AGILab Documentation

AGILAB turns experimental AI/ML notebooks and scripts into executable, portable, evidence-backed applications that can run locally or on distributed workers. Workflows stay portable: export them back to runnable agi-core notebooks, keep reproducibility evidence, and hand off tracking evidence to MLflow when that integration is enabled. The notebook export is an agi-core runtime handoff: you can continue to run the saved project and stage contract with only the stable core runtime, without depending on the AGILAB UI or distributed worker layer. That stable, production-grade core technology remains the smallest supported handoff surface.

If you are new to AGILab, choose one route first:

  • See the UI now: open AGILAB Demo for the public Hugging Face Space.

  • Prove it locally: follow Quick-Start with the built-in flight_telemetry_project or start from a notebook through PROJECT. Default target: pass the flight first proof in 10 minutes.

  • Use the API/notebook: follow agi-core Demo for the smaller AgiEnv / AGI.run(...) surface.

The fastest adoption ladder is browser preview, one local first-proof lane, evidence manifest, then expansion into package mode, external apps, or cluster work.

Golden ML loop

AGILAB’s strongest workflow is deliberately evidence-first:

  1. Import or create a notebook, script, or app project.

  2. Run it through a controlled local environment first.

  3. Capture artifacts plus run_manifest.json evidence.

  4. Inspect and compare outputs in ANALYSIS, notebook export, MLflow handoff, or proof-pack tools.

  5. Promote the result only when the evidence, package contract, and release proof are coherent.

That loop keeps exploratory ML work replayable before a team moves it into a heavier tracker, registry, cluster, or production platform.

AGILAB is not locked to one web frontend. App projects can declare app-owned UI surfaces so the same runtime, artifacts, and evidence contract can be opened through the local Streamlit UI, a hosted Hugging Face backend, or browser-native agi-web UI islands with React-ready component contracts. See Page bundles for the [app_surface] contract and the generic agilab app surface launcher.

If the local first proof fails, use Newcomer First-Failure Recovery before branching into cluster mode, external app repositories, or broader workflows.

For release-level evidence, use Release Proof; it points to the latest public GitHub release, package proof, CI guardrails, and hosted demo status.

This documentation then expands into architecture, service mode, API references, and example projects.

Build

Operations

Indices and tables