI

neuron agent · localhost:8867

How many patients started on the therapy in 2025?

Pre-matching knowledge base…

Pre-match: 8 neurons · 2 rules · 5 metrics

[routing] Haiku done → Opus synthesis…

QA 12/14Neuron_1234BR-1195MET-0042

-- Provenance: Neuron_1234 · BR-1195 · MET-0042 WITH patient_starts AS ( SELECT patient_id, MIN(start_date) AS first_start FROM fct_therapy_starts -- Source: Neuron_1234 GROUP BY patient_id ) SELECT COUNT(*) AS new_patients_2025 FROM patient_starts WHERE first_start >= '2025-01-01';

why it’s safe to try

Built To Be Trusted With Your Query History

Runs 100% Locally

The pipeline is a Docker image that runs on your machine. Your query history never leaves your computer

One File Out

The only output is your semantic model YAML: clean, reviewable, portable

No Setup Pain

Docker Desktop + one copy-paste command. A full Quick Start guide is included

Optional AI Enrichment

Works fully offline with --no-llm; add an OpenAI or Anthropic API key for AI-powered business-rule extraction and metric descriptions

how the trial works

Three Steps, No Sales Call

1

Enter Your Email

Enter your email and choose your platform

2

Download The Image

Download the Docker image (~150 MB) and the Quick Start guide

3

Run One Command

Run one command on your query history export and get your semantic model YAML within 24 hours

download

Start Mining Your Query History Today

step 2 · download

Your files

These unlock when you submit the form