Getting started

This walkthrough uses a minimal synthetic project and does not require timestamps, external infrastructure, or GetDone.

1. Initialize

bash automo init --root demo cd demo automo validate automo doctor

The generated project contains automo.toml and automo_project.py. The initial plugin is intentionally empty but valid.

2. Choose the smallest example that matches your use case

For ordinary tabular fitting:

bash python ../examples/basic-supervised/example.py

For a custom training package or service:

bash python ../examples/custom-trainer/example.py

For stacking/meta-models:

bash python ../examples/meta-model/example.py

For data with IDs but no datetime column:

bash python ../examples/id-only-data/example.py

3. Build the project plugin

A project plugin contributes domain-specific objects while Automo owns the surrounding lifecycle:

```python from automo import ResearchPlugin

def create_plugin() -> ResearchPlugin: return ResearchPlugin( id="my-project", data_sources=(...), feature_computers=(...), feature_sets=(...), objectives=(...), metrics=(...), model_specs=(...), model_runners=(...), model_trainers=(...), evaluators=(...), services={...}, ) ```

Use ModelTrainer when fitting needs full control or an external package. Use ModelOutputInput in a ModelGraphSpec when a model consumes another model's output.

4. Validate before research

bash automo validate automo doctor

Then use the workflow that matches the task:

bash automo research --help automo refresh --help automo models --help

Read trainers-and-graphs.md before implementing a downstream/meta-model so the cross-fitting boundary is clear.

Define the first research hypothesis

A fresh project has no active scientific claim. Define the relevant model structure and a falsifiable hypothesis before governed automated research:

```bash automo research model-add baseline --role standalone automo guidance --task-class hypothesis-planning automo research hypothesis-create H-ROOT \ --statement "The system provides useful out-of-sample predictive value." \ --primary-model baseline \ --evaluation-depth system automo research hypothesis-activate H-ROOT

Once a trained artifact exists, register it as a logical model candidate.

automo research candidate-add BASE-v1 --model baseline --model-spec-id baseline-v1 --artifact-id MODEL-BASE-v1 automo research candidate-select BASE-v1 ```

Project roadmap and milestone direction should remain in GetDone or the project's own planning system. See Research governance and Agent guidance.

Project-specific research guidance

Keep mutable research state under .automo/ and project-owned agent instructions under .project-agent/automo/. automo guidance discovers .project-agent/automo/index.json additively by default; use --no-project-agent for canonical Automo-only guidance. Pin a reviewed composition with automo guidance-lock --write and verify it with automo guidance-check. Project guidance may strengthen or specialize research rules but cannot replace protected sealed-OOS, bounded-search, evidence, or hypothesis-governance rules.