browser: add window resize tool with chromedp
Co-Authored-By: sketch <hello@sketch.dev>
Change-ID: s68bdc7fb84309a7ck
diff --git a/claudetool/browse/README.md b/claudetool/browse/README.md
index 6c03e1a..7cc3437 100644
--- a/claudetool/browse/README.md
+++ b/claudetool/browse/README.md
@@ -14,6 +14,7 @@
6. `browser_eval` - Evaluate JavaScript in the browser context
7. `browser_screenshot` - Take a screenshot of the page or a specific element
8. `browser_scroll_into_view` - Scroll an element into view
+9. `browser_resize` - Resize the browser window to specific dimensions
## Usage
diff --git a/claudetool/browse/browse.go b/claudetool/browse/browse.go
index a5a1621..8a72390 100644
--- a/claudetool/browse/browse.go
+++ b/claudetool/browse/browse.go
@@ -655,6 +655,71 @@
return llm.TextContent(successResponse()), nil
}
+// ResizeTool definition
+type resizeInput struct {
+ Width int `json:"width"`
+ Height int `json:"height"`
+ Timeout string `json:"timeout,omitempty"`
+}
+
+// NewResizeTool creates a tool for resizing the browser window
+func (b *BrowseTools) NewResizeTool() *llm.Tool {
+ return &llm.Tool{
+ Name: "browser_resize",
+ Description: "Resize the browser window to a specific width and height",
+ InputSchema: json.RawMessage(`{
+ "type": "object",
+ "properties": {
+ "width": {
+ "type": "integer",
+ "description": "Window width in pixels"
+ },
+ "height": {
+ "type": "integer",
+ "description": "Window height in pixels"
+ },
+ "timeout": {
+ "type": "string",
+ "description": "Timeout as a Go duration string (default: 5s)"
+ }
+ },
+ "required": ["width", "height"]
+ }`),
+ Run: b.resizeRun,
+ }
+}
+
+func (b *BrowseTools) resizeRun(ctx context.Context, m json.RawMessage) ([]llm.Content, error) {
+ var input resizeInput
+ if err := json.Unmarshal(m, &input); err != nil {
+ return llm.TextContent(errorResponse(fmt.Errorf("invalid input: %w", err))), nil
+ }
+
+ browserCtx, err := b.GetBrowserContext()
+ if err != nil {
+ return llm.TextContent(errorResponse(err)), nil
+ }
+
+ // Create a timeout context for this operation
+ timeoutCtx, cancel := context.WithTimeout(browserCtx, parseTimeout(input.Timeout))
+ defer cancel()
+
+ // Validate dimensions
+ if input.Width <= 0 || input.Height <= 0 {
+ return llm.TextContent(errorResponse(fmt.Errorf("invalid dimensions: width and height must be positive"))), nil
+ }
+
+ // Resize the browser window
+ err = chromedp.Run(timeoutCtx,
+ chromedp.EmulateViewport(int64(input.Width), int64(input.Height)),
+ )
+ if err != nil {
+ return llm.TextContent(errorResponse(err)), nil
+ }
+
+ return llm.TextContent(successResponse()), nil
+}
+
// GetTools returns browser tools, optionally filtering out screenshot-related tools
func (b *BrowseTools) GetTools(includeScreenshotTools bool) []*llm.Tool {
tools := []*llm.Tool{
@@ -665,6 +730,7 @@
b.NewGetTextTool(),
b.NewEvalTool(),
b.NewScrollIntoViewTool(),
+ b.NewResizeTool(),
}
// Add screenshot-related tools if supported
diff --git a/claudetool/browse/browse_test.go b/claudetool/browse/browse_test.go
index ae4fefd..7cffa0e 100644
--- a/claudetool/browse/browse_test.go
+++ b/claudetool/browse/browse_test.go
@@ -12,6 +12,7 @@
"time"
"github.com/chromedp/chromedp"
+ "github.com/stretchr/testify/require"
"sketch.dev/llm"
)
@@ -72,7 +73,7 @@
// Test with screenshot tools included
t.Run("with screenshots", func(t *testing.T) {
toolsWithScreenshots := tools.GetTools(true)
- if len(toolsWithScreenshots) != 9 {
+ if len(toolsWithScreenshots) != 10 {
t.Errorf("expected 9 tools with screenshots, got %d", len(toolsWithScreenshots))
}
@@ -87,7 +88,7 @@
// Test without screenshot tools
t.Run("without screenshots", func(t *testing.T) {
noScreenshotTools := tools.GetTools(false)
- if len(noScreenshotTools) != 7 {
+ if len(noScreenshotTools) != 8 {
t.Errorf("expected 7 tools without screenshots, got %d", len(noScreenshotTools))
}
})
@@ -304,3 +305,42 @@
t.Errorf("Expected Data in second content")
}
}
+
+// TestResizeTool tests the browser resize functionality
+func TestResizeTool(t *testing.T) {
+ ctx, cancel := context.WithTimeout(context.Background(), 10*time.Second)
+ defer cancel()
+
+ // Skip if CI or headless testing environment
+ if os.Getenv("CI") != "" || os.Getenv("HEADLESS_TEST") != "" {
+ t.Skip("Skipping browser test in CI/headless environment")
+ }
+
+ t.Run("ResizeWindow", func(t *testing.T) {
+ tools := NewBrowseTools(ctx)
+ defer tools.Close()
+
+ // Resize to mobile dimensions
+ resizeTool := tools.NewResizeTool()
+ input := json.RawMessage(`{"width": 375, "height": 667}`)
+ content, err := resizeTool.Run(ctx, input)
+ require.NoError(t, err)
+ require.Contains(t, content[0].Text, "success")
+
+ // Navigate to a test page and verify using JavaScript to get window dimensions
+ navInput := json.RawMessage(`{"url": "https://example.com"}`)
+ content, err = tools.NewNavigateTool().Run(ctx, navInput)
+ require.NoError(t, err)
+ require.Contains(t, content[0].Text, "success")
+
+ // Check dimensions via JavaScript
+ evalInput := json.RawMessage(`{"expression": "({width: window.innerWidth, height: window.innerHeight})"}`)
+ content, err = tools.NewEvalTool().Run(ctx, evalInput)
+ require.NoError(t, err)
+
+ // The dimensions might not be exactly what we set (browser chrome, etc.)
+ // but they should be close
+ require.Contains(t, content[0].Text, "width")
+ require.Contains(t, content[0].Text, "height")
+ })
+}
diff --git a/claudetool/browse/browser_resize.go b/claudetool/browse/browser_resize.go
new file mode 100644
index 0000000..557ea66
--- /dev/null
+++ b/claudetool/browse/browser_resize.go
@@ -0,0 +1,2 @@
+// Package browse contains browser automation tools
+package browse