> ## Documentation Index
> Fetch the complete documentation index at: https://aidflow.aigensys.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Overview

Transform raw data into production-ready scorecards with absolute reproducibility and rigorous stability.

# Automated Interactive Development Framework

AIDflow is a pipeline-first framework for *binary classification scorecards* in regulated environments. Built for industries where model governance, temporal stability, and regulatory transparency are non-negotiable.

Framework where a human directs and an AI agent executes the full pipeline — stability-first, audit-ready, deployable. No notebook spaghetti. No 6-week committee cycles.

*Built over 15 years of credit scoring, fintech R\&D, and regulatory practice.*

![AIDflow Preview](https://mintcdn.com/aigensys/5XrnslzLpWDUfjAk/assets/aidflow-preview.gif?s=7c34173263a5423465c41693366223a7)

***

## Who This Is For

* **AI Orchestration Endpoints** — one command surface that lets LLM
  agents drive the full pipeline (genesis → training → champion) and
  return a deployable scoring model without human intervention.
* **Data Scientists** building scorecards in Fintech, Banking, or Insurance
  who need full control over every transformation step.
* **Model Risk Teams** who require complete audit trails, reproducible
  execution, and standardized reporting artifacts.
* **ML Engineers** who need a production scoring artifact with zero
  framework dependency at inference time.

If your scoring model must survive a committee review, AIDflow is your development environment.

***

## Why AIDflow?

* **Human-in-the-Loop.** Every phase is a decision point. Pause, inspect,
  re-configure, re-run — without losing state. All in an interactive UI.

* **Context-Driven.** A single `ProjectContext` links every stage from
  ingestion to champion export. No fragile variable chains.

* **Stability-First + Ensemble.** Temporal screening runs *before* training.
  Four orthogonal agents compete on stabilized data. An objective scoring
  system promotes the champion.

* **Full Feature Lineage.** Trace any data/feature from raw origin to final score.
  Every intermediate state is registered and navigable by stage.

* **Modular Extensibility.** Add a new fitter in minutes. It auto-propagates
  to the UI, panels, and reporting — no manual wiring.

* **Self-Contained Scoring.** The exported artifact encapsulates all
  encoding and model state. Inference requires no training data or prior
  pipeline state.

* **Audit-Ready Governance.** Every production artifact is packaged with full
  lineage, SHAP explainability vectors, and performance benchmarks for seamless
  regulatory approval.

***

## Get Started in Seconds

### Install & Run AIDflow

```bash theme={"theme":{"light":"dracula-soft","dark":"dracula"}}
# Pure CLI
pip install aidflow
aidflow run demo/

# Interactive UI 
pip install aidflow[notebook]
aidflow notebook ui_demo/
```

***

### The Golden Path to Production

AIDflow guides you through a standardized five-phase journey:

1. **Data Phase**: Ingest → Clean → Stabilize
2. **Encoding Phase**: Numerical → Categorical → Interaction → WOE Encoding
3. **Training Phase**: Train an ensemble of 4 specialized ML Agents
4. **Selection Phase**: Predict → Compare → Transition → Promote Champion
5. **Inference Phase**: Prediction Tests → Production Export

[Read the Developer Handbook](https://aidflow.aigensys.com/getting-started/core-concepts.md)


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.