---

name: "interactive-frontends" description: "Operate DeepAnalyze CLI, WebUI v2, and Jupyter frontend workflows." metadata: disco-role: operating disable-model-invocation: true license: MIT

interactive-frontends

Use this sub-skill for terminal, browser, and Jupyter-facing DeepAnalyze tasks.

Route first

• Terminal commands, file IDs, clear / clear-all, history paths, downloads -> CLI reference
• Browser workspace, file preview, download bundles, report export, provider choice, local vs Docker execution, frontend/backend/file ports -> WebUI v2
• Jupyter setup, uv sync, .env, config.toml, uv run CLI.py, notebook workspace, MCP -> Jupyter frontend
• Port conflicts, Node/npm, Docker image, PDF export, fonts, MCP failures, unsupported previews -> Troubleshooting
• Safe preflight audit -> WebUI prereq checker

What this skill must cover

• CLI command help, upload, files, delete, download, status, history, fid, clear, clear-all.
• WebUI v2 provider modes: Local, HeyWhale API, Custom Model.
• Workspace behavior: preview, download, generated-file handling, export, session separation.
• Local vs Docker execution and the ports used by the browser demo.
• Jupyter notebook startup, config, workspace, and MCP connection behavior.

Boundaries

• Do not expand into raw API request/response schemas.
• Do not explain model serving, quantization, training, or evaluation here.
• Use the legacy demo/chat notes only when the user explicitly asks about the older demo.
• Prefer the bundled references in this subtree over memory when answering.

Operating hints

• If a task sounds like "can I start it?" or "why is it failing?", consult Troubleshooting first.
• If the task is about a browser demo coming up cleanly, run or describe scripts/check_webui_prereqs.py before changing ports or deleting files.
• If the task is about notebook-first work, keep the answer centered on uv, config.toml, the generated notebook, and the local workspace directory.