kl-mcp-client 2.0.4__tar.gz → 2.1.0__tar.gz

{kl_mcp_client-2.0.4 → kl_mcp_client-2.1.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: kl-mcp-client
-Version: 2.0.4
+Version: 2.1.0
 Summary: MCP Client for Python
 Author-email: Kyle <hngan.it@gmail.com>
 License: MIT

{kl_mcp_client-2.0.4 → kl_mcp_client-2.1.0}/kl_mcp_client/asyncio/tools.py

@@ -46,7 +46,7 @@ class MCPTools:
 
     async def screenshot(self, sessionId: str) -> Dict[str, Any]:
         """
-        ADK Web expects this structure:
+        ADK Web expects:
         {
             "type": "image",
             "mimeType": "image/png",
@@ -121,7 +121,7 @@ class MCPTools:
     ) -> Dict[str, Any]:
         args = {"sessionId": sessionId, "selector": selector}
         if timeoutMs is not None:
-            args["timeoutMs"] = timeoutMs
+            args["timeoutMs"] = int(timeoutMs)
 
         res = await self.client.call_tool("waitForSelector", args)
         return res.get("structuredContent", {})
@@ -145,10 +145,10 @@ class MCPTools:
         return res.get("structuredContent", {})
 
     # ======================================================
-    # ADVANCED
+    # ADVANCED ACTIONS
     # ======================================================
 
-    async def click_to_text(self, sessionId: str, text: str) -> dict:
+    async def click_to_text(self, sessionId: str, text: str) -> Dict[str, Any]:
         res = await self.client.call_tool(
             "clickToText", {"sessionId": sessionId, "text": text}
         )
@@ -178,25 +178,97 @@ class MCPTools:
         )
         return res.get("structuredContent", {})
 
+    async def get_dom_tree(self, sessionId: str, args: Optional[dict] = None):
+        return await self.client.call_tool(
+            "getDomTree", {"sessionId": sessionId, "args": args or {}}
+        )
+
+    async def get_clickable(self, sessionId: str, args: Optional[dict] = None):
+        return await self.client.call_tool(
+            "getClickable", {"sessionId": sessionId, "args": args or {}}
+        )
+
+    async def selector_map(
+        self, sessionId: str, selector: str, args: Optional[dict] = None
+    ):
+        return await self.client.call_tool(
+            "selectorMap",
+            {"sessionId": sessionId, "selector": selector, "args": args or {}},
+        )
+
+    # ======================================================
+    # AI / CONTENT PARSING
+    # ======================================================
+
     async def parse_html_by_prompt(self, html: str, prompt: str) -> Dict[str, Any]:
+        res = await self.client.call_tool(
+            "parseHTMLByPrompt",
+            {"html": html, "prompt": prompt},
+        )
+        return res.get("structuredContent", {})
+
+    # ======================================================
+    # MOUSE / PERFORM ACTIONS
+    # ======================================================
+
+    async def perform_click_xy(
+        self,
+        sessionId: str,
+        x: float,
+        y: float,
+    ) -> Dict[str, Any]:
+        """
+        Move mouse smoothly to (x, y) and left click.
         """
-        Parse HTML content using AI with dynamic prompt-defined structure.
+        res = await self.client.call_tool(
+            "perform",
+            {
+                "sessionId": sessionId,
+                "action": "click",
+                "x": float(x),
+                "y": float(y),
+            },
+        )
+        return res.get("structuredContent", {})
 
-        Args:
-            html: Raw HTML string (client-provided)
-            prompt: Instruction that defines what to extract and output structure
-                    Example:
-                      - "Hãy lấy nội dung bài viết, struct trả về { content }"
-                      - "Hãy lấy số lượng like, share, comment, trả JSON { like, share, comment }"
+    async def perform_drag(
+        self,
+        sessionId: str,
+        from_x: float,
+        from_y: float,
+        to_x: float,
+        to_y: float,
+    ) -> Dict[str, Any]:
+        """
+        Drag mouse from (from_x, from_y) to (to_x, to_y).
+        """
+        res = await self.client.call_tool(
+            "perform",
+            {
+                "sessionId": sessionId,
+                "action": "drag",
+                "from": {"x": float(from_x), "y": float(from_y)},
+                "to": {"x": float(to_x), "y": float(to_y)},
+            },
+        )
+        return res.get("structuredContent", {})
 
-        Returns:
-            structuredContent (dynamic JSON defined by prompt)
+    async def perform_hover(
+        self,
+        sessionId: str,
+        x: float,
+        y: float,
+    ) -> Dict[str, Any]:
+        """
+        Move mouse smoothly to (x, y) without clicking.
         """
         res = await self.client.call_tool(
-            "parseHTMLByPrompt",
+            "perform",
             {
-                "html": html,
-                "prompt": prompt,
+                "sessionId": sessionId,
+                "action": "hover",
+                "x": float(x),
+                "y": float(y),
             },
         )
         return res.get("structuredContent", {})

{kl_mcp_client-2.0.4 → kl_mcp_client-2.1.0}/kl_mcp_client/tools.py

@@ -126,63 +126,27 @@ class MCPTools:
 
     @_ensure_client
     def upload_file(
-        self,
-        sessionId: str,
-        selector: str,
-        file_path: str,
+        self, sessionId: str, selector: str, filename: str, base64data: str
     ) -> Dict[str, Any]:
         """
-        Upload file (kể cả video lớn) vào input[type=file] theo luồng mới:
-        1. Multipart upload file lên MCP server
-        2. Nhận uploadId
-        3. Gọi MCP tool uploadFile với uploadId
-
+        Upload file vào input[type=file]
         Args:
             sessionId: MCP browser session
             selector: CSS selector, ví dụ 'input[type=file]'
-            file_path: đường dẫn file local (video, pdf, doc, ...)
+            filename: tên file trên browser side
+            base64data: dữ liệu base64 (không kèm header)
+        Returns:
+            structured result từ server
         """
-
-        if not file_path:
-            return {"ok": False, "error": "file_path is required"}
-
-        # --------------------------------------------------
-        # 1️⃣ Multipart upload file lên MCP server
-        # --------------------------------------------------
-        try:
-            with open(file_path, "rb") as f:
-                resp = self.client.http.post(
-                    "/upload",
-                    files={"file": f},
-                    timeout=300,  # upload file lớn
-                )
-        except Exception as e:
-            return {"ok": False, "error": f"upload http failed: {e}"}
-
-        if resp.status_code != 200:
-            return {
-                "ok": False,
-                "error": f"upload http error {resp.status_code}: {resp.text}",
-            }
-
-        data = resp.json()
-        upload_id = data.get("uploadId")
-        if not upload_id:
-            return {"ok": False, "error": "uploadId not returned from server"}
-
-        # --------------------------------------------------
-        # 2️⃣ Gọi MCP tool uploadFile (PATH MODE)
-        # --------------------------------------------------
-        result = self.client.call_tool(
+        return self.client.call_tool(
             "uploadFile",
             {
                 "sessionId": sessionId,
                 "selector": selector,
-                "uploadId": upload_id,
+                "filename": filename,
+                "data": base64data,
             },
-        )
-
-        return result.get("structuredContent", {})
+        ).get("structuredContent", {})
 
     @_ensure_client
     def wait_for_selector(
@@ -265,16 +229,6 @@ class MCPTools:
             {"sessionId": sessionId, "selector": selector, "args": args or {}},
         )
 
-    @_ensure_client
-    def find_element_by_prompt(self, sessionId: str, prompt: str) -> Dict[str, Any]:
-        """
-        Gọi tool findElementByPrompt trên MCP server.
-        Trả về structuredContent gồm: html, nodeId.
-        """
-        return self.client.call_tool(
-            "findElementByPrompt", {"sessionId": sessionId, "prompt": prompt}
-        ).get("structuredContent", {})
-
     # ======================================================
     # AI / CONTENT PARSING
     # ======================================================
@@ -300,3 +254,65 @@ class MCPTools:
                 "prompt": prompt,
             },
         ).get("structuredContent", {})
+    # ======================================================
+    # MOUSE / PERFORM ACTIONS
+    # ======================================================
+    @_ensure_client
+    def perform_click_xy(
+        self,
+        sessionId: str,
+        x: float,
+        y: float,
+    ) -> Dict[str, Any]:
+        """
+        Move mouse smoothly to (x, y) and left click.
+        """
+        return self.client.call_tool(
+            "perform",
+            {
+                "sessionId": sessionId,
+                "action": "click",
+                "x": float(x),
+                "y": float(y),
+            },
+        ).get("structuredContent", {})
+    @_ensure_client
+    def perform_drag(
+        self,
+        sessionId: str,
+        from_x: float,
+        from_y: float,
+        to_x: float,
+        to_y: float,
+    ) -> Dict[str, Any]:
+        """
+        Drag mouse from (from_x, from_y) to (to_x, to_y).
+        """
+        return self.client.call_tool(
+            "perform",
+            {
+                "sessionId": sessionId,
+                "action": "drag",
+                "from": {"x": float(from_x), "y": float(from_y)},
+                "to": {"x": float(to_x), "y": float(to_y)},
+            },
+        ).get("structuredContent", {})
+    @_ensure_client
+    def perform_hover(
+        self,
+        sessionId: str,
+        x: float,
+        y: float,
+    ) -> Dict[str, Any]:
+        """
+        Move mouse smoothly to (x, y) without clicking.
+        """
+        return self.client.call_tool(
+            "perform",
+            {
+                "sessionId": sessionId,
+                "action": "hover",
+                "x": float(x),
+                "y": float(y),
+            },
+        ).get("structuredContent", {})

{kl_mcp_client-2.0.4 → kl_mcp_client-2.1.0}/kl_mcp_client.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: kl-mcp-client
-Version: 2.0.4
+Version: 2.1.0
 Summary: MCP Client for Python
 Author-email: Kyle <hngan.it@gmail.com>
 License: MIT

{kl_mcp_client-2.0.4 → kl_mcp_client-2.1.0}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "kl-mcp-client"
-version = "2.0.4"
+version = "2.1.0"
 description = "MCP Client for Python"
 dependencies = [ "requests>=2.32.5", "httpx>=0.28.1",]
 readme = "README.md"