@koda-sl/baker-cli 0.476.0 → 0.477.0

package/dist/chunk-LKM3FMYC.js

@@ -0,0 +1,363 @@
+import {
+  MAX_VIDEO_INGEST_BYTES,
+  MAX_VIDEO_INGEST_SECONDS,
+  directDedupeKey,
+  downloadTimeoutMs,
+  isDirectMediaUrl,
+  isDrmProtected,
+  isRegionBlocked,
+  isToolingOrBlockFailure,
+  refuseByDuration,
+  resolveIngestSource,
+  worthDownloadingInstead
+} from "./chunk-FKOECEVC.js";
+import {
+  YtDlpError,
+  canonicalUrl,
+  durationSeconds,
+  isAudioOnly,
+  isVimeoAuthFailure,
+  probeYtDlp,
+  runYtDlp,
+  ytDlpBlockSignal
+} from "./chunk-CPTOHQSJ.js";
+import {
+  videosIngestRequestSchema
+} from "./chunk-GAJQPQWS.js";
+import {
+  ApiError,
+  apiPost
+} from "./chunk-7ANWFQEA.js";
+import {
+  captureVimeoCookie,
+  isProxyFailure
+} from "./chunk-H5CGVMQW.js";
+import {
+  registerSchema,
+  writeJson
+} from "./chunk-ANYMWCST.js";
+
+// src/commands/videos/ingest.ts
+import { mkdtemp, rm, stat } from "fs/promises";
+import { tmpdir } from "os";
+import path from "path";
+import { defineCommand } from "citty";
+
+// src/lib/streamUpload.ts
+import { createHash } from "crypto";
+import { createReadStream } from "fs";
+import { Readable } from "stream";
+import { pipeline } from "stream/promises";
+async function hashFile(filePath) {
+  const hash = createHash("sha256");
+  await pipeline(createReadStream(filePath), hash);
+  return hash.digest("hex");
+}
+async function putFileStreamed(args) {
+  const body = Readable.toWeb(createReadStream(args.filePath));
+  return await fetch(args.putUrl, {
+    method: "PUT",
+    body,
+    headers: {
+      "Content-Type": args.contentType,
+      "Content-Length": String(args.sizeBytes)
+    },
+    duplex: "half"
+  });
+}
+
+// src/commands/videos/ingest.ts
+registerSchema({
+  command: "videos.ingest",
+  description: "Add a video to the library from a URL \u2014 a direct file, or a YouTube/TikTok/Vimeo page.",
+  args: {
+    url: { type: "string", description: "Video URL (direct file or a watch/post page)", required: true },
+    source: { type: "string", description: "Provenance: url | youtube", required: false },
+    "external-id": { type: "string", description: "Provider asset id (deduped on)", required: false },
+    "external-url": { type: "string", description: "Canonical page URL", required: false },
+    download: {
+      type: "boolean",
+      description: "Force the download path instead of letting Baker fetch the URL directly",
+      required: false,
+      default: false
+    },
+    country: {
+      type: "string",
+      description: "Two-letter country to fetch from (e.g. US, DE, JP) \u2014 use when a video is blocked in your region",
+      required: false
+    },
+    context: {
+      type: "string",
+      description: "What you know that the video cannot show \u2014 the campaign, whether the person on camera is a real customer, what it is for. Highest-leverage flag here: it is what makes the clip findable later.",
+      required: false
+    },
+    "dry-run": { type: "boolean", description: "Preview the operation without executing", required: false }
+  }
+});
+var ingestCommand = defineCommand({
+  meta: {
+    name: "ingest",
+    description: "Add a video to the library from a URL. A direct file URL is handed straight to Baker, which fetches it. A page URL (YouTube, TikTok, Vimeo, Instagram) is downloaded here first, then uploaded \u2014 and a direct URL that Baker cannot fetch falls back to that same path automatically.\n\nExample: baker videos ingest https://www.youtube.com/watch?v=abc123"
+  },
+  args: {
+    url: { type: "positional", description: "Video URL", required: false },
+    source: { type: "string", description: "Provenance: url | youtube", required: false },
+    "external-id": { type: "string", description: "Provider asset id", required: false },
+    "external-url": { type: "string", description: "Canonical page URL", required: false },
+    download: { type: "boolean", description: "Force the download path", required: false, default: false },
+    country: { type: "string", description: "Two-letter country to fetch from, e.g. US", required: false },
+    context: {
+      type: "string",
+      description: "Context for the analysis \u2014 campaign, who is on camera, intended use",
+      required: false
+    },
+    "dry-run": { type: "boolean", description: "Preview without executing", required: false, default: false }
+  },
+  run: ({ args }) => runVideoIngest({
+    url: args.url,
+    source: args.source,
+    externalId: args["external-id"],
+    externalUrl: args["external-url"],
+    forceDownload: args.download === true,
+    country: args.country,
+    context: args.context,
+    dryRun: args["dry-run"] === true
+  })
+});
+async function runVideoIngest(opts) {
+  const url = opts.url;
+  if (!url) {
+    writeJson({ ok: false, error: { code: "VALIDATION_ERROR", message: "URL is required" } });
+    process.exit(1);
+  }
+  const direct = !opts.forceDownload && isDirectMediaUrl(url);
+  const country = opts.country?.trim() || void 0;
+  const args = {
+    url,
+    source: resolveIngestSource(opts.source, url),
+    externalId: opts.externalId,
+    // Deliberately NOT pre-filled with the pasted URL. Doing so shadowed the
+    // canonical URL the probe resolves (`?? canonicalUrl(...)` below could never
+    // run), so the same video filed under a different `externalUrl` depending on
+    // whether the link was copied from a share sheet, a shorts page or the
+    // address bar. `downloadThenIngest` falls back to `args.url` when the probe
+    // resolves nothing, which is the value this used to hard-code.
+    externalUrl: opts.externalUrl,
+    descriptionContext: opts.context?.trim() || void 0
+  };
+  const validation = videosIngestRequestSchema.safeParse({ ...args, url });
+  if (!validation.success) {
+    const problem = validation.error.issues[0];
+    writeJson({
+      ok: false,
+      error: {
+        code: "VALIDATION_ERROR",
+        message: problem ? `${problem.path.join(".") || "request"}: ${problem.message}` : "Invalid request"
+      }
+    });
+    process.exit(1);
+  }
+  if (opts.dryRun) {
+    reportDryRun(args, direct, country);
+    return;
+  }
+  try {
+    const data = await ingestByRoute(args, direct, country);
+    writeJson({ ok: true, data, hints: ingestHints(data) });
+  } catch (err) {
+    reportIngestFailure(err);
+  }
+}
+function ingestHints(data) {
+  if (data.reducedQuality !== true) return void 0;
+  return [
+    "Added, but at standard quality (360p): the site refused to release the full-quality file and this was the best it would hand over. Tell the user \u2014 the clip is in the library and usable, it is just not high-resolution. If they need the crisp version, ask them for the original file and use `baker videos upload <file>`."
+  ];
+}
+function reportDryRun(args, direct, country) {
+  writeJson({
+    ok: true,
+    dryRun: true,
+    operation: "videos.ingest",
+    params: {
+      url: args.url,
+      source: args.source,
+      // Known up front on the direct route, because it is derived from the URL
+      // alone. The download route's key comes from the probe, which a dry run
+      // does not perform, so there it stays null rather than becoming a guess.
+      externalId: args.externalId ?? (direct ? directDedupeKey(args.source, args.url) : null),
+      // A page URL's canonical form is only known after the probe, which a dry
+      // run does not perform — so this is null rather than a guess.
+      externalUrl: args.externalUrl ?? null,
+      direct,
+      country: country ?? null,
+      descriptionContext: args.descriptionContext ?? null
+    }
+  });
+}
+async function ingestByRoute(args, direct, country) {
+  if (!direct) {
+    return { ...await downloadThenIngest(args, country), via: "download" };
+  }
+  try {
+    const externalId = args.externalId ?? directDedupeKey(args.source, args.url);
+    return { ...await ingestUrl({ ...args, externalId }), via: "direct" };
+  } catch (err) {
+    if (!(err instanceof ApiError) || !worthDownloadingInstead(err.code)) throw err;
+    return { ...await downloadThenIngest(args, country), via: "download", fallbackFrom: err.message };
+  }
+}
+function reportYtDlpFailure(err) {
+  const detail = `${err.stderrTail} ${err.message}`;
+  if (isProxyFailure(ytDlpBlockSignal(detail))) {
+    fail(
+      "Couldn't download that video: our connection to the internet was refused.",
+      "This is Baker's egress proxy, not the link \u2014 its credentials look wrong or expired. The link is probably fine. Report it; retrying won't help until the proxy is fixed."
+    );
+  }
+  if (isDrmProtected(detail)) {
+    fail(
+      "That video is copy-protected, so it can't be added to the library.",
+      "The publisher encrypted this one against downloading. Nothing will change that \u2014 not retrying, not a different region. Ask whoever owns the video for the original file and add it with `baker videos upload <file>`."
+    );
+  }
+  if (isVimeoAuthFailure(detail)) {
+    fail(
+      "Vimeo wouldn't hand over that video \u2014 Baker isn't signed in to Vimeo.",
+      "Vimeo now refuses anonymous downloads entirely, so this needs a signed-in session and the one Baker holds has either expired or was never set. Nothing is wrong with the link. Report it so the Vimeo cookie can be refreshed; meanwhile download the file and use `baker videos upload <file>`."
+    );
+  }
+  if (isRegionBlocked(detail)) {
+    fail(
+      `That video isn't available in the region we're fetching from. ${err.stderrTail || err.message}`,
+      "Re-run with `--country <two-letter code>` for a country where the video plays \u2014 e.g. `--country US`. Pick the one the video belongs to (the uploader's country is the usual answer); each attempt costs a paid connection, so choose rather than work through the list."
+    );
+  }
+  fail(
+    `Couldn't download that video. ${err.stderrTail || err.message}`,
+    isToolingOrBlockFailure(detail) ? "The site wouldn't hand over the video \u2014 usually the downloader being out of date against a site that changed, or the request being blocked. Retrying won't help, and the link is probably fine. Report it so the tool can be updated; meanwhile download the file and use `baker videos upload <file>`." : "Check the link is public and playable. Private or age-restricted videos can't be downloaded \u2014 download the file yourself and use `baker videos upload <file>`."
+  );
+}
+function fail(message, fix, code = "INGEST_FAILED") {
+  writeJson({ ok: false, error: { code, message, fix } });
+  process.exit(1);
+}
+function reportIngestFailure(err) {
+  if (err instanceof ApiError) {
+    writeJson({ ok: false, error: { code: err.code, message: err.message } });
+    process.exit(1);
+  }
+  if (err instanceof YtDlpError) reportYtDlpFailure(err);
+  const detail = err instanceof Error ? err.message : String(err);
+  if (detail.includes("timed out after")) {
+    fail(
+      "That video took too long to download and was stopped part-way.",
+      "Long or slow-serving videos can outrun the download window. Try a shorter clip, or download this one yourself and use `baker videos upload <file>`.",
+      "INTERNAL_ERROR"
+    );
+  }
+  fail(
+    `Couldn't add that video. ${detail}`,
+    "This is a fault on Baker's side, not a problem with the link. Report it.",
+    "INTERNAL_ERROR"
+  );
+}
+function dedupeKey(args, info) {
+  const id = typeof info.id === "string" ? info.id : void 0;
+  if (!id) return void 0;
+  if (info.extractor === "generic") return `${args.source}:${canonicalUrl(info) ?? args.url}`;
+  return `${args.source}:${id}`;
+}
+function ingestUrl(args) {
+  const body = {
+    url: args.url,
+    source: args.source
+  };
+  if (args.externalId) body.externalId = args.externalId;
+  if (args.externalUrl) body.externalUrl = args.externalUrl;
+  if (args.descriptionContext) body.descriptionContext = args.descriptionContext;
+  return apiPost("/api/videos/ingest", body);
+}
+async function downloadThenIngest(args, country) {
+  const vimeoCookie = captureVimeoCookie();
+  const workDir = await mkdtemp(path.join(tmpdir(), "videos-ingest-"));
+  try {
+    const probe = await probeYtDlp({ url: args.url, country, vimeoCookie, cookieDir: workDir });
+    if (isAudioOnly(probe.info)) {
+      throw new ApiError(
+        "VALIDATION_ERROR",
+        "That link is audio, not video, so it can't go in the video library. Music and podcast links have no picture to store."
+      );
+    }
+    const tooLong = refuseByDuration(durationSeconds(probe.info), MAX_VIDEO_INGEST_SECONDS);
+    if (tooLong) {
+      throw new ApiError("VALIDATION_ERROR", tooLong);
+    }
+    const externalId = args.externalId ?? dedupeKey(args, probe.info);
+    const { filePath, tier, reducedQuality } = await runYtDlp({
+      url: args.url,
+      audioOnly: false,
+      workDir,
+      // Deliberately uncapped: take the best the site will release and let Mux
+      // decide what to serve. The 1080p ceiling this replaces was reasoning
+      // about the wrong cost — it saved download bytes by throwing away source
+      // pixels *before* the one system that can make renditions from them ever
+      // saw them, so a 4K master became a 1080p master permanently, for
+      // everything the library is ever asked to do with that clip afterwards.
+      // Downscaling is Mux's job and it is cheap; re-acquiring a master is not.
+      country,
+      vimeoCookie,
+      // Budgeted from the probed length, so the duration we accept and the time
+      // we allow to fetch it cannot drift apart.
+      timeoutMs: downloadTimeoutMs(durationSeconds(probe.info))
+    });
+    const stats = await stat(filePath);
+    if (stats.size > MAX_VIDEO_INGEST_BYTES) {
+      throw new ApiError(
+        "VALIDATION_ERROR",
+        `That video is ${(stats.size / 1024 ** 3).toFixed(1)} GB, over the ${Math.round(MAX_VIDEO_INGEST_BYTES / 1024 ** 3)} GB limit for this route.`
+      );
+    }
+    const publicUrl = await uploadToAssetStore(filePath, stats.size);
+    return {
+      ...await ingestUrl({
+        ...args,
+        url: publicUrl,
+        externalId,
+        externalUrl: args.externalUrl ?? canonicalUrl(probe.info) ?? args.url
+      }),
+      egress: tier,
+      // Named in the result for the same reason `egress` is: a geo-targeted exit
+      // is a deliberate, billable choice and must not be invisible afterwards.
+      ...country ? { country } : {},
+      // Third member of the same family. The site refused the good formats and
+      // the download quietly took the one it would release, which is the right
+      // call — but a clip that is 360p because of that must say so here, or the
+      // only way to find out is to watch it.
+      ...reducedQuality ? { reducedQuality: true } : {}
+    };
+  } finally {
+    await rm(workDir, { recursive: true, force: true }).catch(() => {
+    });
+  }
+}
+async function uploadToAssetStore(filePath, sizeBytes) {
+  const sha256 = await hashFile(filePath);
+  const { putUrl, publicUrl } = await apiPost("/api/canvas/assets/presign", {
+    sha256,
+    mime: "video/mp4",
+    purpose: "source"
+  });
+  const putRes = await putFileStreamed({ putUrl, filePath, contentType: "video/mp4", sizeBytes });
+  if (!putRes.ok) {
+    throw new ApiError("INTERNAL_ERROR", `Storing the downloaded video failed: HTTP ${putRes.status}`);
+  }
+  return publicUrl;
+}
+
+export {
+  ingestCommand,
+  runVideoIngest,
+  ingestHints
+};
+//# sourceMappingURL=chunk-LKM3FMYC.js.map

package/dist/chunk-LOTQBFUP.js

@@ -0,0 +1,106 @@
+import {
+  cacheGet,
+  cacheSet,
+  handleConnectionError,
+  needsConnectionFix,
+  writeAdsJson
+} from "./chunk-AAF7HL4I.js";
+import {
+  ApiError,
+  apiGet
+} from "./chunk-7ANWFQEA.js";
+import {
+  getEnv
+} from "./chunk-H5CGVMQW.js";
+
+// src/commands/ads/x/output.ts
+var ACCOUNTS_CACHE_TTL_MS = 60 * 60 * 1e3;
+var NO_X_ACCOUNTS_EXPLANATION = "The X Ads connection is live, but X lists no ads accounts for it. That is usually an account X has not activated for advertising yet \u2014 most often one with no billing set up. It does not mean the connection failed, and reconnecting will not change it. An account X omits here still works everywhere else: take its id from the ads.x.com URL (ads.x.com/manager/<id>/\u2026) and pass --account-id <id>, or set BAKER_X_ADS_ACCOUNT_ID.";
+async function fetchXAccounts(useCache) {
+  if (useCache) {
+    const cached = cacheGet("x-accounts", "list");
+    if (cached) return cached.data;
+  }
+  const params = !useCache ? { "skip-cache": "true" } : void 0;
+  const data = await apiGet("/api/ads/x/accounts", params);
+  if (useCache) {
+    cacheSet("x-accounts", "list", data, ACCOUNTS_CACHE_TTL_MS);
+  }
+  return data;
+}
+async function resolveXAccountId(args) {
+  const fromArgs = args["account-id"];
+  const accountId = fromArgs || getEnv().BAKER_X_ADS_ACCOUNT_ID;
+  if (accountId) {
+    if (!/^[a-z0-9]+$/.test(accountId)) {
+      writeAdsJson({
+        ok: false,
+        error: {
+          code: "INVALID_ACCOUNT_ID",
+          message: "X Ads account ID must be a base36 string (lowercase letters and digits). Pass --account-id or set BAKER_X_ADS_ACCOUNT_ID."
+        }
+      });
+      process.exit(1);
+    }
+    return accountId;
+  }
+  const useCache = !args["no-cache"];
+  try {
+    const accounts = await fetchXAccounts(useCache);
+    const [single] = accounts;
+    if (accounts.length === 1 && single) {
+      process.stderr.write(`Using account "${single.name}" (${single.id})
+`);
+      return single.id;
+    }
+    if (accounts.length === 0) {
+      writeAdsJson({
+        ok: false,
+        error: {
+          code: "INVALID_ACCOUNT_ID",
+          message: NO_X_ACCOUNTS_EXPLANATION,
+          fix: { action: "reject", explanation: NO_X_ACCOUNTS_EXPLANATION },
+          retryable: false
+        }
+      });
+      process.exit(1);
+    }
+    const list = accounts.map((a) => `  ${a.id}  ${a.name}`).join("\n");
+    writeAdsJson({
+      ok: false,
+      error: {
+        code: "MULTIPLE_ACCOUNTS",
+        message: `Multiple X Ads accounts found. Pass --account-id or set BAKER_X_ADS_ACCOUNT_ID:
+${list}`
+      }
+    });
+    process.exit(1);
+  } catch (err) {
+    if (err instanceof ApiError && needsConnectionFix(err.code, err.message)) {
+      handleConnectionError("x_ads", err.message);
+    }
+    writeAdsJson({
+      ok: false,
+      error: {
+        code: "INVALID_ACCOUNT_ID",
+        message: "Could not auto-detect account. Pass --account-id or set BAKER_X_ADS_ACCOUNT_ID."
+      }
+    });
+    process.exit(1);
+  }
+}
+async function resolveXAccountTimezone(accountId, args) {
+  try {
+    const accounts = await fetchXAccounts(!args["no-cache"]);
+    return accounts.find((a) => a.id === accountId)?.timezone || "UTC";
+  } catch {
+    return "UTC";
+  }
+}
+
+export {
+  NO_X_ACCOUNTS_EXPLANATION,
+  resolveXAccountId,
+  resolveXAccountTimezone
+};
+//# sourceMappingURL=chunk-LOTQBFUP.js.map

package/dist/chunk-LPBLKGH7.js

@@ -0,0 +1,106 @@
+import {
+  apiGet
+} from "./chunk-7ANWFQEA.js";
+
+// src/commands/testimonials/emptyCorpusHints.ts
+function resolveEmptyReason({
+  resultCount,
+  activeFilters,
+  sourceCount
+}) {
+  if (resultCount > 0) {
+    return null;
+  }
+  if (activeFilters.length > 0) {
+    return "filtered_out";
+  }
+  return sourceCount === 0 ? "no_corpus_connected" : "corpus_empty";
+}
+var NO_CORPUS_EXIT = "Testimonials are ingested from the client's public review pages \u2014 an agent cannot write one \u2014 so no rewording will find quotes. If the user named the client's own review page (Google Maps, Trustpilot, Yelp, Tripadvisor, G2, Capterra, Amazon, App Store, Google Play), add it with `baker testimonials sources add <link>` \u2014 its reviews arrive over the next minutes, not in this turn. Stop searching and source proof another way: a result or case study the client already published, a named-customer logo (`baker images logo`), or a founder claim attributed as a claim. Otherwise ship without a testimonial block and file the gap: `baker actions create --name \"Collect customer quotes for <page>\" --priority medium --tags <slug>`. Never invent, paraphrase-into-existence, or pad with unrelated praise.";
+function emptyReasonDetail(reason, sourceCount, activeFilters) {
+  if (reason === "filtered_out") {
+    return `No rows matched, and ${activeFilters.join(", ")} narrowed this read. The corpus was not checked \u2014 drop every filter to see whether the library itself is empty.`;
+  }
+  if (reason === "no_corpus_connected") {
+    return "This company has no review source connected \u2014 checked against its source list, not inferred from the empty result.";
+  }
+  return `${sourceCount} review source${sourceCount === 1 ? " is" : "s are"} connected but hold no testimonials yet \u2014 the source is there, the rows are not. Ingestion may still be running.`;
+}
+function morePagesHint(resultCount, cursor) {
+  const found = resultCount === 0 ? "This page held no matches" : `This page held ${resultCount}`;
+  return `${found}, and the library continues past it \u2014 so this is one page, not the client's whole set. Do not report it as the corpus and do not conclude anything is missing from it. Read the next page with \`baker testimonials list --cursor ${cursor}\` (same filters), or go straight at the rows you want with \`baker testimonials search <query>\`.`;
+}
+function endOfWalkHint(command) {
+  const noun = command === "search" ? "results" : "rows";
+  return `That was the end of the library \u2014 no more ${noun} past the page before this one. Nothing here says the corpus is empty (the earlier pages held it), so answer from what those pages returned rather than re-reading.`;
+}
+async function measureEmptyCorpus({
+  command,
+  resultCount,
+  activeFilters,
+  nextCursor = null,
+  continued = false
+}) {
+  if (nextCursor) {
+    return { hints: [morePagesHint(resultCount, nextCursor)] };
+  }
+  if (resultCount > 0) {
+    return { hints: [] };
+  }
+  if (continued) {
+    return { hints: [endOfWalkHint(command)] };
+  }
+  if (activeFilters.length > 0) {
+    const reason2 = "filtered_out";
+    return {
+      hints: emptyCorpusHints({ command, reason: reason2, sourceCount: 0, activeFilters }),
+      empty: { reason: reason2, detail: emptyReasonDetail(reason2, 0, activeFilters) }
+    };
+  }
+  let sourceCount;
+  try {
+    const sources = await apiGet("/api/testimonials/sources");
+    sourceCount = Array.isArray(sources) ? sources.length : 0;
+  } catch {
+    return {
+      hints: [
+        "No rows, and the check for connected review sources failed \u2014 so this is NOT evidence that the client has no reviews. Do not record an absence you could not verify. Re-run once; if it fails again, say the library could not be read rather than that it is empty."
+      ]
+    };
+  }
+  const reason = resolveEmptyReason({ resultCount, activeFilters, sourceCount });
+  if (reason === null) {
+    return { hints: [] };
+  }
+  return {
+    hints: emptyCorpusHints({ command, reason, sourceCount, activeFilters }),
+    empty: { reason, detail: emptyReasonDetail(reason, sourceCount, activeFilters) }
+  };
+}
+function emptyCorpusHints({
+  command,
+  reason,
+  sourceCount,
+  activeFilters
+}) {
+  if (reason === null) {
+    return [];
+  }
+  if (reason === "filtered_out") {
+    return [
+      `No rows \u2014 ${activeFilters.join(", ")} narrowed this. Drop every filter and re-run once before touching the query; a filter combination and an empty library look identical from here.`
+    ];
+  }
+  const lead = command === "search" ? "No hits" : "No rows";
+  if (reason === "no_corpus_connected") {
+    return [`${lead}, and this company's source list is empty: there is no review corpus to search. ${NO_CORPUS_EXIT}`];
+  }
+  return [
+    `${lead}. ${sourceCount} review source${sourceCount === 1 ? " is" : "s are"} connected but no testimonials have landed yet \u2014 so this is a timing or ingestion gap, not a missing connection, and re-querying will not change it. Ingestion may still be running: say so rather than reporting the client has no reviews. ${NO_CORPUS_EXIT}`
+  ];
+}
+
+export {
+  measureEmptyCorpus
+};
+//# sourceMappingURL=chunk-LPBLKGH7.js.map