Brosh Browser Screenshot

# Changelog All notable changes to this project will be documented in this file. The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.1.0/), and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html). ## [Unreleased] ### Added - Git-tag-based semantic versioning with hatch-vcs - Comprehensive test suite with pytest and coverage reporting - Multiplatform binary releases (Linux, Windows, macOS Intel/ARM) - Local development scripts for build, test, and release - GitHub Actions CI/CD with automated testing and releases - PyInstaller-based binary compilation - Makefile for convenient development workflows - Binary distribution via GitHub Releases - Comprehensive documentation for installation and usage ### Changed - Enhanced GitHub Actions workflows for multiplatform support - Improved package metadata and dependency management - Updated README with binary installation instructions ## [0.1.0] - Initial Release ### Added - Initial public release of brosh (Browser Screenshot Tool) - Playwright-based async implementation for capturing scrolling screenshots - Support for Chrome, Edge, and Safari browsers - Smart section detection for descriptive filenames - Multiple output formats: PNG, JPG, and animated PNG (APNG) - Remote debugging mode to connect to existing browser sessions - MCP (Model Context Protocol) server integration - HTML extraction feature to capture visible element content - Automatic text extraction: converts visible HTML to Markdown format - Configurable scroll steps and starting positions - Automatic browser detection and fallback logic - Comprehensive error handling and retry mechanisms ### Changed - Refactored monolithic script into modular Python package structure - Migrated from script-based to package-based distribution - Updated to use modern Python packaging with pyproject.toml ### Fixed - Browser connection issues with improved retry logic - MCP response format now properly excludes null fields and uses camelCase for field names (e.g., `mimeType` instead of `mime_type`). This completes the MCP data model restructuring for flat JSON output. - MCP server results now handle size limits properly with progressive compression. - Screenshot timeout handling. - Image scaling and format conversion edge cases. - MCP async execution error where Task object was returned instead of actual results. ### Added - PNG optimization using pyoxipng for all captured screenshots. - HTML content compression that removes SVG elements while preserving dimensions. - Progressive compression strategy for MCP results exceeding 1MB size limit. - Automatic downsampling and content reduction for oversized MCP responses. - Separate `capture_webpage_async` API function for async contexts. - New MCP configuration flags for granular output control: - `fetch_image`: Controls whether image data is included in MCP output (default: False). - `fetch_image_path`: Controls whether image paths are included (default: True). - `fetch_text`: Controls whether extracted text is included (default: True). - `trim_text`: Controls whether text is trimmed to 200 characters (default: True). - Comprehensive test updates for all new functionality and refactoring. ### Changed - MCP output structure changed from nested dictionary to flat JSON format, aligning with the data model restructuring. - Old: `{"/path/to/file.png": {"selector": "...", "text": "..."}}` - New: `{"image_path": "/path/to/file.png", "selector": "...", "text": "..."}` - Renamed `html` parameter to `fetch_html` throughout the codebase for consistency. - Text extraction now automatically trims to 200 characters by default (configurable). - Added support for `file://` URLs to capture local HTML files. ### Code Quality - **Imports**: Removed unused imports and organized existing ones using Ruff. - **Security**: Hardened `subprocess` calls by using `shutil.which` to ensure absolute paths for system commands (`system_profiler`, `pkill`, `taskkill`), reducing potential S607 risks. - **Style**: - Replaced `os.path.exists()` with `pathlib.Path().exists()` for consistency and modern practice. - Ensured all `datetime.now()` calls use `timezone.utc` for timezone-aware timestamps. - Made boolean parameters in function definitions keyword-only where appropriate to adhere to FBT001/FBT002 Ruff rules (e.g., in API functions, CLI class, MCP functions). - Replaced various magic numbers with named constants for improved readability and maintainability across several modules (e.g., default dimensions, timeouts, retry counts, sleep durations). - **Testing**: - Refactored `pytest.raises()` blocks in model validation tests to ensure only the exception-raising statement is within the `with` block. - Added `match` parameter to several `pytest.raises()` calls for more specific exception message checking in API and browser tests. - Simplified nested `with patch(...)` statements in API tests by combining them into single `with` statements. - **Async Practices**: Reviewed async functions for `time.sleep` and confirmed usage of `asyncio.sleep` where appropriate. ## [0.1.0] - 2025-06-12 ### Added - Initial implementation as a monolithic script - Basic screenshot capture functionality - Browser management commands (run, quit) - Fire-based CLI interface [Unreleased]: https://github.com/twardoch/brosh/compare/v0.1.0...HEAD [0.1.0]: https://github.com/twardoch/brosh/releases/tag/v0.1.0