diff --git a/streamingpro-mlsql/src/main/java/streaming/core/datasource/impl/MLSQLText.scala b/streamingpro-mlsql/src/main/java/streaming/core/datasource/impl/MLSQLText.scala
index 4ffb12729..bda79ebb8 100644
--- a/streamingpro-mlsql/src/main/java/streaming/core/datasource/impl/MLSQLText.scala
+++ b/streamingpro-mlsql/src/main/java/streaming/core/datasource/impl/MLSQLText.scala
@@ -1,10 +1,12 @@
 package streaming.core.datasource.impl
 
 import org.apache.spark.ml.param.Param
-import org.apache.spark.sql.{DataFrame, DataFrameReader, DataFrameWriter, Row, functions => F}
+import org.apache.spark.sql.mlsql.session.MLSQLException
+import org.apache.spark.sql.{DataFrame, DataFrameReader, DataFrameWriter, Row, SaveMode, functions => F}
 import streaming.core.datasource._
 import streaming.dsl.ScriptSQLExec
 import streaming.dsl.mmlib.algs.param.{BaseParams, WowParams}
+import tech.mlsql.tool.HDFSOperatorV2
 
 /**
  * 2019-03-20 WilliamZhu(allwefantasy@gmail.com)
@@ -26,7 +28,39 @@ class MLSQLText(override val uid: String) extends MLSQLBaseFileSource with WowPa
   }
 
   override def save(writer: DataFrameWriter[Row], config: DataSinkConfig): Unit = {
-    throw new RuntimeException("text not support save")
+    val context = ScriptSQLExec.contextGetOrForTest()
+    val baseDir = resourceRealPath(context.execListener, Option(context.owner), config.path)
+
+    if (HDFSOperatorV2.fileExists(baseDir)) {
+      if (config.mode == SaveMode.Overwrite) {
+        HDFSOperatorV2.deleteDir(baseDir)
+      }
+      if (config.mode == SaveMode.ErrorIfExists) {
+        throw new MLSQLException(s"${baseDir} is exists")
+      }
+    }
+
+    config.config.get(contentColumn.name).map { m =>
+      set(contentColumn, m)
+    }.getOrElse {
+      throw new MLSQLException(s"${contentColumn.name} is required")
+    }
+
+    config.config.get(fileName.name).map { m =>
+      set(fileName, m)
+    }.getOrElse {
+      throw new MLSQLException(s"${fileName.name} is required")
+    }
+
+    val _fileName = $(fileName)
+    val _contentColumn = $(contentColumn)
+
+    val saveContent = (fileName: String, buffer: String) => {
+      HDFSOperatorV2.saveFile(baseDir, fileName, List(("",buffer)).toIterator)
+      baseDir + "/" + fileName
+    }
+
+    config.df.get.rdd.map(r => saveContent(r.getAs[String](_fileName), r.getAs[String](_contentColumn))).count()
   }
 
   override def sourceInfo(config: DataAuthConfig): SourceInfo = {
@@ -50,4 +84,7 @@ class MLSQLText(override val uid: String) extends MLSQLBaseFileSource with WowPa
 
   final val wholetext: Param[Boolean] = new Param[Boolean](this, "wholetext", "`wholetext` (default `false`): If true, read a file as a single row and not split by \"\\n\".")
   final val lineSep: Param[String] = new Param[String](this, "lineSep", "`(default covers all `\\r`, `\\r\\n` and `\\n`): defines the line separator\n   * that should be used for parsing.")
+
+  final val contentColumn: Param[String] = new Param[String](this, "imageColumn", "for save mode")
+  final val fileName: Param[String] = new Param[String](this, "fileName", "for save mode")
 }