fix(core): auto-filter tool arguments based on JSON Schema

- Updated zenml_wrapper.py to fetch tool schema and filter tool_args before execution.
- This prevents 'Unexpected keyword argument' errors from Zem's auto-chaining.
- Reverted **kwargs in argilla/server.py as it is no longer needed and not supported by FastMCP.

Author: OdinHoang03

pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "xfmr-zem"
-version = "0.3.8"
+version = "0.3.9"
 description = "Zem: Unified Data Pipeline Framework (ZenML + NeMo Curator + DataJuicer) for multi-domain processing"
 readme = "README.md"
 requires-python = ">=3.10,<3.13"

src/xfmr_zem/servers/argilla/server.py

@@ -70,7 +70,6 @@ def create_dataset(
     guidelines: str = "",
     api_url: Optional[str] = None,
     api_key: Optional[str] = None,
-    **kwargs: Any,
 ) -> Dict[str, Any]:
     """
     Tạo hoặc lấy dataset trên Argilla server.
@@ -136,7 +135,6 @@ def push_records(
     batch_size: int = 200,
     api_url: Optional[str] = None,
     api_key: Optional[str] = None,
-    **kwargs: Any,
 ) -> Dict[str, Any]:
     """
     Đẩy records từ pipeline vào Argilla dataset.
@@ -456,7 +454,6 @@ def annotation_progress(
     workspace: str = "admin",
     api_url: Optional[str] = None,
     api_key: Optional[str] = None,
-    **kwargs: Any,
 ) -> Dict[str, Any]:
     """
     Thống kê tiến độ annotation của dataset.
@@ -622,7 +619,6 @@ def create_user(
     workspace: Optional[str] = None,
     api_url: Optional[str] = None,
     api_key: Optional[str] = None,
-    **kwargs: Any,
 ) -> Dict[str, Any]:
     """
     Tạo user mới trên Argilla server.

src/xfmr_zem/zenml_wrapper.py

@@ -195,6 +195,25 @@ def mcp_generic_step(
     command = server_config.get("command", "python")
     args = server_config.get("args", [])
     env = server_config.get("env", os.environ.copy())
+
+    # Filter tool_args based on tool's inputSchema
+    try:
+        tools = list_mcp_tools(command, args, env)
+        target_tool = next((t for t in tools if t["name"] == tool_name), None)
+        if target_tool and "inputSchema" in target_tool:
+            schema = target_tool["inputSchema"]
+            properties = schema.get("properties", {})
+            valid_keys = set(properties.keys())
+            
+            # Filter arguments
+            original_args = tool_args.copy()
+            tool_args = {k: v for k, v in tool_args.items() if k in valid_keys}
+            
+            removed_keys = set(original_args.keys()) - valid_keys
+            if removed_keys:
+                logger.info(f"[{server_name}] Lọc bỏ tham số dư thừa cho '{tool_name}': {removed_keys}")
+    except Exception as e:
+        logger.warning(f"[{server_name}] Không thể lấy schema để lọc tham số cho '{tool_name}': {e}")
 
     logger.info(f"[{server_name}] Executing tool '{tool_name}'")
     start_time = time.time()