AGENT SKILL

app-deployment

Generate, inspect, serve locally, and publish Mercury prediction apps from trained MLJAR AutoML models.

License 许可
MIT
最近更新
2026-09-03
Stars
★ 351
Forks
27

01 前言 Front Matter

name
app-deployment
license
MIT
metadata
{ "disco-role": "operating" }
description
Generate, inspect, serve locally, and publish Mercury prediction apps from trained MLJAR AutoML models.
disable-model-invocation
true

02 资源文件 4

03 SKILL.md 正文

App Deployment

Use this sub-skill when a task involves turning an already fitted supervised.AutoML model into a Mercury prediction app, inspecting the generated app workspace, previewing it locally, or preparing a guarded publish flow.

Route here for

  • Generating app files with automl.app(path=..., overwrite=..., title=...).
  • Inspecting generated files such as mljar_app.json, notebooks, automl.zip, config.toml, requirements.txt, and runtime.txt.
  • Starting a local Mercury preview with automl.local_app() when Mercury is installed and the user accepts a foreground server process.
  • Publishing with automl.publish_app(...) only after the user explicitly authorizes browser, network, and platform-auth actions.
  • Diagnosing app workspace, Mercury startup, and publish failures.

Route elsewhere

  • To ../training-core/ when the user still needs to train, configure, evaluate, or retrain an AutoML model.
  • To ../artifacts-reports/ for interpreting saved model folders, leaderboards, structured reports, explainability outputs, or results_path artifacts outside the generated app workspace.
  • To ../fairness-workflows/ for fairness metrics, sensitive features, and fairness dashboard/report interpretation.
  • Exclude maintainer-only CI docs deployment, package release, and platform credential administration tasks.

Operating checklist

  1. Confirm supervised imports and the model was fitted. App generation needs a valid results_path containing fitted AutoML artifacts.
  2. Choose the deployment action:
    • app() for file generation only.
    • local_app() for a local Mercury server and browser preview.
    • publish_app() for the hosted platform workflow; require explicit authorization.
  3. Decide output location and overwrite policy before writing app files. overwrite=True deletes an existing app workspace.
  4. Inspect the generated mljar_app.json manifest before serving or publishing, especially supports.single_sample, feature count, and selected notebooks.
  5. For local serving, install the generated app dependencies in a Python 3.10+ environment and ensure the mercury executable is on PATH.
  6. For publishing, avoid embedding credentials in code. Use open_browser=False in non-interactive sessions and have the user open the printed login URL manually.

Read next

  • App workflows for generation, local preview, manual hosting, and publish recipes.
  • API reference for method signatures, return values, generated files, and state behavior.
  • Troubleshooting for overwrite, unfitted model, Mercury, server logs, publish/auth/network, URL reuse, and wide dataset issues.
  • scripts/generate_app_workspace_smoke.py for a safe synthetic helper that trains a tiny model and generates app files without starting Mercury or publishing.
  • Root package assumptions: ../../references/package-overview.md; cross-cutting install/import issues: ../../references/troubleshooting.md.

04 同仓库更多 Skills

VectorSpaceLab
VectorSpaceLab/AREX-Skill

passnet-triton-opt

Make a single PassNet Triton kernel fast AND numerically faithful: performance model (when a replacement can win at all), block/grid/warp tuning, autotune policy, launch overhead, and per-op numeric recipes to pass the dtype baseline tolerances. Use when a pass matches and is correct but speedup below expected, or when correctness fails by small numeric margins.

★ 351
VectorSpaceLab
VectorSpaceLab/AREX-Skill

passnet-solve

END-TO-END playbook for solving one PassNet sample: analyze the computation graphs, decide the optimization strategy (what to fuse, what to replace, what to leave alone), drive the iteration loop, and maximize the sample score. This is the ENTRY skill — invoke it first for any PassNet optimization task; it tells you when to use passnet-pattern-fusion, passnet-triton-opt and passnet-feedback.

★ 351
VectorSpaceLab
VectorSpaceLab/AREX-Skill

passnet-skill

PassNet GPU kernel optimization via compiler passes. Design and implement Triton-based optimization passes, create pass files under ./pass_dir/, self-evaluate with pass_evaluator, and iterate to maximize GPU speedup.

★ 351
VectorSpaceLab
VectorSpaceLab/AREX-Skill

passnet-pattern-fusion

Author PassNet pass files: write patterns that actually MATCH the FX graph, pick fusion regions (which ops to absorb into one Triton kernel), and structure multi-pass submissions with the mandatory shared-dispatch architecture. Use when creating or fixing pattern()/replacement_args()/replacement_func() files, when a pass "failed to match", or when deciding how to fuse multiple kernels/ops.

★ 351
VectorSpaceLab
VectorSpaceLab/AREX-Skill

passnet-orchestrate

Round/batch planning and decision support for PassNet graph optimization. Use for multi-sample triage, worker allocation, eval-budget policy, and post-eval keep/revert decisions. For solving one concrete sample end to end, use passnet-solve as the entry skill; consult this skill only when you need broader planning or a specific decision gate.

★ 351
VectorSpaceLab
VectorSpaceLab/AREX-Skill

passnet-feedback

Fast, deterministic feedback for PassNet work: pre-flight pattern/match verification WITHOUT burning a GPU evaluation, per-node bottleneck analysis of a sample's graphs, and eval-log parsing into per-variant status + estimated score + failure classification. Use BEFORE every GPU evaluation (check_pattern), at the START of a sample (analyze_graph), and AFTER every evaluation (parse_eval_log).

★ 351