yutu/cmd/caption/insert.go

// Copyright 2025 eat-pray-ai & OpenWaygate
// SPDX-License-Identifier: Apache-2.0

package caption

import (
	"encoding/json/jsontext"
	"fmt"
	"io"

	cobramcp "github.com/eat-pray-ai/cobra-mcp"
	"github.com/eat-pray-ai/yutu/cmd"
	"github.com/eat-pray-ai/yutu/pkg"
	"github.com/eat-pray-ai/yutu/pkg/caption"
	"github.com/eat-pray-ai/yutu/pkg/utils"
	"github.com/google/jsonschema-go/jsonschema"
	"github.com/modelcontextprotocol/go-sdk/mcp"
	"github.com/spf13/cobra"
)

const (
	insertTool    = "caption-insert"
	insertConfirm = "Upload caption: %s"
	insertShort   = "Upload a caption"
	insertLong    = "Upload a caption. Use this tool to upload a caption to a video."
	insertExample = `# Insert a caption to a video
yutu caption insert --file subtitle.srt --videoId dQw4w9WgXcQ
# Insert a caption with language and name
yutu caption insert --file subtitle.srt --videoId dQw4w9WgXcQ --language en --name English
# Insert a standard caption and publish immediately
yutu caption insert --file subtitle.srt --videoId dQw4w9WgXcQ --trackKind standard --isDraft=false`
)

var insertInSchema = &jsonschema.Schema{
	Type:     "object",
	Required: []string{"file", "video_id"},
	Properties: map[string]*jsonschema.Schema{
		"file": {Type: "string", Description: fileUsage},
		"audio_track_type": {
			Type: "string", Description: attUsage,
			Enum:    []any{"unknown", "primary", "commentary", "descriptive"},
			Default: jsontext.Value(`"unknown"`),
		},
		"is_auto_synced": {Type: "boolean", Description: iasUsage},
		"is_cc":          {Type: "boolean", Description: iscUsage},
		"is_draft":       {Type: "boolean", Description: isdUsage},
		"is_easy_reader": {Type: "boolean", Description: iserUsage},
		"is_large":       {Type: "boolean", Description: islUsage},
		"language":       {Type: "string", Description: languageUsage},
		"name":           {Type: "string", Description: nameUsage},
		"track_kind": {
			Type: "string", Description: tkUsage,
			Enum:    []any{"standard", "ASR", "forced"},
			Default: jsontext.Value(`"standard"`),
		},
		"video_id":                   {Type: "string", Description: vidUsage},
		"on_behalf_of":               {Type: "string", Description: pkg.OBOUsage},
		"on_behalf_of_content_owner": {Type: "string", Description: pkg.OBOCOUsage},
		"output": {
			Type: "string", Description: pkg.SilentUsage,
			Enum:    []any{"json", "yaml", "silent"},
			Default: jsontext.Value(`"yaml"`),
		},
	},
}

func init() {
	mcp.AddTool(
		cmd.Server, &mcp.Tool{
			Name: insertTool, Title: insertShort, Description: insertLong,
			InputSchema: insertInSchema, Annotations: &mcp.ToolAnnotations{
				DestructiveHint: new(false),
				IdempotentHint:  false,
				OpenWorldHint:   new(true),
				ReadOnlyHint:    false,
			},
		}, cobramcp.GenToolHandlerWithMRTR(
			insertTool, cobramcp.ConfirmThen(
				func(input caption.Caption) string {
					return fmt.Sprintf(insertConfirm, input.VideoId)
				},
				func(input caption.Caption, w io.Writer) error {
					return input.Insert(w)
				},
			),
		),
	)
	captionCmd.AddCommand(insertCmd)

	insertCmd.Flags().StringVarP(&file, "file", "f", "", fileUsage)
	insertCmd.Flags().StringVarP(
		&audioTrackType, "audioTrackType", "a", "unknown", attUsage,
	)
	insertCmd.Flags().BoolVarP(
		isAutoSynced, "isAutoSynced", "A", true, iasUsage,
	)
	insertCmd.Flags().BoolVarP(isCC, "isCC", "C", false, iscUsage)
	insertCmd.Flags().BoolVarP(isDraft, "isDraft", "D", false, isdUsage)
	insertCmd.Flags().BoolVarP(
		isEasyReader, "isEasyReader", "E", false, iserUsage,
	)
	insertCmd.Flags().BoolVarP(isLarge, "isLarge", "L", false, islUsage)
	insertCmd.Flags().StringVarP(&language, "language", "l", "", languageUsage)
	insertCmd.Flags().StringVarP(&name, "name", "n", "", nameUsage)
	insertCmd.Flags().StringVarP(
		&trackKind, "trackKind", "t", "standard", tkUsage,
	)
	insertCmd.Flags().StringVarP(&videoId, "videoId", "v", "", vidUsage)
	insertCmd.Flags().StringVarP(&onBehalfOf, "onBehalfOf", "b", "", pkg.OBOUsage)
	insertCmd.Flags().StringVarP(
		&onBehalfOfContentOwner, "onBehalfOfContentOwner", "B", "", pkg.OBOCOUsage,
	)
	insertCmd.Flags().StringP("output", "o", "", pkg.SilentUsage)
	insertCmd.Flags().Bool("yes", false, pkg.ConfirmedUsage)
	_ = insertCmd.MarkFlagRequired("file")
	_ = insertCmd.MarkFlagRequired("videoId")
}

var insertCmd = &cobra.Command{
	Use:     "insert",
	Short:   insertShort,
	Long:    insertLong,
	Example: insertExample,
	PreRunE: func(c *cobra.Command, _ []string) error {
		return utils.ConfirmPreRun(c, fmt.Sprintf(insertConfirm, videoId))
	},
	Run: func(c *cobra.Command, _ []string) {
		output, _ := c.Flags().GetString("output")
		input := caption.NewCaption(
			caption.WithFile(file),
			caption.WithAudioTrackType(audioTrackType),
			caption.WithIsAutoSynced(isAutoSynced),
			caption.WithIsCC(isCC),
			caption.WithIsDraft(isDraft),
			caption.WithIsEasyReader(isEasyReader),
			caption.WithIsLarge(isLarge),
			caption.WithLanguage(language),
			caption.WithName(name),
			caption.WithTrackKind(trackKind),
			caption.WithVideoId(videoId),
			caption.WithOnBehalfOf(onBehalfOf),
			caption.WithOnBehalfOfContentOwner(onBehalfOfContentOwner),
			caption.WithOutput(output),
		)
		utils.HandleCmdError(input.Insert(c.OutOrStdout()), c)
	},
}