summaryrefslogtreecommitdiffhomepage
path: root/packages/tool-youtube-transcript/src/tool.ts
blob: b24a6c51d880046245978ed19ce22d48b4dd4b7e (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
/**
 * youtube_transcript tool factory — the imperative shell that binds the pure
 * validate/format functions to the injected TranscriptClient edge.
 *
 * Mirrors the tool-web-search pattern: factory + injected dep + pure helpers +
 * a `ToolResult` returned per call. Errors surface as `{ isError: true }`
 * rather than thrown, so the model can react to the message.
 */

import { mkdirSync, writeFileSync } from "node:fs";
import type { ToolContract, ToolExecuteContext, ToolResult } from "@dispatch/kernel";
import type { TranscriptClient } from "./client.js";
import {
  formatCompleted,
  formatFailed,
  formatQueued,
  type TranscriptResponse,
  truncateOutput,
} from "./format.js";
import { validateUrl } from "./validate.js";

const OUTPUT_CAP = 50_000;
const FULL_OUTPUT_DIR = "/tmp/dispatch/youtube-transcribe";

export interface YoutubeTranscriptToolDeps {
  readonly client: TranscriptClient;
  readonly outputCap?: number;
  /** Injected file writer (defaults to real fs write). */
  readonly writeFile?: (path: string, content: string) => void;
}

const DESCRIPTION =
  "Fetch the transcript/subtitles for a YouTube video from the local transcriber " +
  "service. If the transcript has not been downloaded before, the video will be " +
  "queued for processing and the tool will return the estimated time when the " +
  "transcript will be available. Once available, the tool returns the transcript " +
  "text and timestamped segments (truncated if very long). The full transcript " +
  "is always saved to /tmp/dispatch/youtube-transcribe/{video_id}.txt — use " +
  "read_file to access it. Accepted URL formats: " +
  "youtube.com/watch?v=, youtu.be/, youtube.com/embed/, youtube.com/shorts/";

/**
 * Create the `youtube_transcript` tool. `concurrencySafe: true` — transcript
 * fetches are idempotent and safe to run alongside other tools. The `network`
 * capability is declared on the extension manifest (not the tool contract).
 */
export function createYoutubeTranscriptTool(deps: YoutubeTranscriptToolDeps): ToolContract {
  const client = deps.client;
  const cap = deps.outputCap ?? OUTPUT_CAP;
  const writeFile =
    deps.writeFile ??
    ((path, content) => {
      mkdirSync(FULL_OUTPUT_DIR, { recursive: true });
      writeFileSync(path, content, "utf-8");
    });

  return {
    name: "youtube_transcript",
    description: DESCRIPTION,
    parameters: {
      type: "object",
      properties: {
        url: {
          type: "string",
          description:
            "YouTube video URL (e.g. https://www.youtube.com/watch?v=... or https://youtu.be/...)",
        },
      },
      required: ["url"],
    },
    concurrencySafe: true,
    async execute(args: unknown, ctx: ToolExecuteContext): Promise<ToolResult> {
      const validated = validateUrl(args);
      if (typeof validated !== "string") {
        return { content: validated.error, isError: true };
      }
      const url = validated;
      const span = ctx.log.span("youtube_transcript.execute", { url });
      try {
        const data: TranscriptResponse = await client.getTranscript(url, ctx.signal);
        let output: string;
        let videoId: string | undefined;
        // Check the single-literal discriminants ("completed"/"failed") first,
        // so the final else narrows to QueuedResponse — whose `status` is itself
        // a `"queued" | "processing"` union TS cannot negatively narrow.
        if (data.status === "completed") {
          output = formatCompleted(url, data);
          videoId = data.video_id;
        } else if (data.status === "failed") {
          output = formatFailed(data);
        } else {
          output = formatQueued(url, data, Date.now);
        }
        span.end();

        if (videoId !== undefined) {
          const filePath = `${FULL_OUTPUT_DIR}/${videoId}.txt`;
          try {
            writeFile(filePath, output);
          } catch {
            // File write failed — continue with truncated output only.
          }
          if (output.length > cap) {
            return { content: truncateOutput(output, cap, filePath) };
          }
        }
        return { content: truncateOutput(output, cap) };
      } catch (err: unknown) {
        span.end({ err });
        return {
          content: `Error: ${err instanceof Error ? err.message : String(err)}`,
          isError: true,
        };
      }
    },
  };
}