Strategic Potential =================== This page explains AGILab's strategic potential score, the evidence behind it, and the concrete proof needed before the score should increase. It is a public scorecard, not a marketing claim. Current score ------------- AGILab currently supports a ``Strategic potential`` score of ``4.2 / 5``. That score is based on a narrow position: AGILab is an experimentation and engineering-validation workbench that connects project setup, reproducible execution, evidence, and operator-facing analysis. It is not scored as a production MLOps platform. Strategic wedge --------------- AGILab's strongest wedge is: ``run orchestration + evidence + reproducibility for engineering-grade AI workflows`` The project should be evaluated against that wedge, not against generic dashboard builders, standalone schedulers, or production-serving platforms. What supports the score ----------------------- .. list-table:: :header-rows: 1 :widths: 24 38 38 * - Dimension - Evidence - Limit * - Adoption path - Public Hugging Face demo, local ``flight_project`` first proof, ``run_manifest.json``, and newcomer troubleshooting. - Broader fresh-install validation still needs to run across more external machines. * - Differentiation - AGILab ties execution, artifacts, run evidence, Release Decision, and analysis pages into one path. - The project must keep this evidence-first story clear and avoid drifting into a generic MLOps comparison. * - Reproducibility - Compatibility report, KPI evidence bundle, run-diff evidence, revision-traceability, supply-chain metadata, and CI artifact harvest contracts. - Formal certification and production attestation are intentionally out of scope. * - Orchestration path - Multi-app DAG contract, global DAG reports, dispatch-state evidence, operator-action reports, and static operator UI proof. - Live global runner UI and broader cross-app execution remain roadmap work. * - Data access path - Connector facility, resolution, health planning, runtime adapters, app-local catalogs, and account-free cloud-emulator contracts. - Credentialed live cloud/provider validation remains operator-gated. * - Example quality - Packaged examples now have a learning path, expected inputs/outputs, safe adaptation guidance, and packaging tests. - The example scripts still need a final maturity pass before they should be treated as external SDK-quality examples. Vertical stories ---------------- Two vertical stories should carry most public evaluation: .. list-table:: :header-rows: 1 :widths: 24 38 38 * - Story - Why it matters - Evidence to keep fresh * - Flight + meteo first proof - Demonstrates public onboarding, reproducible file input, generated artifacts, and analysis pages without private repositories. - ``agilab first-proof --json``, ``flight_project``, ``meteo_forecast_project``, compatibility matrix, and packaged examples. * - Data IO 2026 decision workflow - Demonstrates mission-style decision evidence, richer artifacts, connector-aware provenance, and the future DAG/release-decision path. - ``data_io_2026_project``, Release Decision evidence, connector reports, run-diff evidence, and promotion-decision exports. Elasticity opportunity ---------------------- The most credible strategic improvement is elasticity, not another tracking dashboard or model registry. AGILab is strongest at the ``Train -> Test -> Evidence`` loop: run a pipeline, inspect artifacts, and compare outcomes. The first teaching route now exists in the optional industrial optimization examples: Active Mesh Optimization models relay UAVs as controllable agents in a compact centralized PPO policy, then exports movement, topology, and delivery evidence. That is enough to demonstrate the contract shape, but it is not yet a claim of full decentralized MARL certification for aircraft, UAV, or satellite fleets. This opportunity should raise the strategic score only when public evidence shows the full hardening path: baseline versus adaptive-network comparison, failure-injection comparison, service-contract handoff, and reproducible multi-app DAG evidence through the existing artifact layer. Score movement rule ------------------- Do not raise the score only because the docs sound better. Raise it only when new evidence closes a listed gap. .. list-table:: :header-rows: 1 :widths: 20 40 40 * - Target score - Required proof - Still not claimed * - ``4.3 / 5`` - Final beta gate passes with network included; packaged examples pass the external-beta maturity contract; public docs link the strategic scorecard; HF demo, PyPI, and docs are aligned to the same public release. - Production serving, enterprise governance, and formal certification. * - ``4.5 / 5`` - At least two external fresh-machine first proofs are attached through run manifests or artifact indexes; global DAG operator flow has a live UI proof; connector examples demonstrate credentialed operator-gated validation without leaking secrets. - Multi-tenant production operations or cloud/Kubernetes parity. * - ``5.0 / 5`` - AGILab becomes a de facto standard bridge for experimentation-to-handoff workflows with repeated external adoption evidence. - Generic full-stack MLOps replacement. Signals that lower the score ---------------------------- Reduce the score, or keep it flat, if any of these appear in public artifacts: - private app names or symlink-only local state leaking into public gates - stale alpha/beta wording after a promoted release - examples using private AGI internals or scratch-only snippets - HF Space, PyPI, README, and docs pointing to different release states - evidence commands that no longer pass or no longer match documented claims - production-serving or certification language that is not backed by evidence Related pages ------------- - :doc:`agilab-mlops-positioning` for toolchain fit and current category scores - :doc:`compatibility-matrix` for validated public routes and evidence commands - :doc:`beta-readiness` for the beta promotion gate - :doc:`features` for shipped capability evidence