UI はローカライズ済みです。スキルの紹介と説明文書は選択した言語へ AI が自動翻訳し、コード・コマンド・パス・専門用語は原文のままです。原作者の文書が優先されます。
スキルを探す
運用・デプロイ · SKILL.md

web-ui-and-deployment

web-ui-and-deployment

Guides OWL's English Gradio interface, environment-file management, Docker Compose deployment, Xvfb, and browser-runtime troubleshooting.

Independent GitHub snapshot · Automated classification · Original documentation governs

スキルの説明

GitHub

name: web-ui-and-deployment description: "Guides OWL's English Gradio interface, environment-file management, Docker Compose deployment, Xvfb, and browser-runtime troubleshooting." disable-model-invocation: true metadata: disco-role: operating license: Apache 2.0

OWL Web UI and Deployment

Use this route for the English Gradio UI, API-key configuration through the interface, Docker/Compose, Xvfb, Playwright, ports, mounts, or container troubleshooting. It is an operational route, not a promise that a server or Docker daemon is available.

Local UI path

  1. Install OWL and its declared dependencies in Python 3.10–3.12.
  2. Create a private environment file from the package's documented template, fill only the provider and tool keys needed by the selected workflow, and keep it outside version control.
  3. Run the English web application from the project layout expected by the source script. The implementation uses a script-oriented absolute import for utils, so do not assume import owl.webapp from an arbitrary current directory is equivalent.
  4. Read web-ui-reference.md, then use check_web_ui_config.py to compare the selected example module names with the files actually present before submitting a request. The UI's module description table can contain names not present in a particular checkout.
  5. Enter a nonblank task, select a module that exports the expected construct_society interface, and inspect the status/token/log output. Route provider construction to workforce-workflows.

The UI can initialize, read, update, and delete .env values. Treat those callbacks as mutations: do not paste secrets into prompts or logs, and prefer shell environment variables or a protected file when policy disallows UI writes.

Container path

Read docker-operations.md before using Compose, and run check_docker_runtime.sh for non-mutating host diagnostics. The supported pattern uses a service exposing port 7860, a mounted .env, example/data mounts, large shared memory for browser work, and an Xvfb-backed Python command where a display is required. Build and image pulls are networked and potentially expensive; the bundled checker only diagnoses the host and never starts or builds a container.

Use troubleshooting.md for stale module names, port conflicts, missing keys, Gradio errors, browser/Xvfb failures, or Docker health and Compose errors. Local document behavior routes to document-processing, and benchmark work to gaia-evaluation.

保存したスキル