chore: turn vision into capability (#679)

Fixes https://github.com/microsoft/playwright-mcp/issues/420
This commit is contained in:
Pavel Feldman
2025-07-16 16:40:00 -07:00
committed by GitHub
parent 012c906500
commit d61aa16fee
23 changed files with 366 additions and 575 deletions

View File

@@ -34,7 +34,6 @@ test('test snapshot tool list', async ({ client }) => {
'browser_navigate_forward',
'browser_navigate',
'browser_network_requests',
'browser_pdf_save',
'browser_press_key',
'browser_resize',
'browser_snapshot',
@@ -47,45 +46,33 @@ test('test snapshot tool list', async ({ client }) => {
]));
});
test('test vision tool list', async ({ visionClient }) => {
const { tools: visionTools } = await visionClient.listTools();
expect(new Set(visionTools.map(t => t.name))).toEqual(new Set([
'browser_close',
'browser_console_messages',
'browser_file_upload',
'browser_handle_dialog',
'browser_install',
'browser_navigate_back',
'browser_navigate_forward',
'browser_navigate',
'browser_network_requests',
'browser_pdf_save',
'browser_press_key',
'browser_resize',
'browser_screen_capture',
'browser_screen_click',
'browser_screen_drag',
'browser_screen_move_mouse',
'browser_screen_type',
'browser_tab_close',
'browser_tab_list',
'browser_tab_new',
'browser_tab_select',
'browser_wait_for',
]));
});
test('test capabilities', async ({ startClient }) => {
test('test capabilities (pdf)', async ({ startClient }) => {
const { client } = await startClient({
args: ['--caps="core"'],
args: ['--caps=pdf'],
});
const { tools } = await client.listTools();
const toolNames = tools.map(t => t.name);
expect(toolNames).not.toContain('browser_file_upload');
expect(toolNames).not.toContain('browser_pdf_save');
expect(toolNames).not.toContain('browser_screen_capture');
expect(toolNames).not.toContain('browser_screen_click');
expect(toolNames).not.toContain('browser_screen_drag');
expect(toolNames).not.toContain('browser_screen_move_mouse');
expect(toolNames).not.toContain('browser_screen_type');
expect(toolNames).toContain('browser_pdf_save');
});
test('test capabilities (vision)', async ({ startClient }) => {
const { client } = await startClient({
args: ['--caps=vision'],
});
const { tools } = await client.listTools();
const toolNames = tools.map(t => t.name);
expect(toolNames).toContain('browser_mouse_move_xy');
expect(toolNames).toContain('browser_mouse_click_xy');
expect(toolNames).toContain('browser_mouse_drag_xy');
});
test('support for legacy --vision option', async ({ startClient }) => {
const { client } = await startClient({
args: ['--vision'],
});
const { tools } = await client.listTools();
const toolNames = tools.map(t => t.name);
expect(toolNames).toContain('browser_mouse_move_xy');
expect(toolNames).toContain('browser_mouse_click_xy');
expect(toolNames).toContain('browser_mouse_drag_xy');
});