chore(model): revert task input/output refactor

artifact/artifact/v1alpha/file_catalog.proto

@@ -20,77 +20,77 @@ message GetFileCatalogRequest {
 }
 
 // GetFileCatalogResponse
-message GetFileCatalogResponse{
-    // metadata
-    message Metadata {
-        // file uid
-        string file_uid = 1;
-        // file id
-        string file_id = 2;
-        // file type
-        FileType file_type = 3;
-        // file size in bytes
-        int64 file_size = 4;
-        // upload time
-        google.protobuf.Timestamp file_upload_time = 5;
-        // file process status
-        FileProcessStatus file_process_status = 6;
-    }
-    // text message
-    message Text{
-        // pipelines
-        repeated string pipeline_ids = 1;
-        // transformed content
-        string transformed_content = 2;
-        // transformed content uid
-        string transformed_content_uid = 3;
-        // transformed content chunk number
-        int32 transformed_content_chunk_num = 4;
-        // transformed content token number
-        int32 transformed_content_token_num = 5;
-        // transformed content update time 
-        google.protobuf.Timestamp transformed_content_update_time = 6;
-    }
-    // chunk type
-    enum ChunkType {
-        // unspecified
-        CHUNK_TYPE_UNSPECIFIED = 0;
-        // text
-        CHUNK_TYPE_TEXT = 1;
-        // image
-        CHUNK_TYPE_IMAGE = 2;
-        // audio
-        CHUNK_TYPE_AUDIO = 3;
-        // video
-        CHUNK_TYPE_VIDEO = 4;
-    }
-    // chunk message
-    message Chunk {
-        // chunk uid
-        string uid = 1;
-        // chunk type. i.e. text, image, audio, and video
-        ChunkType type = 2;
-        // chunk start position
-        int32 start_pos = 3;
-        // chunk end position
-        int32 end_pos =4;
-        // chunk content
-        string content = 5;
-        // chunk tokens num
-        int32 tokens_num = 6;
-        // embedding. float32 array
-        repeated float embedding = 7;
-        // chunk create time
-        google.protobuf.Timestamp create_time = 8;
-        // chunk retrievable
-        bool retrievable = 9;
-    }
-    // original data is encoded in base64
-    string original_data = 1;
-    // file catalog
-    Metadata metadata = 2;
-    // text 
-    Text text = 3;
-    // chunks
-    repeated Chunk chunks = 4;
+message GetFileCatalogResponse {
+  // metadata
+  message Metadata {
+    // file uid
+    string file_uid = 1;
+    // file id
+    string file_id = 2;
+    // file type
+    FileType file_type = 3;
+    // file size in bytes
+    int64 file_size = 4;
+    // upload time
+    google.protobuf.Timestamp file_upload_time = 5;
+    // file process status
+    FileProcessStatus file_process_status = 6;
+  }
+  // text message
+  message Text {
+    // pipelines
+    repeated string pipeline_ids = 1;
+    // transformed content
+    string transformed_content = 2;
+    // transformed content uid
+    string transformed_content_uid = 3;
+    // transformed content chunk number
+    int32 transformed_content_chunk_num = 4;
+    // transformed content token number
+    int32 transformed_content_token_num = 5;
+    // transformed content update time
+    google.protobuf.Timestamp transformed_content_update_time = 6;
+  }
+  // chunk type
+  enum ChunkType {
+    // unspecified
+    CHUNK_TYPE_UNSPECIFIED = 0;
+    // text
+    CHUNK_TYPE_TEXT = 1;
+    // image
+    CHUNK_TYPE_IMAGE = 2;
+    // audio
+    CHUNK_TYPE_AUDIO = 3;
+    // video
+    CHUNK_TYPE_VIDEO = 4;
+  }
+  // chunk message
+  message Chunk {
+    // chunk uid
+    string uid = 1;
+    // chunk type. i.e. text, image, audio, and video
+    ChunkType type = 2;
+    // chunk start position
+    int32 start_pos = 3;
+    // chunk end position
+    int32 end_pos = 4;
+    // chunk content
+    string content = 5;
+    // chunk tokens num
+    int32 tokens_num = 6;
+    // embedding. float32 array
+    repeated float embedding = 7;
+    // chunk create time
+    google.protobuf.Timestamp create_time = 8;
+    // chunk retrievable
+    bool retrievable = 9;
+  }
+  // original data is encoded in base64
+  string original_data = 1;
+  // file catalog
+  Metadata metadata = 2;
+  // text
+  Text text = 3;
+  // chunks
+  repeated Chunk chunks = 4;
 }

common/task/v1alpha/task.proto

@@ -21,19 +21,15 @@ enum Task {
   // Text to Image - generate images from input text prompts.
   TASK_TEXT_TO_IMAGE = 7;
   // Text Generation - generate texts from input text prompts.
-  reserved 8;
+  TASK_TEXT_GENERATION = 8;
   // Conversational Text Generation - generate text as responses to a dialog input.
-  reserved 9;
+  TASK_TEXT_GENERATION_CHAT = 9;
   // Visual Question Answering - generate text as a response to a visual prompt.
-  reserved 10;
+  TASK_VISUAL_QUESTION_ANSWERING = 10;
   // Image to Image - generate an image from another image.
   TASK_IMAGE_TO_IMAGE = 11;
   // Text Embeddings - generate an embedding (a representation as coordinates) from a text input.
   TASK_TEXT_EMBEDDINGS = 12;
   // Speech Recognition - transcribe the words in an audio input.
   TASK_SPEECH_RECOGNITION = 13;
-  // Conversational Text Generation - generate text as responses to a dialog input.
-  TASK_CHAT = 14;
-  // Completion Text Generation - generate text following the input prompt.
-  TASK_COMPLETION = 15;
 }