Compare commits

...

9 Commits

Author SHA1 Message Date
Karl Seguin
5826caf6dc Merge pull request #2070 from lightpanda-io/mcp-new-action-tools
Some checks are pending
e2e-test / zig build release (push) Waiting to run
e2e-test / demo-scripts (push) Blocked by required conditions
e2e-test / wba-demo-scripts (push) Blocked by required conditions
e2e-test / wba-test (push) Blocked by required conditions
e2e-test / cdp-and-hyperfine-bench (push) Blocked by required conditions
e2e-test / perf-fmt (push) Blocked by required conditions
e2e-test / browser fetch (push) Blocked by required conditions
zig-test / zig fmt (push) Waiting to run
zig-test / zig test using v8 in debug mode (push) Waiting to run
zig-test / zig test (push) Waiting to run
zig-test / perf-fmt (push) Blocked by required conditions
mcp: Add hover, press, selectOption, setChecked
2026-04-04 10:20:54 +08:00
Karl Seguin
b0c6c2d591 Merge pull request #2083 from tmchow/fix/2080-keyboard-event-propagation
Some checks failed
e2e-test / zig build release (push) Has been cancelled
e2e-test / demo-scripts (push) Has been cancelled
e2e-test / wba-demo-scripts (push) Has been cancelled
e2e-test / wba-test (push) Has been cancelled
e2e-test / cdp-and-hyperfine-bench (push) Has been cancelled
e2e-test / perf-fmt (push) Has been cancelled
e2e-test / browser fetch (push) Has been cancelled
zig-test / zig fmt (push) Has been cancelled
zig-test / zig test using v8 in debug mode (push) Has been cancelled
zig-test / zig test (push) Has been cancelled
zig-test / perf-fmt (push) Has been cancelled
fix: propagate keyUp and char keyboard events to JS listeners
2026-04-04 08:19:53 +08:00
Trevin Chow
b33bb54442 fix: propagate keyUp and char keyboard events to JS listeners
dispatchKeyEvent only handled keyDown, returning early for keyUp,
rawKeyDown, and char types. This meant JS keyup and keypress
listeners never fired via CDP.

Now keyUp dispatches as "keyup" and char dispatches as "keypress".
rawKeyDown remains a no-op (Chrome-internal, not used for JS dispatch).

Fixes #2080
Ref #2043
2026-04-03 17:08:09 -07:00
Pierre Tachoire
06fe6c5e7d Merge pull request #1934 from lightpanda-io/http-cache
Some checks failed
e2e-test / zig build release (push) Has been cancelled
e2e-test / demo-scripts (push) Has been cancelled
e2e-test / wba-demo-scripts (push) Has been cancelled
e2e-test / wba-test (push) Has been cancelled
e2e-test / cdp-and-hyperfine-bench (push) Has been cancelled
e2e-test / perf-fmt (push) Has been cancelled
e2e-test / browser fetch (push) Has been cancelled
zig-test / zig fmt (push) Has been cancelled
zig-test / zig test using v8 in debug mode (push) Has been cancelled
zig-test / zig test (push) Has been cancelled
zig-test / perf-fmt (push) Has been cancelled
HTTP Caching
2026-04-03 17:11:06 +02:00
Adrià Arrufat
72229f715a Merge branch 'main' into mcp-new-action-tools 2026-04-03 07:06:10 +02:00
Adrià Arrufat
6c9a5ddab8 Extract shared helpers to reduce duplication
- Extract dispatchInputAndChangeEvents() in actions.zig, used by fill,
  selectOption, and setChecked
- Extract resolveNodeAndPage() in tools.zig, used by click, fill, hover,
  selectOption, setChecked, and nodeDetails handlers
2026-04-02 11:20:28 +02:00
Adrià Arrufat
46a63e0b4b Add focus before fill and findElement MCP tool
- fill action now calls focus() on the element before setting its value,
  ensuring focus/focusin events fire for JS listeners
- Add findElement MCP tool for locating interactive elements by ARIA role
  and/or accessible name (case-insensitive substring match)
- Add tests for findElement (by role, by name, no matches, missing params)
2026-04-02 11:03:49 +02:00
Adrià Arrufat
58143ee3d1 Fix event order and add tests
- Fix setChecked event order: click fires before input/change to match
  browser behavior
- Add tests for hover, press, selectOption, setChecked MCP tools
- Merge all action tests into a single test case sharing one page load
- Add test elements to mcp_actions.html (hover target, key input,
  second select, checkbox, radio)
2026-04-02 10:46:28 +02:00
Adrià Arrufat
5e79af42f4 mcp: Add hover, press, selectOption, setChecked
New browser actions and MCP tools for AI agent interaction:
- hover: dispatches mouseover/mouseenter events on an element
- press: dispatches keydown/keyup keyboard events (Enter, Tab, etc.)
- selectOption: selects a dropdown option by value with input/change events
- setChecked: checks/unchecks checkbox or radio with input/change/click events
2026-04-02 09:47:22 +02:00
4 changed files with 528 additions and 50 deletions

View File

@@ -22,10 +22,23 @@ const DOMNode = @import("webapi/Node.zig");
const Element = @import("webapi/Element.zig"); const Element = @import("webapi/Element.zig");
const Event = @import("webapi/Event.zig"); const Event = @import("webapi/Event.zig");
const MouseEvent = @import("webapi/event/MouseEvent.zig"); const MouseEvent = @import("webapi/event/MouseEvent.zig");
const KeyboardEvent = @import("webapi/event/KeyboardEvent.zig");
const Page = @import("Page.zig"); const Page = @import("Page.zig");
const Session = @import("Session.zig"); const Session = @import("Session.zig");
const Selector = @import("webapi/selector/Selector.zig"); const Selector = @import("webapi/selector/Selector.zig");
fn dispatchInputAndChangeEvents(el: *Element, page: *Page) !void {
const input_evt: *Event = try .initTrusted(comptime .wrap("input"), .{ .bubbles = true }, page);
page._event_manager.dispatch(el.asEventTarget(), input_evt) catch |err| {
lp.log.err(.app, "dispatch input event failed", .{ .err = err });
};
const change_evt: *Event = try .initTrusted(comptime .wrap("change"), .{ .bubbles = true }, page);
page._event_manager.dispatch(el.asEventTarget(), change_evt) catch |err| {
lp.log.err(.app, "dispatch change event failed", .{ .err = err });
};
}
pub fn click(node: *DOMNode, page: *Page) !void { pub fn click(node: *DOMNode, page: *Page) !void {
const el = node.is(Element) orelse return error.InvalidNodeType; const el = node.is(Element) orelse return error.InvalidNodeType;
@@ -43,9 +56,107 @@ pub fn click(node: *DOMNode, page: *Page) !void {
}; };
} }
pub fn hover(node: *DOMNode, page: *Page) !void {
const el = node.is(Element) orelse return error.InvalidNodeType;
const mouseover_event: *MouseEvent = try .initTrusted(comptime .wrap("mouseover"), .{
.bubbles = true,
.cancelable = true,
.composed = true,
}, page);
page._event_manager.dispatch(el.asEventTarget(), mouseover_event.asEvent()) catch |err| {
lp.log.err(.app, "hover mouseover failed", .{ .err = err });
return error.ActionFailed;
};
const mouseenter_event: *MouseEvent = try .initTrusted(comptime .wrap("mouseenter"), .{
.composed = true,
}, page);
page._event_manager.dispatch(el.asEventTarget(), mouseenter_event.asEvent()) catch |err| {
lp.log.err(.app, "hover mouseenter failed", .{ .err = err });
return error.ActionFailed;
};
}
pub fn press(node: ?*DOMNode, key: []const u8, page: *Page) !void {
const target = if (node) |n|
(n.is(Element) orelse return error.InvalidNodeType).asEventTarget()
else
page.document.asNode().asEventTarget();
const keydown_event: *KeyboardEvent = try .initTrusted(comptime .wrap("keydown"), .{
.bubbles = true,
.cancelable = true,
.composed = true,
.key = key,
}, page);
page._event_manager.dispatch(target, keydown_event.asEvent()) catch |err| {
lp.log.err(.app, "press keydown failed", .{ .err = err });
return error.ActionFailed;
};
const keyup_event: *KeyboardEvent = try .initTrusted(comptime .wrap("keyup"), .{
.bubbles = true,
.cancelable = true,
.composed = true,
.key = key,
}, page);
page._event_manager.dispatch(target, keyup_event.asEvent()) catch |err| {
lp.log.err(.app, "press keyup failed", .{ .err = err });
return error.ActionFailed;
};
}
pub fn selectOption(node: *DOMNode, value: []const u8, page: *Page) !void {
const el = node.is(Element) orelse return error.InvalidNodeType;
const select = el.is(Element.Html.Select) orelse return error.InvalidNodeType;
select.setValue(value, page) catch |err| {
lp.log.err(.app, "select setValue failed", .{ .err = err });
return error.ActionFailed;
};
try dispatchInputAndChangeEvents(el, page);
}
pub fn setChecked(node: *DOMNode, checked: bool, page: *Page) !void {
const el = node.is(Element) orelse return error.InvalidNodeType;
const input = el.is(Element.Html.Input) orelse return error.InvalidNodeType;
if (input._input_type != .checkbox and input._input_type != .radio) {
return error.InvalidNodeType;
}
input.setChecked(checked, page) catch |err| {
lp.log.err(.app, "setChecked failed", .{ .err = err });
return error.ActionFailed;
};
// Match browser event order: click fires first, then input and change.
const click_event: *MouseEvent = try .initTrusted(comptime .wrap("click"), .{
.bubbles = true,
.cancelable = true,
.composed = true,
}, page);
page._event_manager.dispatch(el.asEventTarget(), click_event.asEvent()) catch |err| {
lp.log.err(.app, "dispatch click event failed", .{ .err = err });
};
try dispatchInputAndChangeEvents(el, page);
}
pub fn fill(node: *DOMNode, text: []const u8, page: *Page) !void { pub fn fill(node: *DOMNode, text: []const u8, page: *Page) !void {
const el = node.is(Element) orelse return error.InvalidNodeType; const el = node.is(Element) orelse return error.InvalidNodeType;
el.focus(page) catch |err| {
lp.log.err(.app, "fill focus failed", .{ .err = err });
};
if (el.is(Element.Html.Input)) |input| { if (el.is(Element.Html.Input)) |input| {
input.setValue(text, page) catch |err| { input.setValue(text, page) catch |err| {
lp.log.err(.app, "fill input failed", .{ .err = err }); lp.log.err(.app, "fill input failed", .{ .err = err });
@@ -65,15 +176,7 @@ pub fn fill(node: *DOMNode, text: []const u8, page: *Page) !void {
return error.InvalidNodeType; return error.InvalidNodeType;
} }
const input_evt: *Event = try .initTrusted(comptime .wrap("input"), .{ .bubbles = true }, page); try dispatchInputAndChangeEvents(el, page);
page._event_manager.dispatch(el.asEventTarget(), input_evt) catch |err| {
lp.log.err(.app, "dispatch input event failed", .{ .err = err });
};
const change_evt: *Event = try .initTrusted(comptime .wrap("change"), .{ .bubbles = true }, page);
page._event_manager.dispatch(el.asEventTarget(), change_evt) catch |err| {
lp.log.err(.app, "dispatch change event failed", .{ .err = err });
};
} }
pub fn scroll(node: ?*DOMNode, x: ?i32, y: ?i32, page: *Page) !void { pub fn scroll(node: ?*DOMNode, x: ?i32, y: ?i32, page: *Page) !void {

View File

@@ -10,5 +10,20 @@
<div id="scrollbox" style="width: 100px; height: 100px; overflow: scroll;" onscroll="window.scrolled = true;"> <div id="scrollbox" style="width: 100px; height: 100px; overflow: scroll;" onscroll="window.scrolled = true;">
<div style="height: 500px;">Long content</div> <div style="height: 500px;">Long content</div>
</div> </div>
<div id="hoverTarget" onmouseover="window.hovered = true;">Hover Me</div>
<input id="keyTarget" onkeydown="window.keyPressed = event.key;" onkeyup="window.keyReleased = event.key;">
<select id="sel2" onchange="window.sel2Changed = this.value">
<option value="a">Alpha</option>
<option value="b">Beta</option>
<option value="c">Gamma</option>
</select>
<input id="chk" type="checkbox">
<input id="rad" type="radio" name="group1">
<script>
document.getElementById('chk').addEventListener('click', function() { window.chkClicked = true; });
document.getElementById('chk').addEventListener('change', function() { window.chkChanged = true; });
document.getElementById('rad').addEventListener('click', function() { window.radClicked = true; });
document.getElementById('rad').addEventListener('change', function() { window.radChanged = true; });
</script>
</body> </body>
</html> </html>

View File

@@ -52,17 +52,19 @@ fn dispatchKeyEvent(cmd: *CDP.Command) !void {
try cmd.sendResult(null, .{}); try cmd.sendResult(null, .{});
// quickly ignore types we know we don't handle // rawKeyDown is a Chrome-internal event type not used for JS dispatch
switch (params.type) { if (params.type == .rawKeyDown) return;
.keyUp, .rawKeyDown, .char => return,
.keyDown => {},
}
const bc = cmd.browser_context orelse return; const bc = cmd.browser_context orelse return;
const page = bc.session.currentPage() orelse return; const page = bc.session.currentPage() orelse return;
const KeyboardEvent = @import("../../browser/webapi/event/KeyboardEvent.zig"); const KeyboardEvent = @import("../../browser/webapi/event/KeyboardEvent.zig");
const keyboard_event = try KeyboardEvent.initTrusted(comptime .wrap("keydown"), .{ const keyboard_event = try KeyboardEvent.initTrusted(switch (params.type) {
.keyDown => comptime .wrap("keydown"),
.keyUp => comptime .wrap("keyup"),
.char => comptime .wrap("keypress"),
.rawKeyDown => unreachable,
}, .{
.key = params.key, .key = params.key,
.code = params.code, .code = params.code,
.altKey = params.modifiers & 1 == 1, .altKey = params.modifiers & 1 == 1,

View File

@@ -175,6 +175,74 @@ pub const tool_list = [_]protocol.Tool{
\\} \\}
), ),
}, },
.{
.name = "hover",
.description = "Hover over an element, triggering mouseover and mouseenter events. Useful for menus, tooltips, and hover states.",
.inputSchema = protocol.minify(
\\{
\\ "type": "object",
\\ "properties": {
\\ "backendNodeId": { "type": "integer", "description": "The backend node ID of the element to hover over." }
\\ },
\\ "required": ["backendNodeId"]
\\}
),
},
.{
.name = "press",
.description = "Press a keyboard key, dispatching keydown and keyup events. Use key names like 'Enter', 'Tab', 'Escape', 'ArrowDown', 'Backspace', or single characters like 'a', '1'.",
.inputSchema = protocol.minify(
\\{
\\ "type": "object",
\\ "properties": {
\\ "key": { "type": "string", "description": "The key to press (e.g. 'Enter', 'Tab', 'a')." },
\\ "backendNodeId": { "type": "integer", "description": "Optional backend node ID of the element to target. Defaults to the document." }
\\ },
\\ "required": ["key"]
\\}
),
},
.{
.name = "selectOption",
.description = "Select an option in a <select> dropdown element by its value. Dispatches input and change events.",
.inputSchema = protocol.minify(
\\{
\\ "type": "object",
\\ "properties": {
\\ "backendNodeId": { "type": "integer", "description": "The backend node ID of the <select> element." },
\\ "value": { "type": "string", "description": "The value of the option to select." }
\\ },
\\ "required": ["backendNodeId", "value"]
\\}
),
},
.{
.name = "setChecked",
.description = "Check or uncheck a checkbox or radio button. Dispatches input, change, and click events.",
.inputSchema = protocol.minify(
\\{
\\ "type": "object",
\\ "properties": {
\\ "backendNodeId": { "type": "integer", "description": "The backend node ID of the checkbox or radio input element." },
\\ "checked": { "type": "boolean", "description": "Whether to check (true) or uncheck (false) the element." }
\\ },
\\ "required": ["backendNodeId", "checked"]
\\}
),
},
.{
.name = "findElement",
.description = "Find interactive elements by role and/or accessible name. Returns matching elements with their backend node IDs. Useful for locating specific elements without parsing the full semantic tree.",
.inputSchema = protocol.minify(
\\{
\\ "type": "object",
\\ "properties": {
\\ "role": { "type": "string", "description": "Optional ARIA role to match (e.g. 'button', 'link', 'textbox', 'checkbox')." },
\\ "name": { "type": "string", "description": "Optional accessible name substring to match (case-insensitive)." }
\\ }
\\}
),
},
}; };
pub fn handleList(server: *Server, arena: std.mem.Allocator, req: protocol.Request) !void { pub fn handleList(server: *Server, arena: std.mem.Allocator, req: protocol.Request) !void {
@@ -282,6 +350,11 @@ const ToolAction = enum {
fill, fill,
scroll, scroll,
waitForSelector, waitForSelector,
hover,
press,
selectOption,
setChecked,
findElement,
}; };
const tool_map = std.StaticStringMap(ToolAction).initComptime(.{ const tool_map = std.StaticStringMap(ToolAction).initComptime(.{
@@ -300,6 +373,11 @@ const tool_map = std.StaticStringMap(ToolAction).initComptime(.{
.{ "fill", .fill }, .{ "fill", .fill },
.{ "scroll", .scroll }, .{ "scroll", .scroll },
.{ "waitForSelector", .waitForSelector }, .{ "waitForSelector", .waitForSelector },
.{ "hover", .hover },
.{ "press", .press },
.{ "selectOption", .selectOption },
.{ "setChecked", .setChecked },
.{ "findElement", .findElement },
}); });
pub fn handleCall(server: *Server, arena: std.mem.Allocator, req: protocol.Request) !void { pub fn handleCall(server: *Server, arena: std.mem.Allocator, req: protocol.Request) !void {
@@ -334,6 +412,11 @@ pub fn handleCall(server: *Server, arena: std.mem.Allocator, req: protocol.Reque
.fill => try handleFill(server, arena, req.id.?, call_params.arguments), .fill => try handleFill(server, arena, req.id.?, call_params.arguments),
.scroll => try handleScroll(server, arena, req.id.?, call_params.arguments), .scroll => try handleScroll(server, arena, req.id.?, call_params.arguments),
.waitForSelector => try handleWaitForSelector(server, arena, req.id.?, call_params.arguments), .waitForSelector => try handleWaitForSelector(server, arena, req.id.?, call_params.arguments),
.hover => try handleHover(server, arena, req.id.?, call_params.arguments),
.press => try handlePress(server, arena, req.id.?, call_params.arguments),
.selectOption => try handleSelectOption(server, arena, req.id.?, call_params.arguments),
.setChecked => try handleSetChecked(server, arena, req.id.?, call_params.arguments),
.findElement => try handleFindElement(server, arena, req.id.?, call_params.arguments),
} }
} }
@@ -400,17 +483,9 @@ fn handleNodeDetails(server: *Server, arena: std.mem.Allocator, id: std.json.Val
backendNodeId: CDPNode.Id, backendNodeId: CDPNode.Id,
}; };
const args = try parseArgs(Params, arena, arguments, server, id, "nodeDetails"); const args = try parseArgs(Params, arena, arguments, server, id, "nodeDetails");
const resolved = try resolveNodeAndPage(server, id, args.backendNodeId);
_ = server.session.currentPage() orelse { const details = lp.SemanticTree.getNodeDetails(arena, resolved.node, &server.node_registry, resolved.page) catch {
return server.sendError(id, .PageNotLoaded, "Page not loaded");
};
const node = server.node_registry.lookup_by_id.get(args.backendNodeId) orelse {
return server.sendError(id, .InvalidParams, "Node not found");
};
const page = server.session.currentPage().?;
const details = lp.SemanticTree.getNodeDetails(arena, node.dom, &server.node_registry, page) catch {
return server.sendError(id, .InternalError, "Failed to get node details"); return server.sendError(id, .InternalError, "Failed to get node details");
}; };
@@ -510,26 +585,19 @@ fn handleClick(server: *Server, arena: std.mem.Allocator, id: std.json.Value, ar
backendNodeId: CDPNode.Id, backendNodeId: CDPNode.Id,
}; };
const args = try parseArgs(ClickParams, arena, arguments, server, id, "click"); const args = try parseArgs(ClickParams, arena, arguments, server, id, "click");
const resolved = try resolveNodeAndPage(server, id, args.backendNodeId);
const page = server.session.currentPage() orelse { lp.actions.click(resolved.node, resolved.page) catch |err| {
return server.sendError(id, .PageNotLoaded, "Page not loaded");
};
const node = server.node_registry.lookup_by_id.get(args.backendNodeId) orelse {
return server.sendError(id, .InvalidParams, "Node not found");
};
lp.actions.click(node.dom, page) catch |err| {
if (err == error.InvalidNodeType) { if (err == error.InvalidNodeType) {
return server.sendError(id, .InvalidParams, "Node is not an HTML element"); return server.sendError(id, .InvalidParams, "Node is not an HTML element");
} }
return server.sendError(id, .InternalError, "Failed to click element"); return server.sendError(id, .InternalError, "Failed to click element");
}; };
const page_title = page.getTitle() catch null; const page_title = resolved.page.getTitle() catch null;
const result_text = try std.fmt.allocPrint(arena, "Clicked element (backendNodeId: {d}). Page url: {s}, title: {s}", .{ const result_text = try std.fmt.allocPrint(arena, "Clicked element (backendNodeId: {d}). Page url: {s}, title: {s}", .{
args.backendNodeId, args.backendNodeId,
page.url, resolved.page.url,
page_title orelse "(none)", page_title orelse "(none)",
}); });
const content = [_]protocol.TextContent([]const u8){.{ .text = result_text }}; const content = [_]protocol.TextContent([]const u8){.{ .text = result_text }};
@@ -542,27 +610,20 @@ fn handleFill(server: *Server, arena: std.mem.Allocator, id: std.json.Value, arg
text: []const u8, text: []const u8,
}; };
const args = try parseArgs(FillParams, arena, arguments, server, id, "fill"); const args = try parseArgs(FillParams, arena, arguments, server, id, "fill");
const resolved = try resolveNodeAndPage(server, id, args.backendNodeId);
const page = server.session.currentPage() orelse { lp.actions.fill(resolved.node, args.text, resolved.page) catch |err| {
return server.sendError(id, .PageNotLoaded, "Page not loaded");
};
const node = server.node_registry.lookup_by_id.get(args.backendNodeId) orelse {
return server.sendError(id, .InvalidParams, "Node not found");
};
lp.actions.fill(node.dom, args.text, page) catch |err| {
if (err == error.InvalidNodeType) { if (err == error.InvalidNodeType) {
return server.sendError(id, .InvalidParams, "Node is not an input, textarea or select"); return server.sendError(id, .InvalidParams, "Node is not an input, textarea or select");
} }
return server.sendError(id, .InternalError, "Failed to fill element"); return server.sendError(id, .InternalError, "Failed to fill element");
}; };
const page_title = page.getTitle() catch null; const page_title = resolved.page.getTitle() catch null;
const result_text = try std.fmt.allocPrint(arena, "Filled element (backendNodeId: {d}) with \"{s}\". Page url: {s}, title: {s}", .{ const result_text = try std.fmt.allocPrint(arena, "Filled element (backendNodeId: {d}) with \"{s}\". Page url: {s}, title: {s}", .{
args.backendNodeId, args.backendNodeId,
args.text, args.text,
page.url, resolved.page.url,
page_title orelse "(none)", page_title orelse "(none)",
}); });
const content = [_]protocol.TextContent([]const u8){.{ .text = result_text }}; const content = [_]protocol.TextContent([]const u8){.{ .text = result_text }};
@@ -636,6 +697,189 @@ fn handleWaitForSelector(server: *Server, arena: std.mem.Allocator, id: std.json
return server.sendResult(id, protocol.CallToolResult([]const u8){ .content = &content }); return server.sendResult(id, protocol.CallToolResult([]const u8){ .content = &content });
} }
fn handleHover(server: *Server, arena: std.mem.Allocator, id: std.json.Value, arguments: ?std.json.Value) !void {
const Params = struct {
backendNodeId: CDPNode.Id,
};
const args = try parseArgs(Params, arena, arguments, server, id, "hover");
const resolved = try resolveNodeAndPage(server, id, args.backendNodeId);
lp.actions.hover(resolved.node, resolved.page) catch |err| {
if (err == error.InvalidNodeType) {
return server.sendError(id, .InvalidParams, "Node is not an HTML element");
}
return server.sendError(id, .InternalError, "Failed to hover element");
};
const page_title = resolved.page.getTitle() catch null;
const result_text = try std.fmt.allocPrint(arena, "Hovered element (backendNodeId: {d}). Page url: {s}, title: {s}", .{
args.backendNodeId,
resolved.page.url,
page_title orelse "(none)",
});
const content = [_]protocol.TextContent([]const u8){.{ .text = result_text }};
try server.sendResult(id, protocol.CallToolResult([]const u8){ .content = &content });
}
fn handlePress(server: *Server, arena: std.mem.Allocator, id: std.json.Value, arguments: ?std.json.Value) !void {
const Params = struct {
key: []const u8,
backendNodeId: ?CDPNode.Id = null,
};
const args = try parseArgs(Params, arena, arguments, server, id, "press");
const page = server.session.currentPage() orelse {
return server.sendError(id, .PageNotLoaded, "Page not loaded");
};
var target_node: ?*DOMNode = null;
if (args.backendNodeId) |node_id| {
const node = server.node_registry.lookup_by_id.get(node_id) orelse {
return server.sendError(id, .InvalidParams, "Node not found");
};
target_node = node.dom;
}
lp.actions.press(target_node, args.key, page) catch |err| {
if (err == error.InvalidNodeType) {
return server.sendError(id, .InvalidParams, "Node is not an HTML element");
}
return server.sendError(id, .InternalError, "Failed to press key");
};
const page_title = page.getTitle() catch null;
const result_text = try std.fmt.allocPrint(arena, "Pressed key '{s}'. Page url: {s}, title: {s}", .{
args.key,
page.url,
page_title orelse "(none)",
});
const content = [_]protocol.TextContent([]const u8){.{ .text = result_text }};
try server.sendResult(id, protocol.CallToolResult([]const u8){ .content = &content });
}
fn handleSelectOption(server: *Server, arena: std.mem.Allocator, id: std.json.Value, arguments: ?std.json.Value) !void {
const Params = struct {
backendNodeId: CDPNode.Id,
value: []const u8,
};
const args = try parseArgs(Params, arena, arguments, server, id, "selectOption");
const resolved = try resolveNodeAndPage(server, id, args.backendNodeId);
lp.actions.selectOption(resolved.node, args.value, resolved.page) catch |err| {
if (err == error.InvalidNodeType) {
return server.sendError(id, .InvalidParams, "Node is not a <select> element");
}
return server.sendError(id, .InternalError, "Failed to select option");
};
const page_title = resolved.page.getTitle() catch null;
const result_text = try std.fmt.allocPrint(arena, "Selected option '{s}' (backendNodeId: {d}). Page url: {s}, title: {s}", .{
args.value,
args.backendNodeId,
resolved.page.url,
page_title orelse "(none)",
});
const content = [_]protocol.TextContent([]const u8){.{ .text = result_text }};
try server.sendResult(id, protocol.CallToolResult([]const u8){ .content = &content });
}
fn handleSetChecked(server: *Server, arena: std.mem.Allocator, id: std.json.Value, arguments: ?std.json.Value) !void {
const Params = struct {
backendNodeId: CDPNode.Id,
checked: bool,
};
const args = try parseArgs(Params, arena, arguments, server, id, "setChecked");
const resolved = try resolveNodeAndPage(server, id, args.backendNodeId);
lp.actions.setChecked(resolved.node, args.checked, resolved.page) catch |err| {
if (err == error.InvalidNodeType) {
return server.sendError(id, .InvalidParams, "Node is not a checkbox or radio input");
}
return server.sendError(id, .InternalError, "Failed to set checked state");
};
const state_str = if (args.checked) "checked" else "unchecked";
const page_title = resolved.page.getTitle() catch null;
const result_text = try std.fmt.allocPrint(arena, "Set element (backendNodeId: {d}) to {s}. Page url: {s}, title: {s}", .{
args.backendNodeId,
state_str,
resolved.page.url,
page_title orelse "(none)",
});
const content = [_]protocol.TextContent([]const u8){.{ .text = result_text }};
try server.sendResult(id, protocol.CallToolResult([]const u8){ .content = &content });
}
fn handleFindElement(server: *Server, arena: std.mem.Allocator, id: std.json.Value, arguments: ?std.json.Value) !void {
const Params = struct {
role: ?[]const u8 = null,
name: ?[]const u8 = null,
};
const args = try parseArgsOrDefault(Params, arena, arguments, server, id);
if (args.role == null and args.name == null) {
return server.sendError(id, .InvalidParams, "At least one of 'role' or 'name' must be provided");
}
const page = server.session.currentPage() orelse {
return server.sendError(id, .PageNotLoaded, "Page not loaded");
};
const elements = lp.interactive.collectInteractiveElements(page.document.asNode(), arena, page) catch |err| {
log.err(.mcp, "elements collection failed", .{ .err = err });
return server.sendError(id, .InternalError, "Failed to collect interactive elements");
};
var matches: std.ArrayList(lp.interactive.InteractiveElement) = .empty;
for (elements) |el| {
if (args.role) |role| {
const el_role = el.role orelse continue;
if (!std.ascii.eqlIgnoreCase(el_role, role)) continue;
}
if (args.name) |name| {
const el_name = el.name orelse continue;
if (!containsIgnoreCase(el_name, name)) continue;
}
try matches.append(arena, el);
}
const matched = try matches.toOwnedSlice(arena);
lp.interactive.registerNodes(matched, &server.node_registry) catch |err| {
log.err(.mcp, "node registration failed", .{ .err = err });
return server.sendError(id, .InternalError, "Failed to register element nodes");
};
var aw: std.Io.Writer.Allocating = .init(arena);
try std.json.Stringify.value(matched, .{}, &aw.writer);
const content = [_]protocol.TextContent([]const u8){.{ .text = aw.written() }};
try server.sendResult(id, protocol.CallToolResult([]const u8){ .content = &content });
}
fn containsIgnoreCase(haystack: []const u8, needle: []const u8) bool {
if (needle.len > haystack.len) return false;
if (needle.len == 0) return true;
const end = haystack.len - needle.len + 1;
for (0..end) |i| {
if (std.ascii.eqlIgnoreCase(haystack[i..][0..needle.len], needle)) return true;
}
return false;
}
const NodeAndPage = struct { node: *DOMNode, page: *lp.Page };
fn resolveNodeAndPage(server: *Server, id: std.json.Value, node_id: CDPNode.Id) !NodeAndPage {
const page = server.session.currentPage() orelse {
try server.sendError(id, .PageNotLoaded, "Page not loaded");
return error.PageNotLoaded;
};
const node = server.node_registry.lookup_by_id.get(node_id) orelse {
try server.sendError(id, .InvalidParams, "Node not found");
return error.InvalidParams;
};
return .{ .node = node.dom, .page = page };
}
fn ensurePage(server: *Server, id: std.json.Value, url: ?[:0]const u8, timeout: ?u32, waitUntil: ?lp.Config.WaitUntil) !*lp.Page { fn ensurePage(server: *Server, id: std.json.Value, url: ?[:0]const u8, timeout: ?u32, waitUntil: ?lp.Config.WaitUntil) !*lp.Page {
if (url) |u| { if (url) |u| {
try performGoto(server, u, id, timeout, waitUntil); try performGoto(server, u, id, timeout, waitUntil);
@@ -736,7 +980,7 @@ test "MCP - evaluate error reporting" {
} }, out.written()); } }, out.written());
} }
test "MCP - Actions: click, fill, scroll" { test "MCP - Actions: click, fill, scroll, hover, press, selectOption, setChecked" {
defer testing.reset(); defer testing.reset();
const aa = testing.arena_allocator; const aa = testing.arena_allocator;
@@ -797,7 +1041,67 @@ test "MCP - Actions: click, fill, scroll" {
out.clearRetainingCapacity(); out.clearRetainingCapacity();
} }
// Evaluate assertions {
// Test Hover
const el = page.document.getElementById("hoverTarget", page).?.asNode();
const el_id = (try server.node_registry.register(el)).id;
var id_buf: [12]u8 = undefined;
const id_str = std.fmt.bufPrint(&id_buf, "{d}", .{el_id}) catch unreachable;
const msg = try std.mem.concat(aa, u8, &.{ "{\"jsonrpc\":\"2.0\",\"id\":5,\"method\":\"tools/call\",\"params\":{\"name\":\"hover\",\"arguments\":{\"backendNodeId\":", id_str, "}}}" });
try router.handleMessage(server, aa, msg);
try testing.expect(std.mem.indexOf(u8, out.written(), "Hovered element") != null);
out.clearRetainingCapacity();
}
{
// Test Press
const el = page.document.getElementById("keyTarget", page).?.asNode();
const el_id = (try server.node_registry.register(el)).id;
var id_buf: [12]u8 = undefined;
const id_str = std.fmt.bufPrint(&id_buf, "{d}", .{el_id}) catch unreachable;
const msg = try std.mem.concat(aa, u8, &.{ "{\"jsonrpc\":\"2.0\",\"id\":6,\"method\":\"tools/call\",\"params\":{\"name\":\"press\",\"arguments\":{\"key\":\"Enter\",\"backendNodeId\":", id_str, "}}}" });
try router.handleMessage(server, aa, msg);
try testing.expect(std.mem.indexOf(u8, out.written(), "Pressed key") != null);
out.clearRetainingCapacity();
}
{
// Test SelectOption
const el = page.document.getElementById("sel2", page).?.asNode();
const el_id = (try server.node_registry.register(el)).id;
var id_buf: [12]u8 = undefined;
const id_str = std.fmt.bufPrint(&id_buf, "{d}", .{el_id}) catch unreachable;
const msg = try std.mem.concat(aa, u8, &.{ "{\"jsonrpc\":\"2.0\",\"id\":7,\"method\":\"tools/call\",\"params\":{\"name\":\"selectOption\",\"arguments\":{\"backendNodeId\":", id_str, ",\"value\":\"b\"}}}" });
try router.handleMessage(server, aa, msg);
try testing.expect(std.mem.indexOf(u8, out.written(), "Selected option") != null);
out.clearRetainingCapacity();
}
{
// Test SetChecked (checkbox)
const el = page.document.getElementById("chk", page).?.asNode();
const el_id = (try server.node_registry.register(el)).id;
var id_buf: [12]u8 = undefined;
const id_str = std.fmt.bufPrint(&id_buf, "{d}", .{el_id}) catch unreachable;
const msg = try std.mem.concat(aa, u8, &.{ "{\"jsonrpc\":\"2.0\",\"id\":8,\"method\":\"tools/call\",\"params\":{\"name\":\"setChecked\",\"arguments\":{\"backendNodeId\":", id_str, ",\"checked\":true}}}" });
try router.handleMessage(server, aa, msg);
try testing.expect(std.mem.indexOf(u8, out.written(), "checked") != null);
out.clearRetainingCapacity();
}
{
// Test SetChecked (radio)
const el = page.document.getElementById("rad", page).?.asNode();
const el_id = (try server.node_registry.register(el)).id;
var id_buf: [12]u8 = undefined;
const id_str = std.fmt.bufPrint(&id_buf, "{d}", .{el_id}) catch unreachable;
const msg = try std.mem.concat(aa, u8, &.{ "{\"jsonrpc\":\"2.0\",\"id\":9,\"method\":\"tools/call\",\"params\":{\"name\":\"setChecked\",\"arguments\":{\"backendNodeId\":", id_str, ",\"checked\":true}}}" });
try router.handleMessage(server, aa, msg);
try testing.expect(std.mem.indexOf(u8, out.written(), "checked") != null);
out.clearRetainingCapacity();
}
// Evaluate JS assertions for all actions
var ls: js.Local.Scope = undefined; var ls: js.Local.Scope = undefined;
page.js.localScope(&ls); page.js.localScope(&ls);
defer ls.deinit(); defer ls.deinit();
@@ -809,12 +1113,66 @@ test "MCP - Actions: click, fill, scroll" {
const result = try ls.local.exec( const result = try ls.local.exec(
\\ window.clicked === true && window.inputVal === 'hello' && \\ window.clicked === true && window.inputVal === 'hello' &&
\\ window.changed === true && window.selChanged === 'opt2' && \\ window.changed === true && window.selChanged === 'opt2' &&
\\ window.scrolled === true \\ window.scrolled === true &&
\\ window.hovered === true &&
\\ window.keyPressed === 'Enter' && window.keyReleased === 'Enter' &&
\\ window.sel2Changed === 'b' &&
\\ window.chkClicked === true && window.chkChanged === true &&
\\ window.radClicked === true && window.radChanged === true
, null); , null);
try testing.expect(result.isTrue()); try testing.expect(result.isTrue());
} }
test "MCP - findElement" {
defer testing.reset();
const aa = testing.arena_allocator;
var out: std.io.Writer.Allocating = .init(aa);
const server = try testLoadPage("http://localhost:9582/src/browser/tests/mcp_actions.html", &out.writer);
defer server.deinit();
{
// Find by role
const msg =
\\{"jsonrpc":"2.0","id":1,"method":"tools/call","params":{"name":"findElement","arguments":{"role":"button"}}}
;
try router.handleMessage(server, aa, msg);
try testing.expect(std.mem.indexOf(u8, out.written(), "Click Me") != null);
out.clearRetainingCapacity();
}
{
// Find by name (case-insensitive substring)
const msg =
\\{"jsonrpc":"2.0","id":2,"method":"tools/call","params":{"name":"findElement","arguments":{"name":"click"}}}
;
try router.handleMessage(server, aa, msg);
try testing.expect(std.mem.indexOf(u8, out.written(), "Click Me") != null);
out.clearRetainingCapacity();
}
{
// Find with no matches
const msg =
\\{"jsonrpc":"2.0","id":3,"method":"tools/call","params":{"name":"findElement","arguments":{"role":"slider"}}}
;
try router.handleMessage(server, aa, msg);
try testing.expect(std.mem.indexOf(u8, out.written(), "[]") != null);
out.clearRetainingCapacity();
}
{
// Error: no params provided
const msg =
\\{"jsonrpc":"2.0","id":4,"method":"tools/call","params":{"name":"findElement","arguments":{}}}
;
try router.handleMessage(server, aa, msg);
try testing.expect(std.mem.indexOf(u8, out.written(), "error") != null);
out.clearRetainingCapacity();
}
}
test "MCP - waitForSelector: existing element" { test "MCP - waitForSelector: existing element" {
defer testing.reset(); defer testing.reset();
var out: std.io.Writer.Allocating = .init(testing.arena_allocator); var out: std.io.Writer.Allocating = .init(testing.arena_allocator);