AIline documentation

Track ML experiments with reproducible lineage

AIline captures the exact code state, command, environment fingerprint, DVC linkage, and MLflow run behind every experiment. Start in two minutes, then dive into concepts, CLI usage, and internals.

Start quickstart Open CLI reference

Overview

AIline is user-first tracking with zero training-script rewrites.

  • Wrap your normal command with ailine track -- ....
  • Get a live lineage row plus MLflow link while the run is still in progress.
  • Capture snapshots automatically when the Git worktree is dirty.

Start here

Pick your path:

  • Quickstart for first-time setup and first tracked run.
  • Concepts for the mental model (lineage row, snapshot, link strategy).
  • CLI reference for command details and recipes.

Docs map

Quickstart

Install AIline, initialize the workspace, verify setup, and run your first tracked command.

Best for first-time users.

Concepts

Understand lineage rows, snapshot behavior, MLflow link strategies, and configuration boundaries.

Read this before scaling usage.

CLI reference

Command-by-command guide with copy-paste examples and common workflows for track and restore.

Daily-use reference.

Troubleshooting

Fix empty MLflow links, tracking URI mismatches, deleted-experiment errors, and restore blockers.

Use when something behaves unexpectedly.

Internals

Architecture, data model, plugin mechanics, reproducibility scope, and release mechanics for contributors.

For developers and maintainers.

First run in 3 commands

ailine init-workspace
ailine doctor
ailine track -- python train.py --epochs 5

Want UI dashboards? Run ailine serve to open AIline on http://127.0.0.1:5000 and MLflow on http://127.0.0.1:5001.

Contact