Skip to content

Capstone File Guide

Guide Fit

flowchart TD
  family["Reproducible Research"] --> program["Deep Dive DVC"]
  program --> pressure["A concrete learner or reviewer question"]
  pressure --> guide["Capstone File Guide"]
  guide --> next["Modules, capstone, and reference surfaces"]
flowchart TD
  question["Name the exact question you need answered"] --> skim["Skim only the sections that match that pressure"]
  skim --> crosscheck["Open the linked module, proof surface, or capstone route"]
  crosscheck --> next_move["Leave with one next decision, page, or command"]

Read the first diagram as a timing map: this guide is for a named pressure, not for wandering the whole course-book. Read the second diagram as the guide loop: arrive with a concrete question, use only the matching sections, then leave with one smaller and more honest next move.

This page explains which capstone files matter first and what responsibility each one holds.

Use it when the repository feels understandable at a directory level but not yet at a file level.


Start With These Files

File Why it matters
DVC Capstone Guide defines the repository contract and the proof questions it is trying to answer
Capstone Architecture Guide explains which files own declaration, execution, promotion, and verification
capstone/dvc.yaml declares the pipeline graph and stage boundaries
capstone/dvc.lock records executed state and declared evidence
capstone/params.yaml defines the parameter surface that controls comparable runs
capstone/Makefile exposes the learner-facing verification and recovery targets
Experiment Review Guide explains how to inspect comparable experiment changes without mutating the baseline story
Recovery Review Guide explains what the restore drill proves and what it does not prove
Release Review Guide explains how to review the promoted boundary as a downstream contract
DVC Capstone Guide explains the proof bundle generated for learners and reviewers
capstone/publish/v1/manifest.json demonstrates the promoted release evidence boundary

Back to top


Directory Responsibilities

Path Responsibility
capstone/data/raw/ committed source data used to begin the state story
capstone/data/derived/ generated intermediate data produced by the pipeline
capstone/src/incident_escalation_capstone/ implementation of preparation, fitting, evaluation, publication, and verification logic
capstone/metrics/ tracked evaluation outputs
capstone/models/ model artifacts produced by the pipeline
capstone/publish/v1/ promoted downstream contract and evidence bundle
capstone/state/ non-promoted but reviewable intermediate state surfaces
capstone/tests/ executable checks for code-level behavior

Back to top


Best Reading Order

  1. DVC Capstone Guide
  2. capstone/dvc.yaml
  3. capstone/dvc.lock
  4. capstone/params.yaml
  5. capstone/Makefile
  6. Experiment Review Guide, Recovery Review Guide, and Release Review Guide
  7. DVC Capstone Guide
  8. capstone/publish/v1/manifest.json

That order keeps the learner anchored in contract, then declared graph, then recorded state, then verification and promotion.

Back to top


Common Wrong Reading Order

Avoid starting with:

  • implementation files before reading the repository contract
  • promoted artifacts before understanding the baseline state story
  • dvc.lock before reading dvc.yaml

That route teaches fragments without context.

Back to top