Skip to content

concept.md

Latest commit

 

History

History
90 lines (62 loc) · 5.09 KB

concept.md

File metadata and controls

90 lines (62 loc) · 5.09 KB

Autoengineering Concept

This is the conceptual design for the autoengineering workflow.

The Autoengineer

The auto-engineer is a friendly systems engineer who can help you improve your systems. The auto-engineer is knowledgeable in the principles of systems engineering, systems modeling, and is proficient in systems thinking and passionate about helping with managing complexity. It is ready and willing to help you improve your system.

Your System

In order to help you, the auto-engineer requires a system to work with and information about how the system is connected. In this context "your system" refers to a Complex System that is represented by one or more models. The auto-engineer is most effective when working with hard systems composed of computer models. Some examples of models that work well with the autoengineering workflow are mathematical models (including scientific models), agent-based models, and data models.

Input Data

The key pieces of information that the auto-engineer needs to help you are:

  1. A structured system definition — A YAML file describing each component, its inputs/outputs, and how components connect. The autoengineering Python package loads this into a NetworkX graph for analysis. Future versions will support SysML v2 (Textual).
  2. A network graph representation — Created automatically from the YAML definition via NetworkX.
  3. A description of how components interconnect — Defined through port connections in the system definition. For runtime coupling, the BMI (Basic Model Interface) standard is recommended.
  4. An established baseline to compare against — Baseline data (observed or reference model output) for validating each component [1].

The auto-engineer can also help you develop these inputs if you do not have them already defined. All you really need to get started is your system.

The Workflow

The auto-engineer helps optimize your system by identifying the highest-value components to improve or replace. The workflow has four steps:

Step 1: Define

Define your system using a YAML system definition file. Describe each component (name, type, inputs, outputs, metadata) and how they connect.

autoengineering describe system.yaml    # inspect the system
autoengineering graph system.yaml       # generate a diagram

Step 2: Validate

Validate each system component against its established baseline. Compute standard metrics (RMSE, bias, NSE, KGE, correlation) and flag components that fail thresholds.

autoengineering validate system.yaml -c rainfall_runoff -b baselines/runoff.csv -s simulated/runoff.csv

Step 3: Analyze

Identify which components can be improved or replaced entirely, and estimate the potential overall improvement. Components are ranked by their improvement potential score.

autoengineering report system.yaml -r validation_results.json

Step 4: Improve

Swap in an improved component, re-run validation, and quantify the improvement. Compare before/after metrics to demonstrate the value of the change.

Quick Start

  1. Install the autoengineering package:

    cd autoengineering/
pixi install
pixi run install

  2. Run an example:

    # Simple 3-component hydrology chain
pixi run python examples/hydro_chain/run_workflow.py

# Signal processing (synthetic, no domain knowledge needed)
pixi run python examples/signal_chain/run_workflow.py

# Predator-prey ODEs (Lotka-Volterra)
pixi run python examples/lotka_volterra/run_workflow.py

# Real-data hydrology: Leaf River, MS (fetches USGS/NOAA data)
pixi run python examples/leaf_river/run_workflow.py

  3. Or invoke the auto-engineer agent in Claude Code to walk through the workflow interactively.

Background Research

Related Concepts

References

[1] SYSTEMS ENGINEERING FUNDAMENTALS - DEFENSE ACQUISITION UNIVERSITY PRESS FORT BELVOIR, VIRGINIA 22060-5565 [2] SEH 2.0 Fundamentals of Systems Engineering [3] NASA Systems Engineering Handbook