ML Detection & Auto-Configuration

Status: Active Audience: Team, AI agents, curious developers Purpose: Portal into the ML detection knowledge base

This section explains how Niamoto detects semantic column types, how the ML stack is trained and evaluated, and where to find the current research, experiments, and dated planning material.

Start here

• Overview: what the system does, why it exists, current scores, and the main limits.

• Branch Architecture: the hybrid pipeline, product priorities, and the role of autoresearch.

• Training & Evaluation Guide: the reproducible workflow from ml/data/silver to trained models and evaluation results.

If you want to…

• Understand the implementation

  • Overview

  • Branch Architecture

  • Current Training Sources

• Train or evaluate the stack

  • Training & Evaluation Guide

  • ml/README.md

• Understand current autoresearch rules

  • Autoresearch Surrogate Loop

  • ml/programmes/

• Review experiment history

  • Experiments index

• Review exploratory ideas or older planning docs

  • Research notes

  • Archive

• Find the latest dated plans

  • docs/plans/

Structure

Active reference

• Overview

• Branch Architecture

• Training & Evaluation Guide

• Current Training Sources

• Autoresearch Surrogate Loop

• ML Detection Dashboard

Research

• Research notes index

Experiments

• Experiments index

Archive

• Archive index

Related runtime and code-adjacent docs

• ml/README.md — operational commands and where to look in the offline ML workspace

• User Guide — how ML auto-detection plugs into the import workflow

• Reference — API-level documentation and config schemas