@koda-sl/baker-cli 0.543.0 → 0.545.0-dev.2a6beac48

package/dist/chunk-Y2AKJYI4.js

@@ -0,0 +1,191 @@
+import {
+  hintLines,
+  splitHints
+} from "./chunk-D5FY4FAK.js";
+import {
+  isNotConnectedError,
+  writeJsonEnvelope
+} from "./chunk-MGWAZXY6.js";
+import {
+  ApiError,
+  apiPost
+} from "./chunk-B2OHCWYB.js";
+import {
+  requireChatId,
+  resolveChatId
+} from "./chunk-E54435XT.js";
+
+// src/commands/ga4/shared.ts
+import { readFileSync } from "fs";
+function failValidation(message) {
+  writeJsonEnvelope({ ok: false, error: { code: "VALIDATION_ERROR", message } });
+  process.exit(1);
+}
+function requireTarget(args, entity) {
+  const positional = Array.isArray(args._) ? args._[0] : void 0;
+  const target = args.id ?? args.target ?? positional;
+  if (typeof target !== "string" || target.length === 0) {
+    failValidation(`pass the ${entity} id as the positional argument`);
+  }
+  return target;
+}
+function readJsonSource(args, flag) {
+  const inline = args[flag];
+  const file = args.file;
+  const raw = typeof file === "string" && file.length > 0 ? readFileSync(file, "utf8") : typeof inline === "string" && inline.length > 0 ? inline : void 0;
+  if (raw === void 0) {
+    failValidation(`pass --${flag} with inline JSON or --file with a path to a JSON file`);
+  }
+  try {
+    return JSON.parse(raw);
+  } catch {
+    return failValidation(`--${flag} is not valid JSON`);
+  }
+}
+function asObject(value, flag, index) {
+  if (!value || typeof value !== "object" || Array.isArray(value)) {
+    failValidation(
+      index === void 0 ? `--${flag} must be a JSON object` : `--${flag}[${index}] must be a JSON object`
+    );
+  }
+  return value;
+}
+function loadJsonArg(args, flag = "json") {
+  return asObject(readJsonSource(args, flag), flag);
+}
+function loadJsonPayloads(args) {
+  const parsed = readJsonSource(args, "json");
+  if (!Array.isArray(parsed)) {
+    return [asObject(parsed, "json")];
+  }
+  if (parsed.length === 0) {
+    failValidation("--json is an empty list \u2014 pass at least one definition");
+  }
+  return parsed.map((entry, index) => asObject(entry, "json", index));
+}
+var RETRYABLE_CODES = /* @__PURE__ */ new Set(["RATE_LIMITED", "UNAVAILABLE", "INTERNAL_ERROR", "NETWORK_ERROR", "TIMEOUT"]);
+function handleError(err) {
+  if (err instanceof ApiError) {
+    const readOnly = err.code === "FORBIDDEN" && /read the property but not change it/i.test(err.message);
+    writeJsonEnvelope({
+      ok: false,
+      error: {
+        code: err.code,
+        message: err.message,
+        ...readOnly ? {
+          fix: {
+            action: "reconnect",
+            explanation: "The Google Analytics connection is read-only. Ask the user to reconnect Google Analytics \u2014 in the dashboard, Brain \u2192 Integrations \u2192 Platforms \u2192 Google Analytics \u2014 which grants the permission to change the property. Reporting (`baker ga4 query`, `baker ga4 audit`) keeps working meanwhile, so carry on with the rest of the job and tell them what you will set up once they have reconnected. Don't claim anything was configured."
+          }
+        } : err.code === "UNAUTHORIZED" || err.code === "FORBIDDEN" && isNotConnectedError(err.code, err.message) ? {
+          fix: {
+            action: "request_connection",
+            explanation: 'Google Analytics is not connected for this company yet, or no property has been picked. Don\'t send the user to Settings \u2014 call the `request_connection` tool with { platform: "google-analytics", reason } and they connect and pick their property from inside the chat. If the `request_connection` call fails, say so plainly and give the exact route: dashboard \u2192 Brain \u2192 Integrations \u2192 Platforms \u2192 Google Analytics. Then carry on with the rest of the task.'
+          }
+        } : {},
+        retryable: RETRYABLE_CODES.has(err.code)
+      }
+    });
+    process.exit(1);
+  }
+  writeJsonEnvelope({ ok: false, error: { code: "NETWORK_ERROR", message: "Unexpected error", retryable: true } });
+  process.exit(1);
+}
+var STAGE_HINTS = [
+  "Staged only \u2014 nothing has changed in Google Analytics yet. These apply when the chat completes, and an applied change takes effect on the property immediately: there is no version to publish afterwards and no undo. Say the changes are ready and take effect when the user completes the chat \u2014 not that they are done already."
+];
+var WARNING_HINT = "Read the warnings above before telling the user this is ready.";
+async function stageOp(op) {
+  const chatId = requireChatId();
+  try {
+    const { data, hints } = splitHints(await apiPost("/api/ga4/draft/stage", { chatId, op }));
+    writeJsonEnvelope({
+      ok: true,
+      data,
+      hints: [...hints, ...STAGE_HINTS, ...data.warnings.length > 0 ? [WARNING_HINT] : []]
+    });
+  } catch (err) {
+    handleError(err);
+  }
+}
+async function stageOps(ops) {
+  const [only] = ops;
+  if (ops.length === 1 && only) {
+    await stageOp(only);
+    return;
+  }
+  const chatId = requireChatId();
+  try {
+    const { data, hints } = splitHints(
+      await apiPost("/api/ga4/draft/stage-batch", { chatId, ops })
+    );
+    writeJsonEnvelope({
+      ok: true,
+      data,
+      hints: [...hints, ...STAGE_HINTS, ...data.ops.some((op) => op.warnings.length > 0) ? [WARNING_HINT] : []]
+    });
+  } catch (err) {
+    handleError(err);
+  }
+}
+async function draftAction(path, body, chat) {
+  const chatId = resolveChatId(chat);
+  try {
+    const { data, hints } = splitHints(await apiPost(path, { chatId, ...body }));
+    writeJsonEnvelope({ ok: true, data, ...hints.length > 0 ? { hints } : {} });
+    return data;
+  } catch (err) {
+    handleError(err);
+  }
+}
+function renderDraft(response) {
+  if (response.count === 0) {
+    return "No Google Analytics changes staged on this chat.";
+  }
+  const lines = [
+    response.status === "discarded" ? `${response.count} Google Analytics change(s) discarded with their chat \u2014 none will apply` : `${response.count} staged Google Analytics change(s)`,
+    ...response.ops.map((op) => {
+      const status = op.result ? ` [${op.result.status}]` : "";
+      return `  ${op.ref}  ${op.summary}${status}`;
+    })
+  ];
+  if (response.status === "applied") {
+    const failed = response.ops.filter((op) => op.result?.status !== "applied");
+    lines.push(
+      "",
+      failed.length === 0 ? "All applied \u2014 these are in effect on the property now." : `${failed.length} of ${response.count} did not apply. The rest are in effect on the property.`
+    );
+  }
+  if (response.publishError) {
+    lines.push("", response.publishError);
+  }
+  return lines.join("\n");
+}
+async function draftList(json, chat) {
+  const chatId = resolveChatId(chat);
+  try {
+    const response = await apiPost("/api/ga4/draft", { chatId });
+    const { data, hints } = splitHints(response);
+    if (json) {
+      writeJsonEnvelope({ ok: true, data, ...hints.length > 0 ? { hints } : {} });
+      return;
+    }
+    process.stdout.write(`${renderDraft(data)}${hintLines(response)}
+`);
+  } catch (err) {
+    handleError(err);
+  }
+}
+
+export {
+  failValidation,
+  requireTarget,
+  loadJsonArg,
+  loadJsonPayloads,
+  handleError,
+  stageOp,
+  stageOps,
+  draftAction,
+  draftList
+};
+//# sourceMappingURL=chunk-Y2AKJYI4.js.map

package/dist/chunk-YMLJ2H5C.js

@@ -0,0 +1,74 @@
+import {
+  cacheGet,
+  cacheSet,
+  handleConnectionError,
+  needsConnectionFix,
+  writeJsonEnvelope
+} from "./chunk-MGWAZXY6.js";
+import {
+  ApiError,
+  apiGet
+} from "./chunk-B2OHCWYB.js";
+import {
+  getEnv
+} from "./chunk-E54435XT.js";
+
+// src/commands/ga4/resolve.ts
+async function fetchProperties(useCache = true) {
+  if (useCache) {
+    const cached = cacheGet("ga4-properties", "list");
+    if (cached) return cached.data;
+  }
+  const params = !useCache ? { "skip-cache": "true" } : void 0;
+  const data = await apiGet("/api/ga4/properties", params);
+  if (useCache) {
+    cacheSet("ga4-properties", "list", data, 60 * 60 * 1e3);
+  }
+  return data;
+}
+async function resolvePropertyId(args) {
+  const fromArgs = args["property-id"];
+  const fromEnv = getEnv().BAKER_GA4_PROPERTY_ID;
+  const propertyId = fromArgs || fromEnv;
+  if (propertyId) return propertyId;
+  const useCache = !args["no-cache"];
+  try {
+    const properties = await fetchProperties(useCache);
+    const [single] = properties;
+    if (properties.length === 1 && single) {
+      process.stderr.write(`Using property "${single.name}" (${single.externalId})
+`);
+      return single.externalId;
+    }
+    if (properties.length === 0) {
+      handleConnectionError("ga4");
+    }
+    const list = properties.map((p) => `  ${p.externalId}  ${p.name}`).join("\n");
+    writeJsonEnvelope({
+      ok: false,
+      error: {
+        code: "MULTIPLE_PROPERTIES",
+        message: `Multiple properties found. Pass --property-id or set BAKER_GA4_PROPERTY_ID:
+${list}`
+      }
+    });
+    process.exit(1);
+  } catch (err) {
+    if (err instanceof ApiError && needsConnectionFix(err.code, err.message)) {
+      handleConnectionError("ga4", err.message);
+    }
+    writeJsonEnvelope({
+      ok: false,
+      error: {
+        code: "RESOLUTION_ERROR",
+        message: "Could not auto-detect property. Pass --property-id or set BAKER_GA4_PROPERTY_ID."
+      }
+    });
+    process.exit(1);
+  }
+}
+
+export {
+  resolvePropertyId
+};
+//# sourceMappingURL=chunk-YMLJ2H5C.js.map

package/dist/chunk-YQB66YKJ.js

@@ -0,0 +1,98 @@
+import {
+  isRemoteUrl
+} from "./chunk-GI7B47X3.js";
+import {
+  detectImageContentType,
+  uploadLocalImage
+} from "./chunk-EMCTUCCJ.js";
+import {
+  IMAGE_GENERATE_MODELS,
+  MODEL_REGISTRY,
+  VIDEO_GENERATE_MODELS
+} from "./chunk-F3CZFUXZ.js";
+import {
+  ApiError,
+  apiPost
+} from "./chunk-B2OHCWYB.js";
+
+// src/commands/studio/modelOptions.ts
+function unionOfStringEnum(node, models, param) {
+  const values = /* @__PURE__ */ new Set();
+  for (const model of models) {
+    const schema = MODEL_REGISTRY[node][model]?.params[param];
+    if (schema?.kind === "string") for (const value of schema.enum ?? []) values.add(value);
+  }
+  return [...values];
+}
+function unionOfNumberEnum(node, models, param) {
+  const values = /* @__PURE__ */ new Set();
+  for (const model of models) {
+    const schema = MODEL_REGISTRY[node][model]?.params[param];
+    if (schema?.kind === "number") for (const value of schema.enum ?? []) values.add(value);
+  }
+  return [...values].sort((a, b) => a - b);
+}
+var IMAGE_MODEL_IDS = [...IMAGE_GENERATE_MODELS];
+var VIDEO_MODEL_IDS = [...VIDEO_GENERATE_MODELS];
+var IMAGE_SIZES = unionOfStringEnum("image_generate", IMAGE_GENERATE_MODELS, "image_size");
+var IMAGE_QUALITIES = unionOfStringEnum("image_generate", IMAGE_GENERATE_MODELS, "quality");
+var CLIP_RESOLUTIONS = unionOfStringEnum("video_generate", VIDEO_GENERATE_MODELS, "resolution");
+var CLIP_DURATIONS = unionOfNumberEnum("video_generate", VIDEO_GENERATE_MODELS, "duration");
+var CLIP_PERSON_GENERATION = unionOfStringEnum("video_generate", VIDEO_GENERATE_MODELS, "person_generation");
+
+// src/commands/studio/references.ts
+var LIBRARY_ID_RE = /^[a-z0-9]{16,}$/i;
+var FILE_EXTENSION_RE = /\.[a-z0-9]{2,5}$/i;
+var PATH_HINT_RE = /[/\\]/;
+function classify(entry) {
+  if (isRemoteUrl(entry)) return { kind: "url", value: entry };
+  if (PATH_HINT_RE.test(entry) || FILE_EXTENSION_RE.test(entry)) return { kind: "path", value: entry };
+  if (LIBRARY_ID_RE.test(entry)) return { kind: "id", value: entry };
+  throw new ApiError(
+    "VALIDATION_ERROR",
+    `"${entry}" is not a library image id, a file path, or an image URL. Pass a comma-separated mix of those.`
+  );
+}
+function parseImageRefs(spec) {
+  return spec.split(",").map((entry) => entry.trim()).filter(Boolean).map(classify);
+}
+var defaultDeps = {
+  ingest: (url) => apiPost("/api/images/ingest", { url, source: "uploaded" }),
+  upload: (path) => uploadLocalImage({ file: path, contentType: detectImageContentType(path), source: "uploaded" })
+};
+async function resolveLibraryImageIds(spec, limit, deps = defaultDeps) {
+  const refs = parseImageRefs(spec);
+  if (limit && refs.length > limit.max) {
+    throw new ApiError("VALIDATION_ERROR", limit.refuse(refs.length));
+  }
+  const imageIds = [];
+  const added = [];
+  for (const ref of refs) {
+    if (ref.kind === "id") {
+      imageIds.push(ref.value);
+      continue;
+    }
+    const { imageId } = ref.kind === "url" ? await deps.ingest(ref.value) : await deps.upload(ref.value).catch(uploadFailure(ref.value));
+    imageIds.push(imageId);
+    added.push(imageId);
+  }
+  return { imageIds, added };
+}
+function uploadFailure(path) {
+  return (error) => {
+    if (error instanceof ApiError) throw error;
+    throw new ApiError("VALIDATION_ERROR", `Could not read "${path}" as an image.`);
+  };
+}
+
+export {
+  IMAGE_MODEL_IDS,
+  VIDEO_MODEL_IDS,
+  IMAGE_SIZES,
+  IMAGE_QUALITIES,
+  CLIP_RESOLUTIONS,
+  CLIP_DURATIONS,
+  CLIP_PERSON_GENERATION,
+  resolveLibraryImageIds
+};
+//# sourceMappingURL=chunk-YQB66YKJ.js.map

package/dist/chunk-YU55LFZZ.js

@@ -0,0 +1,56 @@
+import {
+  ApiError,
+  apiGet
+} from "./chunk-B2OHCWYB.js";
+import {
+  compactImage,
+  compactVideo,
+  writeJson
+} from "./chunk-WCIS4FZH.js";
+
+// src/commands/mediaGroup.ts
+async function runMediaGroupLookup(input) {
+  if (!input.id && !input.groupKey) {
+    writeJson({
+      ok: false,
+      error: { code: "VALIDATION_ERROR", message: `Pass ${input.idLabel} or --group-key` }
+    });
+    process.exit(1);
+  }
+  try {
+    const query = new URLSearchParams();
+    if (input.id) query.set("id", input.id);
+    if (input.groupKey) query.set("groupKey", input.groupKey);
+    const data = await apiGet(`${input.path}?${query.toString()}`);
+    const images = data.images.map((image) => compactImage(image));
+    const videos = (data.videos ?? []).map((video) => compactVideo(video));
+    const count = images.length + videos.length;
+    writeJson({
+      ok: true,
+      data: {
+        groupKey: data.groupKey,
+        groupKind: data.groupKind ?? null,
+        // The caption or page title the whole set shares — often the only
+        // place the campaign is named.
+        groupTitle: data.groupTitle ?? null,
+        groupUrl: data.groupUrl ?? null,
+        count,
+        images,
+        videos
+      },
+      hints: count > 1 ? ["These arrived together and are ordered. Read them in sequence before using any one of them alone."] : void 0
+    });
+  } catch (err) {
+    if (err instanceof ApiError) {
+      writeJson({ ok: false, error: { code: err.code, message: err.message } });
+      process.exit(1);
+    }
+    writeJson({ ok: false, error: { code: "INTERNAL_ERROR", message: "Unexpected error" } });
+    process.exit(1);
+  }
+}
+
+export {
+  runMediaGroupLookup
+};
+//# sourceMappingURL=chunk-YU55LFZZ.js.map

package/dist/cite-G5PPVGTC.js

@@ -0,0 +1,191 @@
+import {
+  handleResearchError,
+  writeResearchJson
+} from "./chunk-W4W3F3P7.js";
+import {
+  repeatedValues
+} from "./chunk-ERSX3MKH.js";
+import {
+  apiPost
+} from "./chunk-B2OHCWYB.js";
+import "./chunk-SVUFNCOE.js";
+import "./chunk-E54435XT.js";
+import {
+  registerSchema
+} from "./chunk-WCIS4FZH.js";
+import "./chunk-YQSBHWOI.js";
+
+// src/commands/research/cite.ts
+import { defineCommand } from "citty";
+var TIMEOUT_MS = 18e4;
+var STALE_AFTER_DAYS = 14;
+registerSchema({
+  command: "research.cite",
+  description: "Prove a quote-grade claim: read the page, check each span is on it word for word, and record the evidence (url, time, page hash, the span in context). Not-found spans come back found:false \u2014 drop the quotation marks or label the line `Synthesis:`. `--check <evidenceId>` re-reads the page before you reuse a fact.",
+  args: {
+    url: { type: "string", description: "The page that says it", required: false },
+    span: {
+      type: "string",
+      description: "Exact words from the page (8-1000 chars). Repeat --span for several, up to 10",
+      required: false
+    },
+    check: {
+      type: "string",
+      description: "Re-read a cited page by its evidenceId and say whether the spans are still there, and how old the citation is",
+      required: false
+    },
+    full: {
+      type: "boolean",
+      description: "Also return the final URL, page title, content hash and the page's own spelling of each span",
+      required: false,
+      default: false
+    }
+  }
+});
+var FIELDS = {
+  evidenceId: "Id of the recorded evidence \u2014 put it next to the claim, and pass it to --check before reusing the fact",
+  url: "The page as asked for",
+  capturedAt: "When the page was read and the spans checked",
+  spans: "Each span: found (word for word on the page) and, when found, the page's words around it"
+};
+function quoted(text) {
+  return `"${text.length > 80 ? `${text.slice(0, 77)}\u2026` : text}"`;
+}
+function citeHints(result) {
+  const missing = result.spans.filter((span) => !span.found);
+  const hints = [];
+  if (missing.length > 0) {
+    hints.push(
+      `Not on the page word for word: ${missing.map((span) => quoted(span.text)).join(", ")}. Do not present these as quotes: drop the quotation marks and paraphrase with the link, or label the line \`Synthesis:\`. Never attribute them to a person. Carry on with the rest of the job.`
+    );
+    if (result.truncated) {
+      hints.push(
+        `The page was cut at ${result.contentChars} characters, so a span further down reads as not found. Cite a deeper URL (the section's own page) if there is one.`
+      );
+    }
+  }
+  if (missing.length < result.spans.length) {
+    hints.push(
+      `Write the found spans as quotes with the link and the capture date, and keep evidenceId ${result.evidenceId} beside the claim. Before reusing the fact in a later deliverable, run \`baker research cite --check ${result.evidenceId}\`.`
+    );
+  }
+  return hints;
+}
+function citeCheckHints(result) {
+  const hints = [];
+  const gone = result.spans.filter((span) => span.wasFound && !span.found);
+  if (gone.length > 0 && result.truncated) {
+    hints.push(
+      `Not found this time: ${gone.map((span) => quoted(span.text)).join(", ")} \u2014 but the page was cut before its end, so they may sit past the cut rather than be gone. Re-cite them on the section's own URL before reusing them, or state them as "as of ${result.capturedAt.slice(0, 10)}".`
+    );
+  } else if (gone.length > 0) {
+    hints.push(
+      `No longer on the page: ${gone.map((span) => quoted(span.text)).join(", ")}. The page changed since ${result.capturedAt.slice(0, 10)} \u2014 do not reuse these as current facts. Re-research them, or state them as "as of ${result.capturedAt.slice(0, 10)}".`
+    );
+  } else if (result.ageDays >= STALE_AFTER_DAYS) {
+    hints.push(
+      `Still on the page, but first cited ${result.ageDays} days ago. Prices, plans and ownership move: cite again (\`baker research cite\`) so the deliverable carries today's date.`
+    );
+  }
+  return hints;
+}
+function citeView(result, full) {
+  if (full) return result;
+  return {
+    evidenceId: result.evidenceId,
+    url: result.url,
+    capturedAt: result.capturedAt,
+    spans: result.spans.map(({ text, found, context }) => context ? { text, found, context } : { text, found })
+  };
+}
+function checkView(result, full) {
+  if (full) return result;
+  return {
+    evidenceId: result.evidenceId,
+    url: result.url,
+    capturedAt: result.capturedAt,
+    checkedAt: result.checkedAt,
+    ageDays: result.ageDays,
+    pageChanged: result.pageChanged,
+    ...result.truncated ? { truncated: true } : {},
+    spans: result.spans.map(({ text, found, wasFound }) => ({ text, found, wasFound }))
+  };
+}
+function citeFix(code) {
+  if (code === "VALIDATION_ERROR" || code === "BAD_REQUEST") {
+    return {
+      action: "retry_with_changes",
+      explanation: 'Pass the full page address (https://\u2026) and at least one exact span of 8+ characters: baker research cite "<url>" "<exact words>" --span "<more exact words>".'
+    };
+  }
+  if (code === "NOT_FOUND") {
+    return {
+      action: "continue_without",
+      explanation: "The page or the evidence id does not exist. Do not quote from it: drop the quotation marks or label the line `Synthesis:`, and carry on."
+    };
+  }
+  return {
+    action: "continue_without",
+    explanation: "The page could not be read, so nothing on it is verified. Do not present the claim as a quote: drop the quotation marks or label the line `Synthesis:`, name the page as a gap, and finish the rest of the job."
+  };
+}
+function failValidation(message) {
+  writeResearchJson({ ok: false, error: { code: "VALIDATION_ERROR", message, fix: citeFix("VALIDATION_ERROR") } });
+  process.exit(1);
+}
+var citeCommand = defineCommand({
+  meta: {
+    name: "cite",
+    description: `Prove a quote-grade claim before it goes into Knowledge, a landing, an ad or a report.
+
+Start here: pass the page and the exact words you want to quote. The page is read fresh, each span is checked word for word (case, spacing, curly quotes and dashes forgiven \u2014 nothing else), and the evidence is recorded: when, where, a hash of the page and the words around each span.
+
+A span that is not on the page comes back found:false. Do not quote it: drop the quotation marks and paraphrase with the link, or label it \`Synthesis:\`.
+
+Before reusing a cited fact later, re-check it: \`--check <evidenceId>\` re-reads the page and says whether the spans are still there and how old the citation is.
+
+Examples:
+  baker research cite "https://acme.com/pricing" "Plans start at $10 per month"
+  baker research cite "https://acme.com/about" --span "founded in 2014" --span "offices in Berlin and Madrid"
+  baker research cite --check jd7f0m3k2x9y`
+  },
+  args: {
+    url: { type: "positional", description: "The page that says it", required: false },
+    span: { type: "string", description: "Exact words from the page; repeat for several", required: false },
+    check: { type: "string", description: "Re-check an evidenceId", required: false },
+    full: { type: "boolean", description: "Return every field", required: false }
+  },
+  run: async ({ args, rawArgs }) => {
+    const full = Boolean(args.full);
+    const check = typeof args.check === "string" ? args.check.trim() : "";
+    try {
+      if (check) {
+        const body2 = { evidenceId: check };
+        const result2 = await apiPost("/api/research/cite/check", body2, {
+          timeoutMs: TIMEOUT_MS
+        });
+        writeResearchJson({ ok: true, data: checkView(result2, full), hints: citeCheckHints(result2) });
+        return;
+      }
+      const positionals = (Array.isArray(args._) ? args._ : []).filter(
+        (word) => typeof word === "string" && word.trim().length > 0
+      );
+      const [url, ...positionalSpans] = positionals;
+      if (!url) failValidation("A page URL is required (or --check <evidenceId>).");
+      const spans = [...positionalSpans, ...repeatedValues(rawArgs, "span", args.span)];
+      if (spans.length === 0) failValidation("At least one exact span is required.");
+      const body = { url, spans };
+      const result = await apiPost("/api/research/cite", body, { timeoutMs: TIMEOUT_MS });
+      writeResearchJson({ ok: true, data: citeView(result, full), fields: FIELDS, hints: citeHints(result) });
+    } catch (err) {
+      handleResearchError(err, citeFix);
+    }
+  }
+});
+export {
+  STALE_AFTER_DAYS,
+  citeCheckHints,
+  citeCommand,
+  citeHints
+};
+//# sourceMappingURL=cite-G5PPVGTC.js.map

package/dist/cite-G5PPVGTC.js.map

@@ -0,0 +1 @@
+{"version":3,"sources":["../src/commands/research/cite.ts"],"sourcesContent":["import type {\n  ResearchCiteCheckRequest,\n  ResearchCiteCheckResponse,\n  ResearchCiteRequest,\n  ResearchCiteResponse,\n} from \"@baker/api/research-cite\";\nimport { defineCommand } from \"citty\";\nimport { apiPost } from \"../../client.ts\";\nimport { repeatedValues } from \"../../lib/repeatedFlags.ts\";\nimport { registerSchema } from \"../../schemas.ts\";\nimport { handleResearchError, type ResearchFix, writeResearchJson } from \"./output.ts\";\n\n/** A page behind a bot wall can take a while to clear, as with `research fetch`. */\nconst TIMEOUT_MS = 180_000;\n\n/** Past this, a fact about a competitor (price, plan, ownership) is worth re-reading before it ships. */\nexport const STALE_AFTER_DAYS = 14;\n\nregisterSchema({\n  command: \"research.cite\",\n  description:\n    \"Prove a quote-grade claim: read the page, check each span is on it word for word, and record the evidence (url, time, page hash, the span in context). Not-found spans come back found:false — drop the quotation marks or label the line `Synthesis:`. `--check <evidenceId>` re-reads the page before you reuse a fact.\",\n  args: {\n    url: { type: \"string\", description: \"The page that says it\", required: false },\n    span: {\n      type: \"string\",\n      description: \"Exact words from the page (8-1000 chars). Repeat --span for several, up to 10\",\n      required: false,\n    },\n    check: {\n      type: \"string\",\n      description:\n        \"Re-read a cited page by its evidenceId and say whether the spans are still there, and how old the citation is\",\n      required: false,\n    },\n    full: {\n      type: \"boolean\",\n      description: \"Also return the final URL, page title, content hash and the page's own spelling of each span\",\n      required: false,\n      default: false,\n    },\n  },\n});\n\nconst FIELDS: Record<string, string> = {\n  evidenceId: \"Id of the recorded evidence — put it next to the claim, and pass it to --check before reusing the fact\",\n  url: \"The page as asked for\",\n  capturedAt: \"When the page was read and the spans checked\",\n  spans: \"Each span: found (word for word on the page) and, when found, the page's words around it\",\n};\n\nfunction quoted(text: string): string {\n  return `\"${text.length > 80 ? `${text.slice(0, 77)}…` : text}\"`;\n}\n\n/** The nudge after a cite: what to do with a span the page does not carry, and how to reuse one it does. */\nexport function citeHints(result: ResearchCiteResponse): string[] {\n  const missing = result.spans.filter((span) => !span.found);\n  const hints: string[] = [];\n  if (missing.length > 0) {\n    hints.push(\n      `Not on the page word for word: ${missing.map((span) => quoted(span.text)).join(\", \")}. ` +\n        \"Do not present these as quotes: drop the quotation marks and paraphrase with the link, or label the line `Synthesis:`. \" +\n        \"Never attribute them to a person. Carry on with the rest of the job.\",\n    );\n    if (result.truncated) {\n      hints.push(\n        `The page was cut at ${result.contentChars} characters, so a span further down reads as not found. Cite a deeper URL (the section's own page) if there is one.`,\n      );\n    }\n  }\n  if (missing.length < result.spans.length) {\n    hints.push(\n      `Write the found spans as quotes with the link and the capture date, and keep evidenceId ${result.evidenceId} beside the claim. Before reusing the fact in a later deliverable, run \\`baker research cite --check ${result.evidenceId}\\`.`,\n    );\n  }\n  return hints;\n}\n\n/** The nudge after a re-check: which spans went away and whether the citation is old. */\nexport function citeCheckHints(result: ResearchCiteCheckResponse): string[] {\n  const hints: string[] = [];\n  const gone = result.spans.filter((span) => span.wasFound && !span.found);\n  if (gone.length > 0 && result.truncated) {\n    hints.push(\n      `Not found this time: ${gone.map((span) => quoted(span.text)).join(\", \")} — but the page was cut before its end, so they may sit past the cut rather than be gone. Re-cite them on the section's own URL before reusing them, or state them as \"as of ${result.capturedAt.slice(0, 10)}\".`,\n    );\n  } else if (gone.length > 0) {\n    hints.push(\n      `No longer on the page: ${gone.map((span) => quoted(span.text)).join(\", \")}. The page changed since ${result.capturedAt.slice(0, 10)} — do not reuse these as current facts. Re-research them, or state them as \"as of ${result.capturedAt.slice(0, 10)}\".`,\n    );\n  } else if (result.ageDays >= STALE_AFTER_DAYS) {\n    hints.push(\n      `Still on the page, but first cited ${result.ageDays} days ago. Prices, plans and ownership move: cite again (\\`baker research cite\\`) so the deliverable carries today's date.`,\n    );\n  }\n  return hints;\n}\n\nfunction citeView(result: ResearchCiteResponse, full: boolean) {\n  if (full) return result;\n  return {\n    evidenceId: result.evidenceId,\n    url: result.url,\n    capturedAt: result.capturedAt,\n    spans: result.spans.map(({ text, found, context }) => (context ? { text, found, context } : { text, found })),\n  };\n}\n\nfunction checkView(result: ResearchCiteCheckResponse, full: boolean) {\n  if (full) return result;\n  return {\n    evidenceId: result.evidenceId,\n    url: result.url,\n    capturedAt: result.capturedAt,\n    checkedAt: result.checkedAt,\n    ageDays: result.ageDays,\n    pageChanged: result.pageChanged,\n    ...(result.truncated ? { truncated: true } : {}),\n    spans: result.spans.map(({ text, found, wasFound }) => ({ text, found, wasFound })),\n  };\n}\n\n/** A failed read is one source of a larger job: name the gap and keep going. */\nfunction citeFix(code: string): ResearchFix {\n  if (code === \"VALIDATION_ERROR\" || code === \"BAD_REQUEST\") {\n    return {\n      action: \"retry_with_changes\",\n      explanation:\n        'Pass the full page address (https://…) and at least one exact span of 8+ characters: baker research cite \"<url>\" \"<exact words>\" --span \"<more exact words>\".',\n    };\n  }\n  if (code === \"NOT_FOUND\") {\n    return {\n      action: \"continue_without\",\n      explanation:\n        \"The page or the evidence id does not exist. Do not quote from it: drop the quotation marks or label the line `Synthesis:`, and carry on.\",\n    };\n  }\n  return {\n    action: \"continue_without\",\n    explanation:\n      \"The page could not be read, so nothing on it is verified. Do not present the claim as a quote: drop the quotation marks or label the line `Synthesis:`, name the page as a gap, and finish the rest of the job.\",\n  };\n}\n\nfunction failValidation(message: string): never {\n  writeResearchJson({ ok: false, error: { code: \"VALIDATION_ERROR\", message, fix: citeFix(\"VALIDATION_ERROR\") } });\n  process.exit(1);\n}\n\nexport const citeCommand = defineCommand({\n  meta: {\n    name: \"cite\",\n    description: `Prove a quote-grade claim before it goes into Knowledge, a landing, an ad or a report.\n\nStart here: pass the page and the exact words you want to quote. The page is read fresh, each span is checked word for word (case, spacing, curly quotes and dashes forgiven — nothing else), and the evidence is recorded: when, where, a hash of the page and the words around each span.\n\nA span that is not on the page comes back found:false. Do not quote it: drop the quotation marks and paraphrase with the link, or label it \\`Synthesis:\\`.\n\nBefore reusing a cited fact later, re-check it: \\`--check <evidenceId>\\` re-reads the page and says whether the spans are still there and how old the citation is.\n\nExamples:\n  baker research cite \"https://acme.com/pricing\" \"Plans start at $10 per month\"\n  baker research cite \"https://acme.com/about\" --span \"founded in 2014\" --span \"offices in Berlin and Madrid\"\n  baker research cite --check jd7f0m3k2x9y`,\n  },\n  args: {\n    url: { type: \"positional\", description: \"The page that says it\", required: false },\n    span: { type: \"string\", description: \"Exact words from the page; repeat for several\", required: false },\n    check: { type: \"string\", description: \"Re-check an evidenceId\", required: false },\n    full: { type: \"boolean\", description: \"Return every field\", required: false },\n  },\n  run: async ({ args, rawArgs }) => {\n    const full = Boolean(args.full);\n    const check = typeof args.check === \"string\" ? args.check.trim() : \"\";\n    try {\n      if (check) {\n        const body: ResearchCiteCheckRequest = { evidenceId: check };\n        const result = await apiPost<ResearchCiteCheckResponse>(\"/api/research/cite/check\", body, {\n          timeoutMs: TIMEOUT_MS,\n        });\n        writeResearchJson({ ok: true, data: checkView(result, full), hints: citeCheckHints(result) });\n        return;\n      }\n\n      const positionals = (Array.isArray(args._) ? args._ : []).filter(\n        (word): word is string => typeof word === \"string\" && word.trim().length > 0,\n      );\n      const [url, ...positionalSpans] = positionals;\n      if (!url) failValidation(\"A page URL is required (or --check <evidenceId>).\");\n      const spans = [...positionalSpans, ...repeatedValues(rawArgs, \"span\", args.span)];\n      if (spans.length === 0) failValidation(\"At least one exact span is required.\");\n\n      const body: ResearchCiteRequest = { url, spans };\n      const result = await apiPost<ResearchCiteResponse>(\"/api/research/cite\", body, { timeoutMs: TIMEOUT_MS });\n      writeResearchJson({ ok: true, data: citeView(result, full), fields: FIELDS, hints: citeHints(result) });\n    } catch (err) {\n      handleResearchError(err, citeFix);\n    }\n  },\n});\n"],"mappings":";;;;;;;;;;;;;;;;;;AAMA,SAAS,qBAAqB;AAO9B,IAAM,aAAa;AAGZ,IAAM,mBAAmB;AAEhC,eAAe;AAAA,EACb,SAAS;AAAA,EACT,aACE;AAAA,EACF,MAAM;AAAA,IACJ,KAAK,EAAE,MAAM,UAAU,aAAa,yBAAyB,UAAU,MAAM;AAAA,IAC7E,MAAM;AAAA,MACJ,MAAM;AAAA,MACN,aAAa;AAAA,MACb,UAAU;AAAA,IACZ;AAAA,IACA,OAAO;AAAA,MACL,MAAM;AAAA,MACN,aACE;AAAA,MACF,UAAU;AAAA,IACZ;AAAA,IACA,MAAM;AAAA,MACJ,MAAM;AAAA,MACN,aAAa;AAAA,MACb,UAAU;AAAA,MACV,SAAS;AAAA,IACX;AAAA,EACF;AACF,CAAC;AAED,IAAM,SAAiC;AAAA,EACrC,YAAY;AAAA,EACZ,KAAK;AAAA,EACL,YAAY;AAAA,EACZ,OAAO;AACT;AAEA,SAAS,OAAO,MAAsB;AACpC,SAAO,IAAI,KAAK,SAAS,KAAK,GAAG,KAAK,MAAM,GAAG,EAAE,CAAC,WAAM,IAAI;AAC9D;AAGO,SAAS,UAAU,QAAwC;AAChE,QAAM,UAAU,OAAO,MAAM,OAAO,CAAC,SAAS,CAAC,KAAK,KAAK;AACzD,QAAM,QAAkB,CAAC;AACzB,MAAI,QAAQ,SAAS,GAAG;AACtB,UAAM;AAAA,MACJ,kCAAkC,QAAQ,IAAI,CAAC,SAAS,OAAO,KAAK,IAAI,CAAC,EAAE,KAAK,IAAI,CAAC;AAAA,IAGvF;AACA,QAAI,OAAO,WAAW;AACpB,YAAM;AAAA,QACJ,uBAAuB,OAAO,YAAY;AAAA,MAC5C;AAAA,IACF;AAAA,EACF;AACA,MAAI,QAAQ,SAAS,OAAO,MAAM,QAAQ;AACxC,UAAM;AAAA,MACJ,2FAA2F,OAAO,UAAU,wGAAwG,OAAO,UAAU;AAAA,IACvO;AAAA,EACF;AACA,SAAO;AACT;AAGO,SAAS,eAAe,QAA6C;AAC1E,QAAM,QAAkB,CAAC;AACzB,QAAM,OAAO,OAAO,MAAM,OAAO,CAAC,SAAS,KAAK,YAAY,CAAC,KAAK,KAAK;AACvE,MAAI,KAAK,SAAS,KAAK,OAAO,WAAW;AACvC,UAAM;AAAA,MACJ,wBAAwB,KAAK,IAAI,CAAC,SAAS,OAAO,KAAK,IAAI,CAAC,EAAE,KAAK,IAAI,CAAC,qLAAgL,OAAO,WAAW,MAAM,GAAG,EAAE,CAAC;AAAA,IACxR;AAAA,EACF,WAAW,KAAK,SAAS,GAAG;AAC1B,UAAM;AAAA,MACJ,0BAA0B,KAAK,IAAI,CAAC,SAAS,OAAO,KAAK,IAAI,CAAC,EAAE,KAAK,IAAI,CAAC,4BAA4B,OAAO,WAAW,MAAM,GAAG,EAAE,CAAC,0FAAqF,OAAO,WAAW,MAAM,GAAG,EAAE,CAAC;AAAA,IACzP;AAAA,EACF,WAAW,OAAO,WAAW,kBAAkB;AAC7C,UAAM;AAAA,MACJ,sCAAsC,OAAO,OAAO;AAAA,IACtD;AAAA,EACF;AACA,SAAO;AACT;AAEA,SAAS,SAAS,QAA8B,MAAe;AAC7D,MAAI,KAAM,QAAO;AACjB,SAAO;AAAA,IACL,YAAY,OAAO;AAAA,IACnB,KAAK,OAAO;AAAA,IACZ,YAAY,OAAO;AAAA,IACnB,OAAO,OAAO,MAAM,IAAI,CAAC,EAAE,MAAM,OAAO,QAAQ,MAAO,UAAU,EAAE,MAAM,OAAO,QAAQ,IAAI,EAAE,MAAM,MAAM,CAAE;AAAA,EAC9G;AACF;AAEA,SAAS,UAAU,QAAmC,MAAe;AACnE,MAAI,KAAM,QAAO;AACjB,SAAO;AAAA,IACL,YAAY,OAAO;AAAA,IACnB,KAAK,OAAO;AAAA,IACZ,YAAY,OAAO;AAAA,IACnB,WAAW,OAAO;AAAA,IAClB,SAAS,OAAO;AAAA,IAChB,aAAa,OAAO;AAAA,IACpB,GAAI,OAAO,YAAY,EAAE,WAAW,KAAK,IAAI,CAAC;AAAA,IAC9C,OAAO,OAAO,MAAM,IAAI,CAAC,EAAE,MAAM,OAAO,SAAS,OAAO,EAAE,MAAM,OAAO,SAAS,EAAE;AAAA,EACpF;AACF;AAGA,SAAS,QAAQ,MAA2B;AAC1C,MAAI,SAAS,sBAAsB,SAAS,eAAe;AACzD,WAAO;AAAA,MACL,QAAQ;AAAA,MACR,aACE;AAAA,IACJ;AAAA,EACF;AACA,MAAI,SAAS,aAAa;AACxB,WAAO;AAAA,MACL,QAAQ;AAAA,MACR,aACE;AAAA,IACJ;AAAA,EACF;AACA,SAAO;AAAA,IACL,QAAQ;AAAA,IACR,aACE;AAAA,EACJ;AACF;AAEA,SAAS,eAAe,SAAwB;AAC9C,oBAAkB,EAAE,IAAI,OAAO,OAAO,EAAE,MAAM,oBAAoB,SAAS,KAAK,QAAQ,kBAAkB,EAAE,EAAE,CAAC;AAC/G,UAAQ,KAAK,CAAC;AAChB;AAEO,IAAM,cAAc,cAAc;AAAA,EACvC,MAAM;AAAA,IACJ,MAAM;AAAA,IACN,aAAa;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAYf;AAAA,EACA,MAAM;AAAA,IACJ,KAAK,EAAE,MAAM,cAAc,aAAa,yBAAyB,UAAU,MAAM;AAAA,IACjF,MAAM,EAAE,MAAM,UAAU,aAAa,iDAAiD,UAAU,MAAM;AAAA,IACtG,OAAO,EAAE,MAAM,UAAU,aAAa,0BAA0B,UAAU,MAAM;AAAA,IAChF,MAAM,EAAE,MAAM,WAAW,aAAa,sBAAsB,UAAU,MAAM;AAAA,EAC9E;AAAA,EACA,KAAK,OAAO,EAAE,MAAM,QAAQ,MAAM;AAChC,UAAM,OAAO,QAAQ,KAAK,IAAI;AAC9B,UAAM,QAAQ,OAAO,KAAK,UAAU,WAAW,KAAK,MAAM,KAAK,IAAI;AACnE,QAAI;AACF,UAAI,OAAO;AACT,cAAMA,QAAiC,EAAE,YAAY,MAAM;AAC3D,cAAMC,UAAS,MAAM,QAAmC,4BAA4BD,OAAM;AAAA,UACxF,WAAW;AAAA,QACb,CAAC;AACD,0BAAkB,EAAE,IAAI,MAAM,MAAM,UAAUC,SAAQ,IAAI,GAAG,OAAO,eAAeA,OAAM,EAAE,CAAC;AAC5F;AAAA,MACF;AAEA,YAAM,eAAe,MAAM,QAAQ,KAAK,CAAC,IAAI,KAAK,IAAI,CAAC,GAAG;AAAA,QACxD,CAAC,SAAyB,OAAO,SAAS,YAAY,KAAK,KAAK,EAAE,SAAS;AAAA,MAC7E;AACA,YAAM,CAAC,KAAK,GAAG,eAAe,IAAI;AAClC,UAAI,CAAC,IAAK,gBAAe,mDAAmD;AAC5E,YAAM,QAAQ,CAAC,GAAG,iBAAiB,GAAG,eAAe,SAAS,QAAQ,KAAK,IAAI,CAAC;AAChF,UAAI,MAAM,WAAW,EAAG,gBAAe,sCAAsC;AAE7E,YAAM,OAA4B,EAAE,KAAK,MAAM;AAC/C,YAAM,SAAS,MAAM,QAA8B,sBAAsB,MAAM,EAAE,WAAW,WAAW,CAAC;AACxG,wBAAkB,EAAE,IAAI,MAAM,MAAM,SAAS,QAAQ,IAAI,GAAG,QAAQ,QAAQ,OAAO,UAAU,MAAM,EAAE,CAAC;AAAA,IACxG,SAAS,KAAK;AACZ,0BAAoB,KAAK,OAAO;AAAA,IAClC;AAAA,EACF;AACF,CAAC;","names":["body","result"]}