browser: normalize keys and support form submit in press

- Map common key shorthand aliases to canonical KeyboardEvent keys.
- Trigger implicit form submission when pressing Enter on inputs.
- Add CSS selector support to the press tool.

src/browser/actions.zig

@@ -80,16 +80,18 @@ pub fn hover(node: *DOMNode, frame: *Frame) !void {
 }
 
 pub fn press(node: ?*DOMNode, key: []const u8, frame: *Frame) !void {
-    const target = if (node) |n|
-        (n.is(Element) orelse return error.InvalidNodeType).asEventTarget()
+    const target_el: ?*Element = if (node) |n|
+        (n.is(Element) orelse return error.InvalidNodeType)
     else
-        frame.document.asNode().asEventTarget();
+        null;
+    const target = if (target_el) |el| el.asEventTarget() else frame.document.asNode().asEventTarget();
+    const canonical = canonicalKey(key);
 
     const keydown_event: *KeyboardEvent = try .initTrusted(comptime .wrap("keydown"), .{
         .bubbles = true,
         .cancelable = true,
         .composed = true,
-        .key = key,
+        .key = canonical,
     }, frame);
 
     frame._event_manager.dispatch(target, keydown_event.asEvent()) catch |err| {
@@ -97,11 +99,15 @@ pub fn press(node: ?*DOMNode, key: []const u8, frame: *Frame) !void {
         return error.ActionFailed;
     };
 
+    if (std.mem.eql(u8, canonical, "Enter") and !keydown_event.asEvent().getDefaultPrevented()) {
+        if (target_el) |el| try implicitFormSubmit(el, frame);
+    }
+
     const keyup_event: *KeyboardEvent = try .initTrusted(comptime .wrap("keyup"), .{
         .bubbles = true,
         .cancelable = true,
         .composed = true,
-        .key = key,
+        .key = canonical,
     }, frame);
 
     frame._event_manager.dispatch(target, keyup_event.asEvent()) catch |err| {
@@ -110,6 +116,55 @@ pub fn press(node: ?*DOMNode, key: []const u8, frame: *Frame) !void {
     };
 }
 
+/// Map common shorthand to the canonical KeyboardEvent.key string so users
+/// can type "enter" instead of "Enter" without surprises.
+fn canonicalKey(key: []const u8) []const u8 {
+    const aliases = [_]struct { in: []const u8, out: []const u8 }{
+        .{ .in = "enter", .out = "Enter" },
+        .{ .in = "return", .out = "Enter" },
+        .{ .in = "\n", .out = "Enter" },
+        .{ .in = "\\n", .out = "Enter" },
+        .{ .in = "esc", .out = "Escape" },
+        .{ .in = "escape", .out = "Escape" },
+        .{ .in = "tab", .out = "Tab" },
+        .{ .in = "\t", .out = "Tab" },
+        .{ .in = "space", .out = " " },
+        .{ .in = "backspace", .out = "Backspace" },
+        .{ .in = "delete", .out = "Delete" },
+        .{ .in = "del", .out = "Delete" },
+        .{ .in = "up", .out = "ArrowUp" },
+        .{ .in = "down", .out = "ArrowDown" },
+        .{ .in = "left", .out = "ArrowLeft" },
+        .{ .in = "right", .out = "ArrowRight" },
+    };
+    for (aliases) |a| {
+        if (std.ascii.eqlIgnoreCase(key, a.in)) return a.out;
+    }
+    return key;
+}
+
+fn implicitFormSubmit(el: *Element, frame: *Frame) !void {
+    const Input = Element.Html.Input;
+    const Button = Element.Html.Button;
+
+    if (el.is(Input)) |input| {
+        const form = input.getForm(frame) orelse return;
+        const submitter: ?*Element = switch (input._input_type) {
+            .submit, .image => el,
+            // Non-text controls (checkbox, radio, file, ...) don't trigger
+            // implicit submission; only the text-like family does.
+            .text, .password, .email, .url, .tel, .search, .number, .date, .time, .@"datetime-local", .month, .week => null,
+            else => return,
+        };
+        return form.requestSubmit(submitter, frame);
+    }
+    if (el.is(Button)) |button| {
+        if (!std.mem.eql(u8, button.getType(), "submit")) return;
+        const form = button.getForm(frame) orelse return;
+        return form.requestSubmit(el, frame);
+    }
+}
+
 pub fn selectOption(node: *DOMNode, value: []const u8, frame: *Frame) !void {
     const el = node.is(Element) orelse return error.InvalidNodeType;
     const select = el.is(Element.Html.Select) orelse return error.InvalidNodeType;

src/browser/tests/mcp_press_form.html

@@ -0,0 +1,9 @@
+<!DOCTYPE html>
+<html>
+<body>
+    <form id="f" onsubmit="window.submitted = true; window.submittedValue = document.getElementById('q').value; event.preventDefault();">
+        <input id="q" type="text" name="q">
+        <input type="submit" value="Go">
+    </form>
+</body>
+</html>

src/browser/tools.zig

@@ -302,13 +302,14 @@ pub const Tool = enum {
                 ),
             },
             .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'.",
+                .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'. Common shorthand is normalized: 'enter'/'return' → 'Enter', 'esc' → 'Escape', 'up'/'down'/'left'/'right' → 'Arrow*', 'space' → ' '. Pressing 'Enter' on a form input or submit button triggers implicit form submission.",
                 .input_schema = 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." }
+                    \\    "selector": { "type": "string", "description": "Optional CSS selector of the element to target. Preferred over backendNodeId." },
+                    \\    "backendNodeId": { "type": "integer", "description": "Optional backend node ID of the element to target. Defaults to the document when neither selector nor backendNodeId is provided." }
                     \\  },
                     \\  "required": ["key"]
                     \\}
@@ -1026,12 +1027,21 @@ fn execHover(arena: std.mem.Allocator, session: *lp.Session, registry: *CDPNode.
 fn execPress(arena: std.mem.Allocator, session: *lp.Session, registry: *CDPNode.Registry, arguments: ?std.json.Value) ToolError![]const u8 {
     const Params = struct {
         key: []const u8,
+        selector: ?[]const u8 = null,
         backendNodeId: ?CDPNode.Id = null,
     };
     const args = try parseArgs(Params, arena, arguments);
 
-    const page = try requireFrame(session);
-    const target_node = try resolveOptionalNode(registry, args.backendNodeId);
+    var page: *lp.Frame = undefined;
+    var target_node: ?*DOMNode = null;
+    if (args.selector) |sel| {
+        const resolved = try resolveBySelector(session, sel);
+        page = resolved.page;
+        target_node = resolved.node;
+    } else {
+        page = try requireFrame(session);
+        target_node = try resolveOptionalNode(registry, args.backendNodeId);
+    }
 
     lp.actions.press(target_node, args.key, page) catch |err| return mapActionError(err);
 

src/mcp/tools.zig

@@ -898,6 +898,28 @@ test "MCP - waitForSelector: timeout" {
     }, out.written());
 }
 
+test "MCP - press Enter on form input triggers submit (lowercase alias)" {
+    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_press_form.html", &out.writer);
+    defer server.deinit();
+
+    // Fill the input then press "enter" (lowercase alias) on it. The form's
+    // submit handler sets window.submitted and snapshots the input value.
+    const fill = try aa.dupe(u8, "{\"jsonrpc\":\"2.0\",\"id\":1,\"method\":\"tools/call\",\"params\":{\"name\":\"fill\",\"arguments\":{\"selector\":\"#q\",\"value\":\"hello\"}}}");
+    try router.handleMessage(server, aa, fill);
+    out.clearRetainingCapacity();
+
+    const press_msg = try aa.dupe(u8, "{\"jsonrpc\":\"2.0\",\"id\":2,\"method\":\"tools/call\",\"params\":{\"name\":\"press\",\"arguments\":{\"selector\":\"#q\",\"key\":\"enter\"}}}");
+    try router.handleMessage(server, aa, press_msg);
+    out.clearRetainingCapacity();
+
+    const eval_msg = try aa.dupe(u8, "{\"jsonrpc\":\"2.0\",\"id\":3,\"method\":\"tools/call\",\"params\":{\"name\":\"eval\",\"arguments\":{\"script\":\"window.submitted === true && window.submittedValue === 'hello'\"}}}");
+    try router.handleMessage(server, aa, eval_msg);
+    try testing.expect(std.mem.indexOf(u8, out.written(), "true") != null);
+}
+
 test "MCP - getCookies: defaults to current page, url filter, all flag" {
     defer testing.reset();
     var out: std.io.Writer.Allocating = .init(testing.arena_allocator);