CamouFlow v0.4.0

CamouFlow 0.4.0

Describe a browser task, inspect its outputs, then replay a reviewed workflow without another model request.

New

• Isolated catalog demo and three local starters: catalog to JSON, page to text, form fill without submission.
• Structured table/text extraction, source URLs, preview, JSON/CSV export and workspace-scoped AI history.
• Starting URL/host policy, default action confirmations, operator questions, pause/resume and completed step retention on stop.
• Exact input parameterization, read-only replay/schema checks and workflow export.
• Source-only local JSON CLI, optional restricted stdio MCP and an inactive n8n example.
• Fixed stale provider/profile readiness, false click success, editable-content exposure and output path traversal.

Try it

Settings -> AI: configure and test a Chat Completions-compatible provider. Scenarios -> AI -> Try isolated demo -> Start. Inspect/export the table, parameterize the demo URL, check replay and save. The provider may charge for the first AI task; deterministic replay does not call the model.

English guide · Русская инструкция · Recorded real UI demo

Verified

79 automated tests; real-browser fixture extraction/replay and CLI/MCP checks on Camoufox and CloakBrowser; three successful configured DeepSeek runs on each engine; real-provider UI export/replay/save; isolated Windows n8n 2.41.5 import/execution/JSON parsing; packaged runtime checks on both engines; 35 documentation pages built without warnings.

These are synthetic catalog checks, not a benchmark of every model or website. Live cloud auth/list access was checked read-only; production cloud runs were not exercised.

Data and upgrade

Page text, labels, URLs, instructions, outputs and operator answers go to the configured provider. Editable contents are excluded; this is not universal secret detection. Automatic replay checks are read-only. An interrupted action may already have changed a page. CLI/MCP require the source Python environment, not the windowed EXE; close the desktop first. MCP request cancellation does not cancel an already-started local run.

Windows package: unzip into a new folder. Back up your existing settings, profiles and scenaries before moving data. No personal profiles, API keys or cloud credentials are included. Browser binaries may be downloaded on first use.

Кратко по-русски

AI-задача теперь выдаёт проверяемые данные и превращается в параметризованный сценарий. Добавлены таблицы, JSON/CSV, история, подтверждения, пауза, демо и шаблоны. Проверены реальная DeepSeek, оба движка, CLI/MCP и отдельный n8n. Повтор работает без запросов к модели; интерактивные сценарии требуют ручной проверки. Популярность и совместимость со всеми сайтами не гарантируются.

CamouFlow v0.3.1

Release v0.3.1

• Local scenario run history with retained artifact links.
• Release smoke automation and clean PyInstaller builds.
• AI/Python scenario runtime integration from the committed desktop update.

Validation: 55 pytest tests passed; clean packaged runtime checks passed for Camoufox and CloakBrowser.

CamouFlow 0.3.0

Desktop workspace update: scenario marketplace, recorder and step debugger; workspace-aware run queue; profile lifecycle and crash-lock recovery; proxy assignment and interface responsiveness improvements.

Windows x64 portable build. Extract the complete ZIP before running CamouFlow.exe. Back up existing portable data before upgrading. Browser engines are downloaded separately.

Validation: 77 tests, Python lint, all-page QML smoke and packaged startup smoke passed locally. Packaged interactive workflows were not fully acceptance-tested. Backend is maintained separately and is not included.

Commit: 770eea6

CamouFlow 0.2.0

CamouFlow 0.2.0 — Teams for everyone

Windows x64. Распакуйте и запустите CamouFlow.exe. Первый запуск скачивает браузерные движки.
Windows x64. Unpack and run CamouFlow.exe. Browser engines download on first launch.

v1.0.0

main

fix build.bat