--- title: "Run operation" method: POST path: "/v1/operations" tags: ["datasets"] --- # Run operation `POST /v1/operations` Create an operation ## Request body - union - CreateYQLOperationRequest - `type` 'yql' - `params` YQLParams, required - `query` string, required — YQL query to execute. - `src` OperationSrcDataset[], required — List of source datasets for the operation. - `id` string, required — The object identifier, which can be referenced in the API endpoints. - `version` string, nullable, required — Version of the dataset. - `mapping` union — Mapping of dataset fields. - BatchInferenceMessagesMappingResponse - `type` string - `messages` union, required - MappingRecordText - … - MappingRecordColumn - … - `max_tokens` union - MappingRecordNumber - … - MappingRecordColumn - … - MappingRecordText - … - MappingRecordConstant - … - `custom_id` MappingRecordColumn - `type` 'column' - `name` string - BatchInferenceInputMapping - `type` string - `input` MappingRecordColumn, required - `type` 'column' - `name` string - `encoding_format` union - MappingRecordText - … - MappingRecordColumn - … - `dimensions` union - MappingRecordNumber - … - MappingRecordColumn - … - `custom_id` MappingRecordColumn - `type` 'column' - `name` string - `dst` OperationDstDataset[] — Optional destination datasets. If not provided, a temporary dataset will be created. - `id` string, required — The object identifier, which can be referenced in the API endpoints. - `version` string, nullable, required — Version of the dataset. - `ai_project_id` string, nullable — AI Studio project ID to associate with the dataset. - CreateMaterializeHistoryOperationRequest - `type` 'materialize_history' - `params` MaterializeHistoryParams, required - `start_datetime` string, date-time, required — Start datetime for history materialization (ISO 8601 format). - `end_datetime` string, date-time, required — End datetime for history materialization (ISO 8601 format). - `dataset_name` string, required — Name of the dataset to materialize history into. - `model_name` string, nullable — Optional model name to filter by model_flavor_id. - `ai_project_id` string, nullable — AI Studio project ID to associate with the dataset. - CreateBatchInferenceOperationRequest - `type` 'batch_inference' - `params` BatchInferenceParams, required - `model` string, required — ID of the model to use. - `completion_window` string, required - `src` CreateOperationSrcDataset[], required — List of source datasets for the operation. - `id` string, required — The object identifier, which can be referenced in the API endpoints. - `version` string, nullable, required — Version of the dataset. - `mapping` union — Mapping of dataset fields. - BatchInferenceMessagesMapping - `type` string - `messages` union, required - MappingRecordText - … - MappingRecordColumn - … - `max_tokens` union - MappingRecordNumber - … - MappingRecordColumn - … - MappingRecordText - … - `custom_id` MappingRecordColumn - `type` 'column' - `name` string - BatchInferenceInputMapping - `type` string - `input` MappingRecordColumn, required - `type` 'column' - `name` string - `encoding_format` union - MappingRecordText - … - MappingRecordColumn - … - `dimensions` union - MappingRecordNumber - … - MappingRecordColumn - … - `custom_id` MappingRecordColumn - `type` 'column' - `name` string - `dst` OperationDstDataset[] — Optional destination datasets. If not provided, a temporary dataset will be created. - `id` string, required — The object identifier, which can be referenced in the API endpoints. - `version` string, nullable, required — Version of the dataset. - `ai_project_id` string, nullable — AI Studio project ID to associate with the dataset. - CreateFineTuningOperationRequest - `type` 'finetuning' - `params` FineTuningParamsInput, required - `model_name` string, required — The name of the model to fine-tune - `hyperparameters` SupervisedHParameters - `batch_size` integer — Batch size for training - `learning_rate` number — Learning rate for training - `n_epochs` integer — Number of epochs for training - `warmup_ratio` number — Warmup ratio for learning rate - `weight_decay` number — Weight decay for optimizer - `packing` boolean — Whether to use packing for training - `max_grad_norm` number — Maximum gradient norm for optimizer - `context_length` integer — Context length for training - `lora` boolean — Whether to use LoRA (Low-Rank Adaptation) for training - `lora_r` integer — Rank for LoRA - `lora_alpha` integer — Alpha parameter for LoRA - `lora_dropout` number — Dropout rate for LoRA - `method` union — Fine-tuning method configuration - SupervisedMethodConfig - `type` 'supervised' - `supervised` SupervisedConfig, required - `hyperparameters` SupervisedHParameters, required - … - SpecDraftMethodConfig - `type` 'spec-draft' - `spec_draft` SpecDraftConfig, required - `hyperparameters` SpecDraftHParameters, required - … - `from_checkpoint` union — Checkpoint to continue training from - HFCheckpointParametersRequestInput - `type` union - string - 'hf' - `repo` string, required - `revision` string, nullable - `token` string, password, nullable - FTCheckpointParameters - `type` union - string - 'ft' - HFCheckpointParametersResponse - `type` union - string - 'hf' - `repo` string, required - `revision` string, nullable - `integrations` union[], nullable — Job integration parameters - union - WandbIntegrationRequest - `type` union - … - `wandb` WandbConfigRequest, required - … - MlflowIntegrationRequest - `type` 'mlflow' - `mlflow` MlflowConfigRequest, required - … - HfExportIntegrationRequest - `type` union - … - `hf` HfExportConfigRequest, required - … - WandbIntegrationResponse - `type` union - … - `wandb` WandbConfigResponse, required - … - MlflowIntegrationResponse - `type` 'mlflow' - `mlflow` MlflowConfigResponse, required - … - HfExportIntegrationResponse - `type` union - … - `hf` HfExportConfigResponse, required - … - `seed` integer, nullable — The seed controls the reproducibility of the job - `suffix` string, nullable — The suffix that will be used for output model name - `tags` string[], nullable — List of tags to associate with the job - `trained_steps` integer, nullable — The number of steps completed by the fine-tuning job. - `total_steps` integer, nullable — The total number of steps that the fine-tuning job is expected to complete. - `trained_tokens` integer, nullable — The total number of billable tokens processed by this fine-tuning job. The value will be null if the fine-tuning job is still running. - `estimated_finish` integer, nullable — The Unix timestamp (in seconds) for when the fine-tuning job is estimated to finish. The value will be null if the fine-tuning job is not running. - `src` union[], required — List of train and validation datasets for the finetuning operation. - union - OperationSrcTrainDataset - `id` string, required — The object identifier, which can be referenced in the API endpoints. - `version` string, nullable, required — Version of the dataset. - `mapping` union, required — Mapping for finetuning dataset, one of: text, messages, prompts, pretokenized. - FineTuningTextMapping - … - FineTuningPromptsMapping - … - FineTuningMessagesMapping - … - FineTuningPretokenizedMapping - … - `type` 'train' - OperationSrcValidationDataset - `id` string, required — The object identifier, which can be referenced in the API endpoints. - `version` string, nullable, required — Version of the dataset. - `mapping` union, required — Mapping for finetuning dataset, one of: text, messages, prompts, pretokenized. - FineTuningTextMapping - … - FineTuningPromptsMapping - … - FineTuningMessagesMapping - … - FineTuningPretokenizedMapping - … - `type` 'validation' - `ai_project_id` string, nullable — AI Studio project ID to associate with the dataset. - CreateFileToDatasetOperationRequest - `type` 'file_to_dataset' - `params` FileToDatasetParams, required - `file_id` string, required — ID of the source file to convert into a dataset. - `dataset_name` string, nullable — Name of the dataset. - `ai_project_id` string, nullable — AI Studio project ID to associate with the dataset. - CreateDatasetToFileOperationRequest - `type` 'dataset_to_file' - `params` DatasetToFileParams, required - `purpose` 'fine_tuning' | 'inference', required - `model` string, nullable — Model id for inference batch export. - `src` OperationSrcDatasetToFile[], required — List of source datasets for the dataset-to-file operation. - `dataset` OperationDatasetToFileDataset, required - `id` string, required — The object identifier, which can be referenced in the API endpoints. - `version` string, required — Version of the dataset. - `mapping` union, required — Mapping for dataset-to-file export. - FineTuningTextMapping - `type` 'text' — FineTuning mapping type for text format - `text` MappingRecordColumn, required - `type` 'column' - `name` string - FineTuningPromptsMapping - `type` 'prompts' — FineTuning mapping type for prompts format - `prompt` MappingRecordColumn, required - `type` 'column' - `name` string - `completion` MappingRecordColumn, required - `type` 'column' - `name` string - FineTuningMessagesMapping - `type` 'messages' — FineTuning mapping type for messages format - `messages` MappingRecordColumn, required - `type` 'column' - `name` string - DatasetToFileBatchInferenceMapping - `type` 'text_messages' - `messages` MappingRecordColumn, required - `type` 'column' - `name` string - `max_tokens` MappingRecordColumn - `type` 'column' - `name` string - `custom_id` MappingRecordColumn - `type` 'column' - `name` string - `dst` OperationDstDataset[] — Optional destination datasets. If not provided, a temporary dataset will be created. - `id` string, required — The object identifier, which can be referenced in the API endpoints. - `version` string, nullable, required — Version of the dataset. - `ai_project_id` string, nullable — AI Studio project ID to associate with the dataset. - CreateMergeOperationRequest - `type` 'merge' - `params` MergeParams - `dataset_name` string, nullable — Name of the merged dataset. A generated name is used when omitted. - `src` OperationSrcDataset[], required — List of source datasets for the operation. - `id` string, required — The object identifier, which can be referenced in the API endpoints. - `version` string, nullable, required — Version of the dataset. - `mapping` union — Mapping of dataset fields. - BatchInferenceMessagesMappingResponse - `type` string - `messages` union, required - MappingRecordText - … - MappingRecordColumn - … - `max_tokens` union - MappingRecordNumber - … - MappingRecordColumn - … - MappingRecordText - … - MappingRecordConstant - … - `custom_id` MappingRecordColumn - `type` 'column' - `name` string - BatchInferenceInputMapping - `type` string - `input` MappingRecordColumn, required - `type` 'column' - `name` string - `encoding_format` union - MappingRecordText - … - MappingRecordColumn - … - `dimensions` union - MappingRecordNumber - … - MappingRecordColumn - … - `custom_id` MappingRecordColumn - `type` 'column' - `name` string - `ai_project_id` string, nullable — AI Studio project ID to associate with the dataset. ## Response `200` Successful Response - CreateOperationResponse - `id` string, required — Operation id - `type` 'yql' | 'materialize_history' | 'batch_inference' | 'finetuning' | 'file_to_dataset' | 'dataset_to_file' | 'merge', required - `params` union, required — Parameters specific to the operation type. - YQLParams - `query` string, required — YQL query to execute. - MaterializeHistoryParams - `start_datetime` string, date-time, required — Start datetime for history materialization (ISO 8601 format). - `end_datetime` string, date-time, required — End datetime for history materialization (ISO 8601 format). - `dataset_name` string, required — Name of the dataset to materialize history into. - `model_name` string, nullable — Optional model name to filter by model_flavor_id. - BatchInferenceParams - `model` string, required — ID of the model to use. - `completion_window` string, required - FineTuningParamsOutput - `model_name` string, required — The name of the model to fine-tune - `hyperparameters` SupervisedHParameters - `batch_size` integer — Batch size for training - `learning_rate` number — Learning rate for training - `n_epochs` integer — Number of epochs for training - `warmup_ratio` number — Warmup ratio for learning rate - `weight_decay` number — Weight decay for optimizer - `packing` boolean — Whether to use packing for training - `max_grad_norm` number — Maximum gradient norm for optimizer - `context_length` integer — Context length for training - `lora` boolean — Whether to use LoRA (Low-Rank Adaptation) for training - `lora_r` integer — Rank for LoRA - `lora_alpha` integer — Alpha parameter for LoRA - `lora_dropout` number — Dropout rate for LoRA - `method` union — Fine-tuning method configuration - SupervisedMethodConfig - `type` 'supervised' - `supervised` SupervisedConfig, required - `hyperparameters` SupervisedHParameters, required - … - SpecDraftMethodConfig - `type` 'spec-draft' - `spec_draft` SpecDraftConfig, required - `hyperparameters` SpecDraftHParameters, required - … - `from_checkpoint` union — Checkpoint to continue training from - HFCheckpointParametersRequestOutput - `type` union - string - 'hf' - `repo` string, required - `revision` string, nullable - `token` string, nullable - FTCheckpointParameters - `type` union - string - 'ft' - HFCheckpointParametersResponse - `type` union - string - 'hf' - `repo` string, required - `revision` string, nullable - `integrations` union[], nullable — Job integration parameters - union - WandbIntegrationRequest - `type` union - … - `wandb` WandbConfigRequest, required - … - MlflowIntegrationRequest - `type` 'mlflow' - `mlflow` MlflowConfigRequest, required - … - HfExportIntegrationRequest - `type` union - … - `hf` HfExportConfigRequest, required - … - WandbIntegrationResponse - `type` union - … - `wandb` WandbConfigResponse, required - … - MlflowIntegrationResponse - `type` 'mlflow' - `mlflow` MlflowConfigResponse, required - … - HfExportIntegrationResponse - `type` union - … - `hf` HfExportConfigResponse, required - … - `seed` integer, nullable — The seed controls the reproducibility of the job - `suffix` string, nullable — The suffix that will be used for output model name - `tags` string[], nullable — List of tags to associate with the job - `trained_steps` integer, nullable — The number of steps completed by the fine-tuning job. - `total_steps` integer, nullable — The total number of steps that the fine-tuning job is expected to complete. - `trained_tokens` integer, nullable — The total number of billable tokens processed by this fine-tuning job. The value will be null if the fine-tuning job is still running. - `estimated_finish` integer, nullable — The Unix timestamp (in seconds) for when the fine-tuning job is estimated to finish. The value will be null if the fine-tuning job is not running. - FileToDatasetParams - `file_id` string, required — ID of the source file to convert into a dataset. - `dataset_name` string, nullable — Name of the dataset. - DatasetToFileParams - `purpose` 'fine_tuning' | 'inference', required - `model` string, nullable — Model id for inference batch export. - MergeParams - `dataset_name` string, nullable — Name of the merged dataset. A generated name is used when omitted. - `src` union[], required — List of source datasets for the operation. - union - OperationSrcDataset - `id` string, required — The object identifier, which can be referenced in the API endpoints. - `version` string, nullable, required — Version of the dataset. - `mapping` union — Mapping of dataset fields. - BatchInferenceMessagesMappingResponse - `type` string - `messages` union, required - … - `max_tokens` union - … - `custom_id` MappingRecordColumn - … - BatchInferenceInputMapping - `type` string - `input` MappingRecordColumn, required - … - `encoding_format` union - … - `dimensions` union - … - `custom_id` MappingRecordColumn - … - OperationSrcTrainDataset - `id` string, required — The object identifier, which can be referenced in the API endpoints. - `version` string, nullable, required — Version of the dataset. - `mapping` union, required — Mapping for finetuning dataset, one of: text, messages, prompts, pretokenized. - FineTuningTextMapping - `type` 'text' — FineTuning mapping type for text format - `text` MappingRecordColumn, required - … - FineTuningPromptsMapping - `type` 'prompts' — FineTuning mapping type for prompts format - `prompt` MappingRecordColumn, required - … - `completion` MappingRecordColumn, required - … - FineTuningMessagesMapping - `type` 'messages' — FineTuning mapping type for messages format - `messages` MappingRecordColumn, required - … - FineTuningPretokenizedMapping - `type` 'pretokenized' — FineTuning mapping type for pretokenized format - `input_ids` MappingRecordColumn, required - … - `labels` MappingRecordColumn - … - `attention_mask` MappingRecordColumn - … - `type` 'train' - OperationSrcValidationDataset - `id` string, required — The object identifier, which can be referenced in the API endpoints. - `version` string, nullable, required — Version of the dataset. - `mapping` union, required — Mapping for finetuning dataset, one of: text, messages, prompts, pretokenized. - FineTuningTextMapping - `type` 'text' — FineTuning mapping type for text format - `text` MappingRecordColumn, required - … - FineTuningPromptsMapping - `type` 'prompts' — FineTuning mapping type for prompts format - `prompt` MappingRecordColumn, required - … - `completion` MappingRecordColumn, required - … - FineTuningMessagesMapping - `type` 'messages' — FineTuning mapping type for messages format - `messages` MappingRecordColumn, required - … - FineTuningPretokenizedMapping - `type` 'pretokenized' — FineTuning mapping type for pretokenized format - `input_ids` MappingRecordColumn, required - … - `labels` MappingRecordColumn - … - `attention_mask` MappingRecordColumn - … - `type` 'validation' - OperationSrcDatasetToFile - `dataset` OperationDatasetToFileDataset, required - `id` string, required — The object identifier, which can be referenced in the API endpoints. - `version` string, required — Version of the dataset. - `mapping` union, required — Mapping for dataset-to-file export. - FineTuningTextMapping - `type` 'text' — FineTuning mapping type for text format - `text` MappingRecordColumn, required - … - FineTuningPromptsMapping - `type` 'prompts' — FineTuning mapping type for prompts format - `prompt` MappingRecordColumn, required - … - `completion` MappingRecordColumn, required - … - FineTuningMessagesMapping - `type` 'messages' — FineTuning mapping type for messages format - `messages` MappingRecordColumn, required - … - DatasetToFileBatchInferenceMapping - `type` 'text_messages' - `messages` MappingRecordColumn, required - … - `max_tokens` MappingRecordColumn - … - `custom_id` MappingRecordColumn - … - `dst` OperationDstDataset[], required — List of destination datasets for the operation. - `id` string, required — The object identifier, which can be referenced in the API endpoints. - `version` string, nullable, required — Version of the dataset. - `status` 'queued' | 'running' | 'succeeded' | 'failed' | 'cancelled' | 'unknown', required - `created_at` integer, required — The Unix timestamp (in seconds) for when the operation was created. - `in_progress_at` integer, nullable — The Unix timestamp (in seconds) for when the operation started processing. - `finished_at` integer, nullable — The Unix timestamp (in seconds) for when the operation was completed. - `ai_project_id` string, nullable — AI Studio project ID to associate with the dataset. ## Other responses - `422` — Validation Error --- [API](https://skmtc.net/nebius/apis/nebius-openai-compatible-inference-api.md) · [All operations](https://skmtc.net/nebius/apis/nebius-openai-compatible-inference-api/llms.txt) · [OpenAPI document](https://skmtc-service-staging.skmtc.workers.dev/v1/apis/nebius/nebius-openai-compatible-inference-api/versions/0fb323abba3c/schema)