[Feature][Transforms-V2] LLM transforms Support custom field name

apache · Sep 13, 2024 · 5793222 · 5793222
1 parent 4dae15c
commit 5793222
Show file tree

Hide file tree

Showing 4 changed files with 18 additions and 23 deletions.
diff --git a/docs/en/transform-v2/llm.md b/docs/en/transform-v2/llm.md
@@ -11,10 +11,10 @@ more.
 ## Options
 
 | name                   | type   | required | default value |
-|------------------------| ------ | -------- | ------------- |
+|------------------------| ------ | -------- |---------------|
 | model_provider         | enum   | yes      |               |
 | output_data_type       | enum   | no       | String        |
-| output_column_name     | string | no       | String        |
+| output_column_name     | string | no       | llm_output    |
 | prompt                 | string | yes      |               |
 | inference_columns      | list   | no       |               |
 | model                  | string | yes      |               |

diff --git a/docs/zh/transform-v2/llm.md b/docs/zh/transform-v2/llm.md
@@ -8,20 +8,20 @@
 
 ## 属性
 
-| 名称                     | 类型   | 是否必须 | 默认值    |
-|------------------------| ------ | -------- |--------|
-| model_provider         | enum   | yes      |        |
-| output_data_type       | enum   | no       | String |
-| output_column_name     | string | no       |        |
-| prompt                 | string | yes      |        |
-| inference_columns      | list   | no       |        |
-| model                  | string | yes      |        |
-| api_key                | string | yes      |        |
-| api_path               | string | no       |        |
-| custom_config          | map    | no       |        |
-| custom_response_parse  | string | no       |        |
-| custom_request_headers | map    | no       |        |
-| custom_request_body    | map    | no       |        |
+| 名称                     | 类型   | 是否必须 | 默认值         |
+|------------------------| ------ | -------- |-------------|
+| model_provider         | enum   | yes      |             |
+| output_data_type       | enum   | no       | String      |
+| output_column_name     | string | no       | llm_output   |
+| prompt                 | string | yes      |             |
+| inference_columns      | list   | no       |             |
+| model                  | string | yes      |             |
+| api_key                | string | yes      |             |
+| api_path               | string | no       |             |
+| custom_config          | map    | no       |             |
+| custom_response_parse  | string | no       |             |
+| custom_request_headers | map    | no       |             |
+| custom_request_body    | map    | no       |             |
 
 ### model_provider
 

diff --git a/...transforms-v2/src/main/java/org/apache/seatunnel/transform/nlpmodel/llm/LLMTransform.java b/...transforms-v2/src/main/java/org/apache/seatunnel/transform/nlpmodel/llm/LLMTransform.java
@@ -32,8 +32,6 @@
 import org.apache.seatunnel.transform.nlpmodel.llm.remote.custom.CustomModel;
 import org.apache.seatunnel.transform.nlpmodel.llm.remote.openai.OpenAIModel;
 
-import org.apache.commons.lang3.StringUtils;
-
 import lombok.NonNull;
 import lombok.SneakyThrows;
 
@@ -147,10 +145,7 @@ protected Column getOutputColumn() {
         boolean isExist = Arrays.asList(fieldNames).contains(customFieldName);
         if (isExist) {
             throw new IllegalArgumentException(
-                    String.format("Field name %s already exists", customFieldName));
-        }
-        if (StringUtils.isEmpty(customFieldName)) {
-            customFieldName = "llm_output";
+                    String.format("llm inference field name %s already exists", customFieldName));
         }
         return PhysicalColumn.of(
                 customFieldName, outputDataType, (Long) null, true, null, "Output column of LLM");

diff --git a/...orms-v2/src/main/java/org/apache/seatunnel/transform/nlpmodel/llm/LLMTransformConfig.java b/...orms-v2/src/main/java/org/apache/seatunnel/transform/nlpmodel/llm/LLMTransformConfig.java
@@ -40,7 +40,7 @@ public class LLMTransformConfig extends ModelTransformConfig {
     public static final Option<String> OUTPUT_COLUMN_NAME =
             Options.key("output_column_name")
                     .stringType()
-                    .noDefaultValue()
+                    .defaultValue("llm_output")
                     .withDescription("custom field name for the llm output data");
 
     public static final Option<Integer> INFERENCE_BATCH_SIZE =