Merge feat/new-tools-and-richer-list: 4 tools nuevas + browser_list enriquecido

2026-06-16 20:25:35 +02:00
parent 15949bf4ed 6b7f71c39f
commit 82993179dc
3 changed files with 186 additions and 2 deletions
@@ -27,9 +27,74 @@ func registerDomTools(s *server.MCPServer, d *deps) {
 		s.AddTool(domTypeRefTool(), mcp.NewTypedToolHandler(d.handleDomTypeRef))
 		s.AddTool(domHoverRefTool(), mcp.NewTypedToolHandler(d.handleDomHoverRef))
 		s.AddTool(domClickXYTool(), mcp.NewTypedToolHandler(d.handleDomClickXY))
+		s.AddTool(domSelectOptionTool(), mcp.NewTypedToolHandler(d.handleDomSelectOption))
+		s.AddTool(domSetFilesTool(), mcp.NewTypedToolHandler(d.handleDomSetFiles))
 	}
 }

+// ---- dom_select_option (MUTA) ----
+
+type domSelectOptionArgs struct {
+	Port     int    `json:"port"`
+	Selector string `json:"selector"`
+	Value    string `json:"value"`
+}
+
+func domSelectOptionTool() mcp.Tool {
+	return mcp.NewTool("dom_select_option",
+		mcp.WithDescription("Select an <option> in a native <select> element (by CSS selector), matching by option value first, then by visible text, and firing input/change events so React/Vue react. For custom (non-<select>) dropdowns use dom_click_ref on the trigger then on the option instead."),
+		mcp.WithNumber("port", mcp.Description("CDP port. Default 9333 (Chrome isolated del MCP); usa 9222 explícito solo para adjuntarte al navegador diario.")),
+		mcp.WithString("selector", mcp.Required(), mcp.Description("CSS selector of the <select> element.")),
+		mcp.WithString("value", mcp.Required(), mcp.Description("Option value (or visible text if no value matches).")),
+	)
+}
+
+func (d *deps) handleDomSelectOption(_ context.Context, _ mcp.CallToolRequest, a domSelectOptionArgs) (*mcp.CallToolResult, error) {
+	if a.Selector == "" || a.Value == "" {
+		return mcp.NewToolResultError("selector and value are required"), nil
+	}
+	err := d.withConn(portOr(a.Port), func(c *browser.CDPConn) error {
+		return browser.CdpSelectOption(c, a.Selector, a.Value)
+	})
+	if err != nil {
+		return mcp.NewToolResultError(err.Error()), nil
+	}
+	return mcp.NewToolResultText(fmt.Sprintf("selected %q in %s", a.Value, a.Selector)), nil
+}
+
+// ---- dom_set_files (MUTA) ----
+
+type domSetFilesArgs struct {
+	Port     int      `json:"port"`
+	Selector string   `json:"selector"`
+	Paths    []string `json:"paths"`
+}
+
+func domSetFilesTool() mcp.Tool {
+	return mcp.NewTool("dom_set_files",
+		mcp.WithDescription("Upload files to an <input type=\"file\"> (by CSS selector) via DOM.setFileInputFiles, without driving the OS file picker. Paths must be absolute and readable by the Chrome process."),
+		mcp.WithNumber("port", mcp.Description("CDP port. Default 9333 (Chrome isolated del MCP); usa 9222 explícito solo para adjuntarte al navegador diario.")),
+		mcp.WithString("selector", mcp.Required(), mcp.Description("CSS selector of the file input element.")),
+		mcp.WithArray("paths", mcp.Required(), mcp.Description("Absolute file paths to attach."), mcp.Items(map[string]any{"type": "string"})),
+	)
+}
+
+func (d *deps) handleDomSetFiles(_ context.Context, _ mcp.CallToolRequest, a domSetFilesArgs) (*mcp.CallToolResult, error) {
+	if a.Selector == "" {
+		return mcp.NewToolResultError("selector is required"), nil
+	}
+	if len(a.Paths) == 0 {
+		return mcp.NewToolResultError("paths is required (at least one file)"), nil
+	}
+	err := d.withConn(portOr(a.Port), func(c *browser.CDPConn) error {
+		return browser.CdpSetFileInput(c, a.Selector, a.Paths)
+	})
+	if err != nil {
+		return mcp.NewToolResultError(err.Error()), nil
+	}
+	return mcp.NewToolResultText(fmt.Sprintf("attached %d file(s) to %s", len(a.Paths), a.Selector)), nil
+}
+
 // defaultMode es el modo de velocidad cuando ni la llamada ni la sesión fijan uno.
 // "auto" = rápido (movimiento de ratón mínimo, escritura en un solo evento, settle
 // breve) — el modo por defecto del MCP. "human" (Bézier + esperas aleatorias) se
@@ -16,6 +16,8 @@ import (

 	"github.com/mark3labs/mcp-go/mcp"
 	"github.com/mark3labs/mcp-go/server"
+
+	"fn-registry/functions/browser"
 )

 // registerLifecycleTools wires the per-profile Chromium lifecycle tools:
@@ -55,7 +57,10 @@ type chromiumMaster struct {
 	UserDataDir string `json:"user_data_dir"` // value of --user-data-dir
 	CDPPort     string `json:"cdp_port"`      // value of --remote-debugging-port ("" if none)
 	HasCDP      bool   `json:"has_cdp"`
-	Headless    bool   `json:"headless"` // true if launched with --headless / --headless=new / --headless=old
+	Headless    bool   `json:"headless"`               // true if launched with --headless / --headless=new / --headless=old
+	Pages       int    `json:"pages"`                  // count of "page" targets (best-effort via GET /json; 0 if no CDP or unreachable)
+	ActiveTitle string `json:"active_title,omitempty"` // title of the first "page" target
+	ActiveURL   string `json:"active_url,omitempty"`   // URL of the first "page" target
 }

 // parseCmdline turns the raw bytes of /proc/<pid>/cmdline into argv.
@@ -271,7 +276,7 @@ type browserListArgs struct{}

 func browserListTool() mcp.Tool {
 	return mcp.NewTool("browser_list",
-		mcp.WithDescription("List the running Chromium MASTER processes (one per user-data-dir master, NOT zygote/gpu/renderer children). For each: pid, profile (--profile-directory value), user_data_dir, cdp_port (--remote-debugging-port value, empty if none), has_cdp, headless (true if launched with --headless). Returns a JSON array. Read-only."),
+		mcp.WithDescription("List the running Chromium MASTER processes (one per user-data-dir master, NOT zygote/gpu/renderer children). For each: pid, profile (--profile-directory value), user_data_dir, cdp_port (--remote-debugging-port value, empty if none), has_cdp, headless (true if launched with --headless), pages (count of open page targets via GET /json, best-effort), active_title/active_url (first open page). Returns a JSON array. Read-only."),
 	)
 }

@@ -283,10 +288,43 @@ func (d *deps) handleBrowserList(_ context.Context, _ mcp.CallToolRequest, _ bro
 	if masters == nil {
 		masters = []chromiumMaster{}
 	}
+	// Enriquecer cada master con CDP con su nº de páginas y la primera página
+	// (título/URL) consultando GET /json. Best-effort: si el puerto no responde,
+	// se dejan los campos a cero — el listado de procesos nunca falla por esto.
+	for i := range masters {
+		enrichMasterTabs(&masters[i])
+	}
 	b, _ := json.MarshalIndent(masters, "", "  ")
 	return mcp.NewToolResultText(string(b)), nil
 }

+// enrichMasterTabs rellena Pages/ActiveTitle/ActiveURL de un master consultando
+// sus targets CDP por HTTP. No devuelve error: cualquier fallo (sin CDP, puerto
+// caído, timeout) deja los campos en su cero y el master se reporta igual.
+func enrichMasterTabs(m *chromiumMaster) {
+	if m.CDPPort == "" {
+		return
+	}
+	port, err := strconv.Atoi(m.CDPPort)
+	if err != nil {
+		return
+	}
+	tabs, err := browser.CdpListTabs("localhost", port)
+	if err != nil {
+		return
+	}
+	for _, t := range tabs {
+		if t.Type != "page" {
+			continue
+		}
+		m.Pages++
+		if m.ActiveURL == "" {
+			m.ActiveTitle = t.Title
+			m.ActiveURL = t.URL
+		}
+	}
+}
+
 // ---- browser_launch_profile (MUTA) ----

 type launchProfileArgs struct {
@@ -3,6 +3,8 @@ package main
 import (
 	"context"
 	"encoding/base64"
+	"encoding/json"
+	"fmt"
 	"os"

 	"github.com/mark3labs/mcp-go/mcp"
@@ -20,12 +22,91 @@ func registerReadTools(s *server.MCPServer, d *deps) {
 	s.AddTool(pageGetTextTool(), mcp.NewTypedToolHandler(d.handlePageGetText))
 	s.AddTool(pagePerceiveTool(), mcp.NewTypedToolHandler(d.handlePagePerceive))
 	s.AddTool(pageScreenshotTool(), mcp.NewTypedToolHandler(d.handlePageScreenshot))
+	s.AddTool(pageCollectConsoleTool(), mcp.NewTypedToolHandler(d.handlePageCollectConsole))
+	s.AddTool(pagePDFTool(), mcp.NewTypedToolHandler(d.handlePagePDF))

 	if !d.readOnly {
 		s.AddTool(pageEvalJSTool(), mcp.NewTypedToolHandler(d.handlePageEvalJS))
 	}
 }

+// ---- page_collect_console ----
+
+type pageCollectConsoleArgs struct {
+	Port       int `json:"port"`
+	DurationMs int `json:"duration_ms"`
+}
+
+func pageCollectConsoleTool() mcp.Tool {
+	return mcp.NewTool("page_collect_console",
+		mcp.WithDescription("Capture the page's console output (console.log/info/warn/error), uncaught JS exceptions and browser log entries during a time window, and return them as JSON. It is a SNAPSHOT: it records what happens during duration_ms AFTER the call starts, not past history — so trigger the action you want to observe (reload, click) right before or during the window. Use this to debug why a page misbehaves without flying blind."),
+		mcp.WithNumber("port", mcp.Description("CDP port. Default 9333 (Chrome isolated del MCP); usa 9222 explícito solo para adjuntarte al navegador diario.")),
+		mcp.WithNumber("duration_ms", mcp.Description("Capture window in milliseconds. Default 1500.")),
+	)
+}
+
+func (d *deps) handlePageCollectConsole(_ context.Context, _ mcp.CallToolRequest, a pageCollectConsoleArgs) (*mcp.CallToolResult, error) {
+	var entries []browser.ConsoleEntry
+	err := d.withConn(portOr(a.Port), func(c *browser.CDPConn) error {
+		var e error
+		entries, e = browser.CdpCollectConsole(c, a.DurationMs)
+		return e
+	})
+	if err != nil {
+		return mcp.NewToolResultError(err.Error()), nil
+	}
+	if entries == nil {
+		entries = []browser.ConsoleEntry{}
+	}
+	b, _ := json.MarshalIndent(entries, "", "  ")
+	return mcp.NewToolResultText(truncate(string(b), htmlMax)), nil
+}
+
+// ---- page_pdf ----
+
+type pagePDFArgs struct {
+	Port            int     `json:"port"`
+	Path            string  `json:"path"`
+	Landscape       bool    `json:"landscape"`
+	PrintBackground bool    `json:"print_background"`
+	Scale           float64 `json:"scale"`
+}
+
+func pagePDFTool() mcp.Tool {
+	return mcp.NewTool("page_pdf",
+		mcp.WithDescription("Render the current page to a PDF (Page.printToPDF) and write it to a local file path. Use for archiving an article/invoice/report exactly as laid out, when a screenshot is not enough (multi-page, selectable text)."),
+		mcp.WithNumber("port", mcp.Description("CDP port. Default 9333 (Chrome isolated del MCP); usa 9222 explícito solo para adjuntarte al navegador diario.")),
+		mcp.WithString("path", mcp.Required(), mcp.Description("Output .pdf file path.")),
+		mcp.WithBoolean("landscape", mcp.Description("Landscape orientation. Default false (portrait).")),
+		mcp.WithBoolean("print_background", mcp.Description("Include background graphics/colors. Default false.")),
+		mcp.WithNumber("scale", mcp.Description("Render scale. Default 1.0.")),
+	)
+}
+
+func (d *deps) handlePagePDF(_ context.Context, _ mcp.CallToolRequest, a pagePDFArgs) (*mcp.CallToolResult, error) {
+	if a.Path == "" {
+		return mcp.NewToolResultError("path is required"), nil
+	}
+	opts := browser.CdpPrintPDFOpts{
+		Landscape:       a.Landscape,
+		PrintBackground: a.PrintBackground,
+		Scale:           a.Scale,
+	}
+	var data []byte
+	err := d.withConn(portOr(a.Port), func(c *browser.CDPConn) error {
+		var e error
+		data, e = browser.CdpPrintPDF(c, opts)
+		return e
+	})
+	if err != nil {
+		return mcp.NewToolResultError(err.Error()), nil
+	}
+	if e := os.WriteFile(a.Path, data, 0o644); e != nil {
+		return mcp.NewToolResultError("saving pdf to " + a.Path + ": " + e.Error()), nil
+	}
+	return mcp.NewToolResultText(fmt.Sprintf("pdf saved to %s (%d bytes)", a.Path, len(data))), nil
+}
+
 // ---- page_get_text ----

 type pageGetTextArgs struct {