From d5344ade75a22a30df1924cd03ce2e7ebde7a6b3 Mon Sep 17 00:00:00 2001 From: STJ Date: Fri, 13 Mar 2026 17:11:28 -0700 Subject: [PATCH] simplify actions, remove ones that cannot be used --- strix/tools/browser/browser_actions.py | 2 - .../tools/browser/browser_actions_schema.xml | 40 ++----------------- 2 files changed, 3 insertions(+), 39 deletions(-) diff --git a/strix/tools/browser/browser_actions.py b/strix/tools/browser/browser_actions.py index dba89bad..82b0da00 100644 --- a/strix/tools/browser/browser_actions.py +++ b/strix/tools/browser/browser_actions.py @@ -38,13 +38,11 @@ BrowserUseLocalAction = Literal[ "search_page", "find_elements", "screenshot", - "save_as_pdf", "dropdown_options", "select_dropdown", "evaluate", "switch", "close_tab", - "done", ] _TASK_TIMEOUT = 300 diff --git a/strix/tools/browser/browser_actions_schema.xml b/strix/tools/browser/browser_actions_schema.xml index 1a7fc650..276a658c 100644 --- a/strix/tools/browser/browser_actions_schema.xml +++ b/strix/tools/browser/browser_actions_schema.xml @@ -44,7 +44,6 @@ **Visual Analysis:** [Requires to be running in non-sandbox environment] - screenshot: Capture screenshot. Optional 'file_name'. - - save_as_pdf: Save page as PDF. Optional 'file_name', 'print_background', 'landscape', 'scale', 'paper_format'. **Form Controls:** - dropdown_options: Get dropdown options. Requires 'index'. @@ -57,8 +56,7 @@ - switch: Switch between tabs. Requires 'tab_id'. - close_tab: Close browser tab. Requires 'tab_id'. - **Task Completion:** - - done: Complete the task. Requires 'text', optional 'success', 'files_to_display'. + Required for 'run' action. A natural-language description of what to do in @@ -84,7 +82,7 @@ Element index from the DOM state. Used by: click, input, upload_file, dropdown_options, select_dropdown, scroll (for scrollable elements). - Text content. Required for 'input', 'select_dropdown', 'find_text', and 'done' actions. + Text content. Required for 'input', 'select_dropdown', and 'find_text' actions. Clear existing text before input. Used by 'input' action. Default: false. @@ -138,19 +136,7 @@ Include element text content for 'find_elements' action. Default: true. - File name for 'screenshot', 'save_as_pdf', 'write_file', 'read_file', 'replace_file' actions. - - - Include background graphics in PDF for 'save_as_pdf' action. Default: false. - - - Landscape orientation for 'save_as_pdf' action. Default: false. - - - Scale factor for 'save_as_pdf' action. Range: 0.1-2.0. - - - Paper format for 'save_as_pdf' action. E.g. "A4", "Letter". + File name for 'screenshot' action. JavaScript code to execute. Required for 'evaluate' action. @@ -158,12 +144,6 @@ Tab ID for 'switch' and 'close_tab' actions. Last 4 chars of target_id from browser state. - - Task success status for 'done' action. - - - Files to display for 'done' action. - Only used with action='run'. A list of history fields to include in the response under a "fields" key. By default only the final result is returned. Use this @@ -306,13 +286,6 @@ page_screenshot.png - # Save page as PDF - - save_as_pdf - report.pdf - true - - # Get dropdown options dropdown_options @@ -344,13 +317,6 @@ a3f2 - # Complete the task - - done - Successfully extracted all product data - true - - # Close the browser when completely done close_browser