@koda-sl/baker-cli 0.513.0 → 0.514.0-dev.555f7007e

package/dist/transcript-CGPX7Y37.js

@@ -0,0 +1,165 @@
+import {
+  transcriptHints
+} from "./chunk-3CHYJYSS.js";
+import {
+  handleResearchError,
+  writeResearchJson
+} from "./chunk-GSKZ3ZDO.js";
+import "./chunk-NZ34KM46.js";
+import "./chunk-CAZGM2NX.js";
+import "./chunk-NZS3KEQ2.js";
+import "./chunk-GIWHHCD6.js";
+import "./chunk-SOPVIIUW.js";
+import "./chunk-G4FG5EQO.js";
+import "./chunk-NSBNIPIW.js";
+import "./chunk-MUS2UTWU.js";
+import {
+  apiPost
+} from "./chunk-X7VQFLHI.js";
+import "./chunk-LU4HI4KI.js";
+import "./chunk-LGK7UWBN.js";
+import {
+  registerSchema
+} from "./chunk-Q7US4OOA.js";
+import "./chunk-YQSBHWOI.js";
+
+// src/commands/research/transcript.ts
+import { mkdirSync, writeFileSync } from "fs";
+import { dirname, resolve } from "path";
+import { defineCommand } from "citty";
+var EXCERPT_CHARS = 3e3;
+var TIMEOUT_MS = 54e4;
+registerSchema({
+  command: "research.transcript",
+  description: "The words of one long recording \u2014 a keynote, earnings call, webinar or podcast \u2014 so a leader's own phrasing can feed 1:1 ad and page messaging. A YouTube link is read from its captions; any other link must be the audio or video file itself. Nothing is added to the video library. Start here: `baker research transcript <url> --out transcript.txt`.",
+  args: {
+    url: {
+      type: "string",
+      description: "A YouTube link, or a direct link to an audio/video file (.mp3, .mp4, .m4a, \u2026) (positional)",
+      required: true
+    },
+    full: {
+      type: "boolean",
+      description: `Return the whole text instead of the first ${EXCERPT_CHARS} characters`,
+      required: false,
+      default: false
+    },
+    out: {
+      type: "string",
+      description: "Write the whole text to this file and return its path instead of the text",
+      required: false
+    }
+  }
+});
+var FIELDS = {
+  provider: "youtube \u2014 the video's captions; deepgram \u2014 the recording transcribed",
+  title: "The video's title (YouTube only), or null",
+  duration_sec: "Length of the recording in seconds, or null when unknown",
+  language: "The language spoken, as the provider names it",
+  word_count: "Words in the whole recording",
+  text_chars: "Length of the whole text available on the server",
+  excerpt: "The opening of the text (default output)",
+  text: "The whole text (--full)",
+  file: "Where the whole text was written (--out)",
+  truncated: "The recording was longer than one transcript carries, so the end is missing"
+};
+function transcriptFix(code) {
+  if (code === "VALIDATION_ERROR" || code === "BAD_REQUEST") {
+    return {
+      action: "retry_with_changes",
+      explanation: "Pass a YouTube link, or the direct link to the audio or video file \u2014 for a podcast, the episode's download link (its RSS enclosure), not the episode page."
+    };
+  }
+  if (code === "CONFLICT") {
+    return {
+      action: "continue_without",
+      explanation: "There are no words to read in this recording. Look for another recording of the same person, or write the brief from what you have."
+    };
+  }
+  if (code === "RATE_LIMITED") {
+    return {
+      action: "continue_without",
+      explanation: "The transcriber is busy. Do the rest of the job first and try this recording once more later."
+    };
+  }
+  return {
+    action: "continue_without",
+    explanation: "This recording couldn't be transcribed. Don't loop on it \u2014 finish the brief from your other sources and name this recording as a gap."
+  };
+}
+var transcriptCommand = defineCommand({
+  meta: {
+    name: "transcript",
+    description: `The words of one long recording \u2014 a keynote, earnings call, webinar or podcast \u2014 so a leader's own phrasing can feed 1:1 ad and page messaging.
+
+A YouTube link is read from its captions. Any other link must be the audio or video file itself (.mp3, .mp4, .m4a, \u2026): for a podcast, the episode's download link, not its page. Nothing is added to the video library.
+
+Default output is compact \u2014 title, length, word count and the opening of the text. Use --out to write the whole text to a file and read the parts you need.
+
+Examples:
+  baker research transcript "https://www.youtube.com/watch?v=<id>"
+  baker research transcript "https://www.youtube.com/watch?v=<id>" --out keynote.txt
+  baker research transcript "https://traffic.megaphone.fm/<episode>.mp3" --full`
+  },
+  args: {
+    url: { type: "positional", description: "A YouTube link, or a direct audio/video file link", required: false },
+    full: { type: "boolean", description: "Return the whole text", required: false },
+    out: { type: "string", description: "Write the whole text to this file", required: false }
+  },
+  run: async ({ args }) => {
+    const url = args.url;
+    if (!url) {
+      writeResearchJson({
+        ok: false,
+        error: {
+          code: "VALIDATION_ERROR",
+          message: "A recording link is required",
+          fix: transcriptFix("VALIDATION_ERROR")
+        }
+      });
+      process.exit(1);
+    }
+    const body = { url };
+    try {
+      const transcript = await apiPost("/api/research/transcript", body, {
+        timeoutMs: TIMEOUT_MS
+      });
+      const out = args.out;
+      let delivered;
+      let excerpted = false;
+      if (out) {
+        const file = resolve(out);
+        mkdirSync(dirname(file), { recursive: true });
+        writeFileSync(file, transcript.text, "utf-8");
+        delivered = { file };
+      } else if (args.full) {
+        delivered = { text: transcript.text };
+      } else {
+        excerpted = transcript.text.length > EXCERPT_CHARS;
+        delivered = { excerpt: transcript.text.slice(0, EXCERPT_CHARS) };
+      }
+      writeResearchJson({
+        ok: true,
+        data: {
+          url: transcript.url,
+          provider: transcript.provider,
+          title: transcript.title,
+          duration_sec: transcript.durationSec,
+          language: transcript.language,
+          word_count: transcript.wordCount,
+          text_chars: transcript.textChars,
+          truncated: transcript.truncated,
+          ...delivered
+        },
+        fields: FIELDS,
+        hints: transcriptHints({ excerpted, truncated: transcript.truncated })
+      });
+    } catch (err) {
+      handleResearchError(err, transcriptFix);
+    }
+  }
+});
+export {
+  transcriptCommand
+};
+//# sourceMappingURL=transcript-CGPX7Y37.js.map

package/dist/transcript-CGPX7Y37.js.map

@@ -0,0 +1 @@
+{"version":3,"sources":["../src/commands/research/transcript.ts"],"sourcesContent":["import { mkdirSync, writeFileSync } from \"node:fs\";\nimport { dirname, resolve } from \"node:path\";\nimport type { ResearchTranscriptRequest, ResearchTranscriptResponse } from \"@baker/api\";\nimport { defineCommand } from \"citty\";\nimport { apiPost } from \"../../client.ts\";\nimport { registerSchema } from \"../../schemas.ts\";\nimport { transcriptHints } from \"./hints.ts\";\nimport { handleResearchError, type ResearchFix, writeResearchJson } from \"./output.ts\";\n\n/**\n * A keynote is tens of thousands of words, more than a turn wants to carry, so\n * the default is the opening — enough to tell whether it is the right talk.\n */\nconst EXCERPT_CHARS = 3_000;\n\n/** An hour-long recording takes Deepgram well under a minute, but the default 60s leaves no margin. */\nconst TIMEOUT_MS = 540_000;\n\nregisterSchema({\n  command: \"research.transcript\",\n  description:\n    \"The words of one long recording — a keynote, earnings call, webinar or podcast — so a leader's own phrasing can feed 1:1 ad and page messaging. A YouTube link is read from its captions; any other link must be the audio or video file itself. Nothing is added to the video library. Start here: `baker research transcript <url> --out transcript.txt`.\",\n  args: {\n    url: {\n      type: \"string\",\n      description: \"A YouTube link, or a direct link to an audio/video file (.mp3, .mp4, .m4a, …) (positional)\",\n      required: true,\n    },\n    full: {\n      type: \"boolean\",\n      description: `Return the whole text instead of the first ${EXCERPT_CHARS} characters`,\n      required: false,\n      default: false,\n    },\n    out: {\n      type: \"string\",\n      description: \"Write the whole text to this file and return its path instead of the text\",\n      required: false,\n    },\n  },\n});\n\nconst FIELDS: Record<string, string> = {\n  provider: \"youtube — the video's captions; deepgram — the recording transcribed\",\n  title: \"The video's title (YouTube only), or null\",\n  duration_sec: \"Length of the recording in seconds, or null when unknown\",\n  language: \"The language spoken, as the provider names it\",\n  word_count: \"Words in the whole recording\",\n  text_chars: \"Length of the whole text available on the server\",\n  excerpt: \"The opening of the text (default output)\",\n  text: \"The whole text (--full)\",\n  file: \"Where the whole text was written (--out)\",\n  truncated: \"The recording was longer than one transcript carries, so the end is missing\",\n};\n\nfunction transcriptFix(code: string): ResearchFix {\n  if (code === \"VALIDATION_ERROR\" || code === \"BAD_REQUEST\") {\n    return {\n      action: \"retry_with_changes\",\n      explanation:\n        \"Pass a YouTube link, or the direct link to the audio or video file — for a podcast, the episode's download link (its RSS enclosure), not the episode page.\",\n    };\n  }\n  if (code === \"CONFLICT\") {\n    return {\n      action: \"continue_without\",\n      explanation:\n        \"There are no words to read in this recording. Look for another recording of the same person, or write the brief from what you have.\",\n    };\n  }\n  if (code === \"RATE_LIMITED\") {\n    return {\n      action: \"continue_without\",\n      explanation: \"The transcriber is busy. Do the rest of the job first and try this recording once more later.\",\n    };\n  }\n  return {\n    action: \"continue_without\",\n    explanation:\n      \"This recording couldn't be transcribed. Don't loop on it — finish the brief from your other sources and name this recording as a gap.\",\n  };\n}\n\nexport const transcriptCommand = defineCommand({\n  meta: {\n    name: \"transcript\",\n    description: `The words of one long recording — a keynote, earnings call, webinar or podcast — so a leader's own phrasing can feed 1:1 ad and page messaging.\n\nA YouTube link is read from its captions. Any other link must be the audio or video file itself (.mp3, .mp4, .m4a, …): for a podcast, the episode's download link, not its page. Nothing is added to the video library.\n\nDefault output is compact — title, length, word count and the opening of the text. Use --out to write the whole text to a file and read the parts you need.\n\nExamples:\n  baker research transcript \"https://www.youtube.com/watch?v=<id>\"\n  baker research transcript \"https://www.youtube.com/watch?v=<id>\" --out keynote.txt\n  baker research transcript \"https://traffic.megaphone.fm/<episode>.mp3\" --full`,\n  },\n  args: {\n    url: { type: \"positional\", description: \"A YouTube link, or a direct audio/video file link\", required: false },\n    full: { type: \"boolean\", description: \"Return the whole text\", required: false },\n    out: { type: \"string\", description: \"Write the whole text to this file\", required: false },\n  },\n  run: async ({ args }) => {\n    const url = args.url as string | undefined;\n    if (!url) {\n      writeResearchJson({\n        ok: false,\n        error: {\n          code: \"VALIDATION_ERROR\",\n          message: \"A recording link is required\",\n          fix: transcriptFix(\"VALIDATION_ERROR\"),\n        },\n      });\n      process.exit(1);\n    }\n\n    const body: ResearchTranscriptRequest = { url };\n\n    try {\n      const transcript = await apiPost<ResearchTranscriptResponse>(\"/api/research/transcript\", body, {\n        timeoutMs: TIMEOUT_MS,\n      });\n      const out = args.out as string | undefined;\n      let delivered: Record<string, string>;\n      let excerpted = false;\n      if (out) {\n        const file = resolve(out);\n        mkdirSync(dirname(file), { recursive: true });\n        writeFileSync(file, transcript.text, \"utf-8\");\n        delivered = { file };\n      } else if (args.full) {\n        delivered = { text: transcript.text };\n      } else {\n        excerpted = transcript.text.length > EXCERPT_CHARS;\n        delivered = { excerpt: transcript.text.slice(0, EXCERPT_CHARS) };\n      }\n\n      writeResearchJson({\n        ok: true,\n        data: {\n          url: transcript.url,\n          provider: transcript.provider,\n          title: transcript.title,\n          duration_sec: transcript.durationSec,\n          language: transcript.language,\n          word_count: transcript.wordCount,\n          text_chars: transcript.textChars,\n          truncated: transcript.truncated,\n          ...delivered,\n        },\n        fields: FIELDS,\n        hints: transcriptHints({ excerpted, truncated: transcript.truncated }),\n      });\n    } catch (err) {\n      handleResearchError(err, transcriptFix);\n    }\n  },\n});\n"],"mappings":";;;;;;;;;;;;;;;;;;;;;;;;;;AAAA,SAAS,WAAW,qBAAqB;AACzC,SAAS,SAAS,eAAe;AAEjC,SAAS,qBAAqB;AAU9B,IAAM,gBAAgB;AAGtB,IAAM,aAAa;AAEnB,eAAe;AAAA,EACb,SAAS;AAAA,EACT,aACE;AAAA,EACF,MAAM;AAAA,IACJ,KAAK;AAAA,MACH,MAAM;AAAA,MACN,aAAa;AAAA,MACb,UAAU;AAAA,IACZ;AAAA,IACA,MAAM;AAAA,MACJ,MAAM;AAAA,MACN,aAAa,8CAA8C,aAAa;AAAA,MACxE,UAAU;AAAA,MACV,SAAS;AAAA,IACX;AAAA,IACA,KAAK;AAAA,MACH,MAAM;AAAA,MACN,aAAa;AAAA,MACb,UAAU;AAAA,IACZ;AAAA,EACF;AACF,CAAC;AAED,IAAM,SAAiC;AAAA,EACrC,UAAU;AAAA,EACV,OAAO;AAAA,EACP,cAAc;AAAA,EACd,UAAU;AAAA,EACV,YAAY;AAAA,EACZ,YAAY;AAAA,EACZ,SAAS;AAAA,EACT,MAAM;AAAA,EACN,MAAM;AAAA,EACN,WAAW;AACb;AAEA,SAAS,cAAc,MAA2B;AAChD,MAAI,SAAS,sBAAsB,SAAS,eAAe;AACzD,WAAO;AAAA,MACL,QAAQ;AAAA,MACR,aACE;AAAA,IACJ;AAAA,EACF;AACA,MAAI,SAAS,YAAY;AACvB,WAAO;AAAA,MACL,QAAQ;AAAA,MACR,aACE;AAAA,IACJ;AAAA,EACF;AACA,MAAI,SAAS,gBAAgB;AAC3B,WAAO;AAAA,MACL,QAAQ;AAAA,MACR,aAAa;AAAA,IACf;AAAA,EACF;AACA,SAAO;AAAA,IACL,QAAQ;AAAA,IACR,aACE;AAAA,EACJ;AACF;AAEO,IAAM,oBAAoB,cAAc;AAAA,EAC7C,MAAM;AAAA,IACJ,MAAM;AAAA,IACN,aAAa;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAUf;AAAA,EACA,MAAM;AAAA,IACJ,KAAK,EAAE,MAAM,cAAc,aAAa,qDAAqD,UAAU,MAAM;AAAA,IAC7G,MAAM,EAAE,MAAM,WAAW,aAAa,yBAAyB,UAAU,MAAM;AAAA,IAC/E,KAAK,EAAE,MAAM,UAAU,aAAa,qCAAqC,UAAU,MAAM;AAAA,EAC3F;AAAA,EACA,KAAK,OAAO,EAAE,KAAK,MAAM;AACvB,UAAM,MAAM,KAAK;AACjB,QAAI,CAAC,KAAK;AACR,wBAAkB;AAAA,QAChB,IAAI;AAAA,QACJ,OAAO;AAAA,UACL,MAAM;AAAA,UACN,SAAS;AAAA,UACT,KAAK,cAAc,kBAAkB;AAAA,QACvC;AAAA,MACF,CAAC;AACD,cAAQ,KAAK,CAAC;AAAA,IAChB;AAEA,UAAM,OAAkC,EAAE,IAAI;AAE9C,QAAI;AACF,YAAM,aAAa,MAAM,QAAoC,4BAA4B,MAAM;AAAA,QAC7F,WAAW;AAAA,MACb,CAAC;AACD,YAAM,MAAM,KAAK;AACjB,UAAI;AACJ,UAAI,YAAY;AAChB,UAAI,KAAK;AACP,cAAM,OAAO,QAAQ,GAAG;AACxB,kBAAU,QAAQ,IAAI,GAAG,EAAE,WAAW,KAAK,CAAC;AAC5C,sBAAc,MAAM,WAAW,MAAM,OAAO;AAC5C,oBAAY,EAAE,KAAK;AAAA,MACrB,WAAW,KAAK,MAAM;AACpB,oBAAY,EAAE,MAAM,WAAW,KAAK;AAAA,MACtC,OAAO;AACL,oBAAY,WAAW,KAAK,SAAS;AACrC,oBAAY,EAAE,SAAS,WAAW,KAAK,MAAM,GAAG,aAAa,EAAE;AAAA,MACjE;AAEA,wBAAkB;AAAA,QAChB,IAAI;AAAA,QACJ,MAAM;AAAA,UACJ,KAAK,WAAW;AAAA,UAChB,UAAU,WAAW;AAAA,UACrB,OAAO,WAAW;AAAA,UAClB,cAAc,WAAW;AAAA,UACzB,UAAU,WAAW;AAAA,UACrB,YAAY,WAAW;AAAA,UACvB,YAAY,WAAW;AAAA,UACvB,WAAW,WAAW;AAAA,UACtB,GAAG;AAAA,QACL;AAAA,QACA,QAAQ;AAAA,QACR,OAAO,gBAAgB,EAAE,WAAW,WAAW,WAAW,UAAU,CAAC;AAAA,MACvE,CAAC;AAAA,IACH,SAAS,KAAK;AACZ,0BAAoB,KAAK,aAAa;AAAA,IACxC;AAAA,EACF;AACF,CAAC;","names":[]}