mlflow--mlflow/mlflow/protos/databricks_filesystem_service.proto

// Check Databricks Protobuf Style Guide (go/protostyleguide) about recommended Protobuf practices.
// See go/protostyleguide/syntax.
syntax = "proto2";

// See go/protostyleguide/packages.
package mlflow.filesystem;

import "databricks.proto";
import "scalapb/scalapb.proto";

option java_generate_equals_and_hash = true;
option java_package = "com.databricks.api.proto.filesystem";
option py_generic_services = true;
option (scalapb.options) = {flat_package: true};

service FilesystemService {
  // Only for internal usage for now. Endpoint and request format are to be finalized.
  rpc CreateDownloadUrl(CreateDownloadUrlRequest) returns (CreateDownloadUrlResponse) {
    option (rpc) = {
      endpoints: {
        method: "POST"
        path: "/fs/create-download-url"
        since: {
          major: 2
          minor: 0
        }
      }
      visibility: PUBLIC_UNDOCUMENTED
    };
  }

  rpc CreateUploadUrl(CreateUploadUrlRequest) returns (CreateUploadUrlResponse) {
    option (rpc) = {
      endpoints: {
        method: "POST"
        path: "/fs/create-upload-url"
        since: {
          major: 2
          minor: 0
        }
      }
      visibility: PUBLIC_UNDOCUMENTED
    };
  }
}

message HttpHeader {
  optional string name = 1;
  optional string value = 2;
}

message CreateDownloadUrlRequest {
  option (scalapb.message).extends = "com.databricks.rpc.RPC[CreateDownloadUrlResponse]";

  // The path to the file for which we would like a pre-signed download URL from which you may GET
  // the file's contents. This path can be an HDFS-style path starting with a scheme, or a
  // Unix-style absolute path. For example:
  // "dbfs:/Volumes/<catalog>/<schema>/<volume>/<path_to_file>" or
  // "/Volumes/<catalog>/<schema>/<volume>/<path_to_file>".
  optional string path = 1;
}

message CreateDownloadUrlResponse {
  option (scalapb.message).extends = "com.databricks.rpc.DoNotLogContents";
  // The pre-signed download URL which you may use to GET the specified file.
  //
  // To download the file, issue an HTTP GET request with the headers specified below
  // (see 'headers') and the body of the response should be the file's contents. If the file
  // does not exist or can't be accessed, you'll receive a cloud-service-provider-specific
  // error response.
  optional string url = 1;

  // These headers must be included in your HTTP request to the given 'url'. If all of the headers
  // are not included exactly, the behavior of the request is unspecified (most likely you will get
  // an error response).
  repeated HttpHeader headers = 2;
}

message CreateUploadUrlRequest {
  option (scalapb.message).extends = "com.databricks.rpc.RPC[CreateUploadUrlResponse]";

  // The path to the file for which we would like a pre-signed upload URL to which you may PUT
  // the file's contents. This path can be an HDFS-style path starting with a scheme, or a
  // Unix-style absolute path. For example:
  // "dbfs:/Volumes/<catalog>/<schema>/<volume>/<path_to_file>" or
  // "/Volumes/<catalog>/<schema>/<volume>/<path_to_file>".
  optional string path = 1;
}

message CreateUploadUrlResponse {
  option (scalapb.message).extends = "com.databricks.rpc.DoNotLogContents";

  // The pre-signed download URL which you may use to PUT the specified file's contents.
  //
  // To upload the file, issue an HTTP PUT request with the headers specified below, and with the
  // body of the request being the file contents you wish to upload.
  //
  // You must add the "Content-Length" header to your request with the value being the number of
  // bytes you are uploading. The maximum possible Content-Length depends on the cloud service
  // provider, but is at least 5GB.
  //
  // The headers include a "Content-Type" of "application/octet-stream". You must encode your file
  // contents in the request body as an octet-stream.
  //
  // You'll receive a cloud-service-provider-specific response.
  optional string url = 1;

  // These headers must be included in your HTTP request to the given 'url'. If all of the headers
  // are not included exactly, the behavior of the request is unspecified (most likely you will get
  // an error response).
  repeated HttpHeader headers = 2;
}

message DirectoryEntry {
  option (scalapb.message).extends = "com.databricks.rpc.DoNotLogContents";

  // The path of the file or directory.
  // Directories have a trailing slash.
  // Example 1: "/Volumes/catalog/schema/volume/directory/file.txt"
  // Example 2: "/Volumes/catalog/schema/volume/directory/"
  optional string path = 1;

  // True if the path is a directory.
  optional bool is_directory = 2;

  // The length of the file in bytes or null if the path is a directory.
  optional int64 file_size = 3;

  // Last modification time of given file/directory in milliseconds since the Unix Epoch.
  // NOTE: This field is not set for directories.
  optional int64 last_modified = 4;

  // Name of the file or directory.
  // There is no slash at the end of the directory names.
  // Example: "file.txt" or "mydirectory"
  optional string name = 5;
}

// The response structure for listing files with the /fs/directories endpoint.
message ListDirectoryResponse {
  option (scalapb.message).extends = "com.databricks.rpc.DoNotLogContents";

  // The files and directories in the specified path.
  // Note that a file and a directory can share the same path.
  repeated DirectoryEntry contents = 1;

  // A token that can be sent as `page_token` to retrieve the next page.
  // If this field is omitted, there are no subsequent pages.
  optional string next_page_token = 2;
}