@fs/mycroft 0.2.0 → 0.3.0

package/dist/chunk-6DLQHHCC.js

@@ -0,0 +1,249 @@
+// src/services/batch-embedder.ts
+import OpenAI from "openai";
+
+// src/services/constants.ts
+import { mkdir as mkdir2 } from "fs/promises";
+
+// src/config.ts
+import { mkdir, readFile } from "fs/promises";
+import { homedir } from "os";
+import { dirname, join, resolve } from "path";
+var DEFAULT_CONFIG = {
+  dataDir: "~/.local/share/mycroft",
+  askEnabled: true,
+  models: {
+    embedding: "text-embedding-3-small",
+    summary: "gpt-5-nano",
+    chat: "gpt-5.1"
+  }
+};
+var expandHome = (input) => {
+  if (!input.startsWith("~")) return input;
+  return join(homedir(), input.slice(1));
+};
+var resolvePath = (input) => resolve(expandHome(input));
+var getConfigPath = () => {
+  const override = process.env.MYCROFT_CONFIG;
+  if (override) return resolvePath(override);
+  return resolvePath("~/.config/mycroft/config.json");
+};
+var normalizeModels = (models) => ({
+  embedding: models?.embedding || DEFAULT_CONFIG.models.embedding,
+  summary: models?.summary || DEFAULT_CONFIG.models.summary,
+  chat: models?.chat || DEFAULT_CONFIG.models.chat
+});
+var overrides = {};
+var setConfigOverrides = (next) => {
+  overrides = { ...overrides, ...next };
+};
+var normalizeConfig = (input) => {
+  const dataDirEnv = process.env.MYCROFT_DATA_DIR;
+  const dataDir = overrides.dataDir || dataDirEnv || input?.dataDir || DEFAULT_CONFIG.dataDir;
+  return {
+    dataDir,
+    askEnabled: input?.askEnabled ?? DEFAULT_CONFIG.askEnabled,
+    models: normalizeModels(input?.models)
+  };
+};
+var readConfigFile = async (path) => {
+  try {
+    const contents = await readFile(path, "utf-8");
+    return JSON.parse(contents);
+  } catch {
+    return null;
+  }
+};
+var loadConfig = async () => {
+  const configPath2 = getConfigPath();
+  const data = await readConfigFile(configPath2);
+  const normalized = normalizeConfig(data);
+  return {
+    ...normalized,
+    dataDir: resolvePath(normalized.dataDir)
+  };
+};
+var ensureConfigDirs = async (configPath2) => {
+  const path = configPath2 || getConfigPath();
+  await mkdir(dirname(path), { recursive: true });
+};
+var configPath = () => getConfigPath();
+
+// src/commands/io.ts
+import chalk from "chalk";
+var isTTY = () => Boolean(process.stdout.isTTY);
+var isInteractive = () => Boolean(process.stdin.isTTY && process.stdout.isTTY);
+var formatError = (text) => isTTY() ? chalk.red(text) : text;
+var formatWarn = (text) => isTTY() ? chalk.yellow(text) : text;
+var stdout = (message) => {
+  process.stdout.write(message.endsWith("\n") ? message : `${message}
+`);
+};
+var stderr = (message) => {
+  process.stderr.write(message.endsWith("\n") ? message : `${message}
+`);
+};
+var printError = (message) => {
+  stderr(formatError(`Error: ${message}`));
+};
+var logInfo = (message) => {
+  stderr(message);
+};
+var logWarn = (message) => {
+  stderr(formatWarn(message));
+};
+var handleSigint = (onCancel) => {
+  const handler = () => {
+    if (onCancel) onCancel();
+    stderr("\nCancelled.");
+    process.exit(130);
+  };
+  process.once("SIGINT", handler);
+  return () => process.off("SIGINT", handler);
+};
+
+// src/services/constants.ts
+var CHUNK_SIZE = 1e3;
+var CHUNK_OVERLAP = 100;
+var SEPARATORS = ["\n\n", "\n", ". ", " ", ""];
+var SUMMARY_MAX_TOKENS = 3e4;
+var SUMMARY_CONCURRENCY = 3;
+var SUMMARY_TARGET_WORDS = 250;
+var resolvePaths = async () => {
+  const config = await loadConfig();
+  const dataDir = config.dataDir;
+  return {
+    dataDir,
+    booksDir: `${dataDir}/books`,
+    vectorsDir: `${dataDir}/vectors`,
+    ingestDir: `${dataDir}/ingest`,
+    dbPath: `${dataDir}/metadata.db`
+  };
+};
+var ensureDataDirs = async () => {
+  const paths = await resolvePaths();
+  await mkdir2(paths.dataDir, { recursive: true });
+  await mkdir2(paths.booksDir, { recursive: true });
+  await mkdir2(paths.vectorsDir, { recursive: true });
+  await mkdir2(paths.ingestDir, { recursive: true });
+  return paths;
+};
+var getModels = async () => {
+  const config = await loadConfig();
+  return config.models;
+};
+var isAskEnabled = async () => {
+  const config = await loadConfig();
+  return config.askEnabled;
+};
+var requireOpenAIKey = () => {
+  if (!process.env.OPENAI_API_KEY) {
+    throw new Error("OPENAI_API_KEY is not set. Export it to use embeddings and chat.");
+  }
+};
+
+// src/services/batch-embedder.ts
+var buildJsonl = (chunks, model) => chunks.map(
+  (chunk, i) => ({
+    custom_id: String(i),
+    method: "POST",
+    url: "/v1/embeddings",
+    body: { model, input: chunk.content }
+  })
+).map((line) => JSON.stringify(line)).join("\n");
+var submitBatchEmbeddings = async (chunks) => {
+  const models = await getModels();
+  const client = new OpenAI();
+  logInfo(`[BatchEmbedder] Preparing batch request for ${chunks.length} chunks`);
+  const jsonl = buildJsonl(chunks, models.embedding);
+  const blob = new Blob([jsonl], { type: "application/jsonl" });
+  const file = await client.files.create({
+    file: new File([blob], "embeddings.jsonl", { type: "application/jsonl" }),
+    purpose: "batch"
+  });
+  logInfo(`[BatchEmbedder] Uploaded input file ${file.id}`);
+  const batch = await client.batches.create({
+    input_file_id: file.id,
+    endpoint: "/v1/embeddings",
+    completion_window: "24h"
+  });
+  logInfo(`[BatchEmbedder] Created batch ${batch.id} \u2014 status: ${batch.status}`);
+  return { batchId: batch.id, inputFileId: file.id };
+};
+var checkBatchStatus = async (batchId) => {
+  const client = new OpenAI();
+  const batch = await client.batches.retrieve(batchId);
+  return {
+    status: batch.status,
+    completed: batch.request_counts?.completed ?? 0,
+    total: batch.request_counts?.total ?? 0,
+    outputFileId: batch.output_file_id ?? null
+  };
+};
+var downloadBatchResults = async (outputFileId, chunks) => {
+  const client = new OpenAI();
+  logInfo(`[BatchEmbedder] Downloading results from ${outputFileId}`);
+  const response = await client.files.content(outputFileId);
+  const text = await response.text();
+  const lines = text.trim().split("\n");
+  const vectors = /* @__PURE__ */ new Map();
+  for (const line of lines) {
+    const result = JSON.parse(line);
+    const idx = Number(result.custom_id);
+    if (result.response?.status_code === 200) {
+      const embedding = result.response.body?.data?.[0]?.embedding;
+      if (embedding) {
+        vectors.set(idx, embedding);
+      }
+    } else {
+      logWarn(
+        `[BatchEmbedder] Request ${idx} failed: ${JSON.stringify(result.response?.body?.error ?? result.error)}`
+      );
+    }
+  }
+  const embedded = chunks.map((chunk, i) => ({
+    ...chunk,
+    vector: vectors.get(i) ?? []
+  }));
+  const missing = embedded.filter((e) => e.vector.length === 0).length;
+  if (missing > 0) {
+    logWarn(`[BatchEmbedder] ${missing} chunk(s) have empty embeddings due to batch errors`);
+  }
+  logInfo(`[BatchEmbedder] Successfully processed ${embedded.length} chunks via batch API`);
+  return embedded;
+};
+var cleanupBatchFiles = async (inputFileId, outputFileId) => {
+  const client = new OpenAI();
+  await client.files.del(inputFileId).catch(() => void 0);
+  if (outputFileId) {
+    await client.files.del(outputFileId).catch(() => void 0);
+  }
+};
+
+export {
+  setConfigOverrides,
+  loadConfig,
+  ensureConfigDirs,
+  configPath,
+  isInteractive,
+  stdout,
+  printError,
+  logInfo,
+  logWarn,
+  handleSigint,
+  CHUNK_SIZE,
+  CHUNK_OVERLAP,
+  SEPARATORS,
+  SUMMARY_MAX_TOKENS,
+  SUMMARY_CONCURRENCY,
+  SUMMARY_TARGET_WORDS,
+  resolvePaths,
+  ensureDataDirs,
+  getModels,
+  isAskEnabled,
+  requireOpenAIKey,
+  submitBatchEmbeddings,
+  checkBatchStatus,
+  downloadBatchResults,
+  cleanupBatchFiles
+};
+//# sourceMappingURL=chunk-6DLQHHCC.js.map

package/dist/chunk-6DLQHHCC.js.map

@@ -0,0 +1 @@
+{"version":3,"sources":["../src/services/batch-embedder.ts","../src/services/constants.ts","../src/config.ts","../src/commands/io.ts"],"sourcesContent":["import OpenAI from \"openai\";\nimport type { BookChunk } from \"../shared/types.js\";\nimport type { EmbeddedChunk } from \"./embedder.js\";\nimport { getModels, logInfo, logWarn } from \"./constants.js\";\n\ntype BatchRequestLine = {\n  custom_id: string;\n  method: \"POST\";\n  url: \"/v1/embeddings\";\n  body: { model: string; input: string };\n};\n\nconst buildJsonl = (chunks: BookChunk[], model: string): string =>\n  chunks\n    .map(\n      (chunk, i): BatchRequestLine => ({\n        custom_id: String(i),\n        method: \"POST\",\n        url: \"/v1/embeddings\",\n        body: { model, input: chunk.content },\n      })\n    )\n    .map((line) => JSON.stringify(line))\n    .join(\"\\n\");\n\nexport type BatchSubmitResult = {\n  batchId: string;\n  inputFileId: string;\n};\n\nexport const submitBatchEmbeddings = async (chunks: BookChunk[]): Promise<BatchSubmitResult> => {\n  const models = await getModels();\n  const client = new OpenAI();\n\n  logInfo(`[BatchEmbedder] Preparing batch request for ${chunks.length} chunks`);\n\n  const jsonl = buildJsonl(chunks, models.embedding);\n  const blob = new Blob([jsonl], { type: \"application/jsonl\" });\n  const file = await client.files.create({\n    file: new File([blob], \"embeddings.jsonl\", { type: \"application/jsonl\" }),\n    purpose: \"batch\",\n  });\n  logInfo(`[BatchEmbedder] Uploaded input file ${file.id}`);\n\n  const batch = await client.batches.create({\n    input_file_id: file.id,\n    endpoint: \"/v1/embeddings\",\n    completion_window: \"24h\",\n  });\n  logInfo(`[BatchEmbedder] Created batch ${batch.id} — status: ${batch.status}`);\n\n  return { batchId: batch.id, inputFileId: file.id };\n};\n\nexport type BatchStatus = {\n  status: string;\n  completed: number;\n  total: number;\n  outputFileId: string | null;\n};\n\nexport const checkBatchStatus = async (batchId: string): Promise<BatchStatus> => {\n  const client = new OpenAI();\n  const batch = await client.batches.retrieve(batchId);\n  return {\n    status: batch.status,\n    completed: batch.request_counts?.completed ?? 0,\n    total: batch.request_counts?.total ?? 0,\n    outputFileId: batch.output_file_id ?? null,\n  };\n};\n\nexport const downloadBatchResults = async (\n  outputFileId: string,\n  chunks: BookChunk[],\n): Promise<EmbeddedChunk[]> => {\n  const client = new OpenAI();\n\n  logInfo(`[BatchEmbedder] Downloading results from ${outputFileId}`);\n  const response = await client.files.content(outputFileId);\n  const text = await response.text();\n  const lines = text.trim().split(\"\\n\");\n\n  const vectors = new Map<number, number[]>();\n  for (const line of lines) {\n    const result = JSON.parse(line);\n    const idx = Number(result.custom_id);\n    if (result.response?.status_code === 200) {\n      const embedding = result.response.body?.data?.[0]?.embedding;\n      if (embedding) {\n        vectors.set(idx, embedding);\n      }\n    } else {\n      logWarn(\n        `[BatchEmbedder] Request ${idx} failed: ${JSON.stringify(result.response?.body?.error ?? result.error)}`\n      );\n    }\n  }\n\n  const embedded: EmbeddedChunk[] = chunks.map((chunk, i) => ({\n    ...chunk,\n    vector: vectors.get(i) ?? [],\n  }));\n\n  const missing = embedded.filter((e) => e.vector.length === 0).length;\n  if (missing > 0) {\n    logWarn(`[BatchEmbedder] ${missing} chunk(s) have empty embeddings due to batch errors`);\n  }\n\n  logInfo(`[BatchEmbedder] Successfully processed ${embedded.length} chunks via batch API`);\n  return embedded;\n};\n\nexport const cleanupBatchFiles = async (inputFileId: string, outputFileId?: string | null) => {\n  const client = new OpenAI();\n  await client.files.del(inputFileId).catch(() => undefined);\n  if (outputFileId) {\n    await client.files.del(outputFileId).catch(() => undefined);\n  }\n};\n","import { mkdir } from \"node:fs/promises\";\nimport { loadConfig } from \"../config.js\";\nimport { logInfo, logWarn } from \"../commands/io.js\";\n\nexport const CHUNK_SIZE: number = 1000;\nexport const CHUNK_OVERLAP: number = 100;\nexport const SEPARATORS = [\"\\n\\n\", \"\\n\", \". \", \" \", \"\"] as const;\n\nexport const SUMMARY_MAX_TOKENS = 30000;\nexport const SUMMARY_CONCURRENCY = 3;\nexport const SUMMARY_TARGET_WORDS = 250;\n\nexport type ResolvedPaths = {\n  dataDir: string;\n  booksDir: string;\n  vectorsDir: string;\n  ingestDir: string;\n  dbPath: string;\n};\n\nexport const resolvePaths = async (): Promise<ResolvedPaths> => {\n  const config = await loadConfig();\n  const dataDir = config.dataDir;\n  return {\n    dataDir,\n    booksDir: `${dataDir}/books`,\n    vectorsDir: `${dataDir}/vectors`,\n    ingestDir: `${dataDir}/ingest`,\n    dbPath: `${dataDir}/metadata.db`,\n  };\n};\n\nexport const ensureDataDirs = async () => {\n  const paths = await resolvePaths();\n  await mkdir(paths.dataDir, { recursive: true });\n  await mkdir(paths.booksDir, { recursive: true });\n  await mkdir(paths.vectorsDir, { recursive: true });\n  await mkdir(paths.ingestDir, { recursive: true });\n  return paths;\n};\n\nexport const getModels = async () => {\n  const config = await loadConfig();\n  return config.models;\n};\n\nexport const isAskEnabled = async () => {\n  const config = await loadConfig();\n  return config.askEnabled;\n};\n\nexport const requireOpenAIKey = () => {\n  if (!process.env.OPENAI_API_KEY) {\n    throw new Error(\"OPENAI_API_KEY is not set. Export it to use embeddings and chat.\");\n  }\n};\n\nexport { logInfo, logWarn };\n","import { mkdir, readFile } from \"node:fs/promises\";\nimport { homedir } from \"node:os\";\nimport { dirname, join, resolve } from \"node:path\";\n\nexport type ConfigModels = {\n  embedding: string;\n  summary: string;\n  chat: string;\n};\n\nexport type AppConfig = {\n  dataDir: string;\n  askEnabled: boolean;\n  models: ConfigModels;\n};\n\nconst DEFAULT_CONFIG: AppConfig = {\n  dataDir: \"~/.local/share/mycroft\",\n  askEnabled: true,\n  models: {\n    embedding: \"text-embedding-3-small\",\n    summary: \"gpt-5-nano\",\n    chat: \"gpt-5.1\",\n  },\n};\n\nconst expandHome = (input: string): string => {\n  if (!input.startsWith(\"~\")) return input;\n  return join(homedir(), input.slice(1));\n};\n\nconst resolvePath = (input: string): string => resolve(expandHome(input));\n\nconst getConfigPath = (): string => {\n  const override = process.env.MYCROFT_CONFIG;\n  if (override) return resolvePath(override);\n  return resolvePath(\"~/.config/mycroft/config.json\");\n};\n\nconst normalizeModels = (models?: Partial<ConfigModels>): ConfigModels => ({\n  embedding: models?.embedding || DEFAULT_CONFIG.models.embedding,\n  summary: models?.summary || DEFAULT_CONFIG.models.summary,\n  chat: models?.chat || DEFAULT_CONFIG.models.chat,\n});\n\ntype ConfigOverrides = {\n  dataDir?: string;\n};\n\nlet overrides: ConfigOverrides = {};\n\nexport const setConfigOverrides = (next: ConfigOverrides) => {\n  overrides = { ...overrides, ...next };\n};\n\nconst normalizeConfig = (input: Partial<AppConfig> | null): AppConfig => {\n  const dataDirEnv = process.env.MYCROFT_DATA_DIR;\n  const dataDir = overrides.dataDir || dataDirEnv || input?.dataDir || DEFAULT_CONFIG.dataDir;\n  return {\n    dataDir,\n    askEnabled: input?.askEnabled ?? DEFAULT_CONFIG.askEnabled,\n    models: normalizeModels(input?.models),\n  };\n};\n\nconst readConfigFile = async (path: string): Promise<Partial<AppConfig> | null> => {\n  try {\n    const contents = await readFile(path, \"utf-8\");\n    return JSON.parse(contents) as Partial<AppConfig>;\n  } catch {\n    return null;\n  }\n};\n\nexport const loadConfig = async (): Promise<AppConfig> => {\n  const configPath = getConfigPath();\n  const data = await readConfigFile(configPath);\n  const normalized = normalizeConfig(data);\n  return {\n    ...normalized,\n    dataDir: resolvePath(normalized.dataDir),\n  };\n};\n\nexport const ensureConfigDirs = async (configPath?: string) => {\n  const path = configPath || getConfigPath();\n  await mkdir(dirname(path), { recursive: true });\n};\n\nexport const configPath = () => getConfigPath();\n","import chalk from \"chalk\";\n\nconst isTTY = () => Boolean(process.stdout.isTTY);\nexport const isInteractive = () => Boolean(process.stdin.isTTY && process.stdout.isTTY);\n\nexport const formatDim = (text: string) => (isTTY() ? chalk.dim(text) : text);\nexport const formatError = (text: string) => (isTTY() ? chalk.red(text) : text);\nexport const formatBold = (text: string) => (isTTY() ? chalk.bold(text) : text);\nexport const formatWarn = (text: string) => (isTTY() ? chalk.yellow(text) : text);\n\nexport const stdout = (message: string) => {\n  process.stdout.write(message.endsWith(\"\\n\") ? message : `${message}\\n`);\n};\n\nexport const stderr = (message: string) => {\n  process.stderr.write(message.endsWith(\"\\n\") ? message : `${message}\\n`);\n};\n\nexport const printError = (message: string) => {\n  stderr(formatError(`Error: ${message}`));\n};\n\nexport const logInfo = (message: string) => {\n  stderr(message);\n};\n\nexport const logWarn = (message: string) => {\n  stderr(formatWarn(message));\n};\n\nexport const handleSigint = (onCancel?: () => void) => {\n  const handler = () => {\n    if (onCancel) onCancel();\n    stderr(\"\\nCancelled.\");\n    process.exit(130);\n  };\n  process.once(\"SIGINT\", handler);\n  return () => process.off(\"SIGINT\", handler);\n};\n"],"mappings":";AAAA,OAAO,YAAY;;;ACAnB,SAAS,SAAAA,cAAa;;;ACAtB,SAAS,OAAO,gBAAgB;AAChC,SAAS,eAAe;AACxB,SAAS,SAAS,MAAM,eAAe;AAcvC,IAAM,iBAA4B;AAAA,EAChC,SAAS;AAAA,EACT,YAAY;AAAA,EACZ,QAAQ;AAAA,IACN,WAAW;AAAA,IACX,SAAS;AAAA,IACT,MAAM;AAAA,EACR;AACF;AAEA,IAAM,aAAa,CAAC,UAA0B;AAC5C,MAAI,CAAC,MAAM,WAAW,GAAG,EAAG,QAAO;AACnC,SAAO,KAAK,QAAQ,GAAG,MAAM,MAAM,CAAC,CAAC;AACvC;AAEA,IAAM,cAAc,CAAC,UAA0B,QAAQ,WAAW,KAAK,CAAC;AAExE,IAAM,gBAAgB,MAAc;AAClC,QAAM,WAAW,QAAQ,IAAI;AAC7B,MAAI,SAAU,QAAO,YAAY,QAAQ;AACzC,SAAO,YAAY,+BAA+B;AACpD;AAEA,IAAM,kBAAkB,CAAC,YAAkD;AAAA,EACzE,WAAW,QAAQ,aAAa,eAAe,OAAO;AAAA,EACtD,SAAS,QAAQ,WAAW,eAAe,OAAO;AAAA,EAClD,MAAM,QAAQ,QAAQ,eAAe,OAAO;AAC9C;AAMA,IAAI,YAA6B,CAAC;AAE3B,IAAM,qBAAqB,CAAC,SAA0B;AAC3D,cAAY,EAAE,GAAG,WAAW,GAAG,KAAK;AACtC;AAEA,IAAM,kBAAkB,CAAC,UAAgD;AACvE,QAAM,aAAa,QAAQ,IAAI;AAC/B,QAAM,UAAU,UAAU,WAAW,cAAc,OAAO,WAAW,eAAe;AACpF,SAAO;AAAA,IACL;AAAA,IACA,YAAY,OAAO,cAAc,eAAe;AAAA,IAChD,QAAQ,gBAAgB,OAAO,MAAM;AAAA,EACvC;AACF;AAEA,IAAM,iBAAiB,OAAO,SAAqD;AACjF,MAAI;AACF,UAAM,WAAW,MAAM,SAAS,MAAM,OAAO;AAC7C,WAAO,KAAK,MAAM,QAAQ;AAAA,EAC5B,QAAQ;AACN,WAAO;AAAA,EACT;AACF;AAEO,IAAM,aAAa,YAAgC;AACxD,QAAMC,cAAa,cAAc;AACjC,QAAM,OAAO,MAAM,eAAeA,WAAU;AAC5C,QAAM,aAAa,gBAAgB,IAAI;AACvC,SAAO;AAAA,IACL,GAAG;AAAA,IACH,SAAS,YAAY,WAAW,OAAO;AAAA,EACzC;AACF;AAEO,IAAM,mBAAmB,OAAOA,gBAAwB;AAC7D,QAAM,OAAOA,eAAc,cAAc;AACzC,QAAM,MAAM,QAAQ,IAAI,GAAG,EAAE,WAAW,KAAK,CAAC;AAChD;AAEO,IAAM,aAAa,MAAM,cAAc;;;ACzF9C,OAAO,WAAW;AAElB,IAAM,QAAQ,MAAM,QAAQ,QAAQ,OAAO,KAAK;AACzC,IAAM,gBAAgB,MAAM,QAAQ,QAAQ,MAAM,SAAS,QAAQ,OAAO,KAAK;AAG/E,IAAM,cAAc,CAAC,SAAkB,MAAM,IAAI,MAAM,IAAI,IAAI,IAAI;AAEnE,IAAM,aAAa,CAAC,SAAkB,MAAM,IAAI,MAAM,OAAO,IAAI,IAAI;AAErE,IAAM,SAAS,CAAC,YAAoB;AACzC,UAAQ,OAAO,MAAM,QAAQ,SAAS,IAAI,IAAI,UAAU,GAAG,OAAO;AAAA,CAAI;AACxE;AAEO,IAAM,SAAS,CAAC,YAAoB;AACzC,UAAQ,OAAO,MAAM,QAAQ,SAAS,IAAI,IAAI,UAAU,GAAG,OAAO;AAAA,CAAI;AACxE;AAEO,IAAM,aAAa,CAAC,YAAoB;AAC7C,SAAO,YAAY,UAAU,OAAO,EAAE,CAAC;AACzC;AAEO,IAAM,UAAU,CAAC,YAAoB;AAC1C,SAAO,OAAO;AAChB;AAEO,IAAM,UAAU,CAAC,YAAoB;AAC1C,SAAO,WAAW,OAAO,CAAC;AAC5B;AAEO,IAAM,eAAe,CAAC,aAA0B;AACrD,QAAM,UAAU,MAAM;AACpB,QAAI,SAAU,UAAS;AACvB,WAAO,cAAc;AACrB,YAAQ,KAAK,GAAG;AAAA,EAClB;AACA,UAAQ,KAAK,UAAU,OAAO;AAC9B,SAAO,MAAM,QAAQ,IAAI,UAAU,OAAO;AAC5C;;;AFlCO,IAAM,aAAqB;AAC3B,IAAM,gBAAwB;AAC9B,IAAM,aAAa,CAAC,QAAQ,MAAM,MAAM,KAAK,EAAE;AAE/C,IAAM,qBAAqB;AAC3B,IAAM,sBAAsB;AAC5B,IAAM,uBAAuB;AAU7B,IAAM,eAAe,YAAoC;AAC9D,QAAM,SAAS,MAAM,WAAW;AAChC,QAAM,UAAU,OAAO;AACvB,SAAO;AAAA,IACL;AAAA,IACA,UAAU,GAAG,OAAO;AAAA,IACpB,YAAY,GAAG,OAAO;AAAA,IACtB,WAAW,GAAG,OAAO;AAAA,IACrB,QAAQ,GAAG,OAAO;AAAA,EACpB;AACF;AAEO,IAAM,iBAAiB,YAAY;AACxC,QAAM,QAAQ,MAAM,aAAa;AACjC,QAAMC,OAAM,MAAM,SAAS,EAAE,WAAW,KAAK,CAAC;AAC9C,QAAMA,OAAM,MAAM,UAAU,EAAE,WAAW,KAAK,CAAC;AAC/C,QAAMA,OAAM,MAAM,YAAY,EAAE,WAAW,KAAK,CAAC;AACjD,QAAMA,OAAM,MAAM,WAAW,EAAE,WAAW,KAAK,CAAC;AAChD,SAAO;AACT;AAEO,IAAM,YAAY,YAAY;AACnC,QAAM,SAAS,MAAM,WAAW;AAChC,SAAO,OAAO;AAChB;AAEO,IAAM,eAAe,YAAY;AACtC,QAAM,SAAS,MAAM,WAAW;AAChC,SAAO,OAAO;AAChB;AAEO,IAAM,mBAAmB,MAAM;AACpC,MAAI,CAAC,QAAQ,IAAI,gBAAgB;AAC/B,UAAM,IAAI,MAAM,kEAAkE;AAAA,EACpF;AACF;;;AD3CA,IAAM,aAAa,CAAC,QAAqB,UACvC,OACG;AAAA,EACC,CAAC,OAAO,OAAyB;AAAA,IAC/B,WAAW,OAAO,CAAC;AAAA,IACnB,QAAQ;AAAA,IACR,KAAK;AAAA,IACL,MAAM,EAAE,OAAO,OAAO,MAAM,QAAQ;AAAA,EACtC;AACF,EACC,IAAI,CAAC,SAAS,KAAK,UAAU,IAAI,CAAC,EAClC,KAAK,IAAI;AAOP,IAAM,wBAAwB,OAAO,WAAoD;AAC9F,QAAM,SAAS,MAAM,UAAU;AAC/B,QAAM,SAAS,IAAI,OAAO;AAE1B,UAAQ,+CAA+C,OAAO,MAAM,SAAS;AAE7E,QAAM,QAAQ,WAAW,QAAQ,OAAO,SAAS;AACjD,QAAM,OAAO,IAAI,KAAK,CAAC,KAAK,GAAG,EAAE,MAAM,oBAAoB,CAAC;AAC5D,QAAM,OAAO,MAAM,OAAO,MAAM,OAAO;AAAA,IACrC,MAAM,IAAI,KAAK,CAAC,IAAI,GAAG,oBAAoB,EAAE,MAAM,oBAAoB,CAAC;AAAA,IACxE,SAAS;AAAA,EACX,CAAC;AACD,UAAQ,uCAAuC,KAAK,EAAE,EAAE;AAExD,QAAM,QAAQ,MAAM,OAAO,QAAQ,OAAO;AAAA,IACxC,eAAe,KAAK;AAAA,IACpB,UAAU;AAAA,IACV,mBAAmB;AAAA,EACrB,CAAC;AACD,UAAQ,iCAAiC,MAAM,EAAE,mBAAc,MAAM,MAAM,EAAE;AAE7E,SAAO,EAAE,SAAS,MAAM,IAAI,aAAa,KAAK,GAAG;AACnD;AASO,IAAM,mBAAmB,OAAO,YAA0C;AAC/E,QAAM,SAAS,IAAI,OAAO;AAC1B,QAAM,QAAQ,MAAM,OAAO,QAAQ,SAAS,OAAO;AACnD,SAAO;AAAA,IACL,QAAQ,MAAM;AAAA,IACd,WAAW,MAAM,gBAAgB,aAAa;AAAA,IAC9C,OAAO,MAAM,gBAAgB,SAAS;AAAA,IACtC,cAAc,MAAM,kBAAkB;AAAA,EACxC;AACF;AAEO,IAAM,uBAAuB,OAClC,cACA,WAC6B;AAC7B,QAAM,SAAS,IAAI,OAAO;AAE1B,UAAQ,4CAA4C,YAAY,EAAE;AAClE,QAAM,WAAW,MAAM,OAAO,MAAM,QAAQ,YAAY;AACxD,QAAM,OAAO,MAAM,SAAS,KAAK;AACjC,QAAM,QAAQ,KAAK,KAAK,EAAE,MAAM,IAAI;AAEpC,QAAM,UAAU,oBAAI,IAAsB;AAC1C,aAAW,QAAQ,OAAO;AACxB,UAAM,SAAS,KAAK,MAAM,IAAI;AAC9B,UAAM,MAAM,OAAO,OAAO,SAAS;AACnC,QAAI,OAAO,UAAU,gBAAgB,KAAK;AACxC,YAAM,YAAY,OAAO,SAAS,MAAM,OAAO,CAAC,GAAG;AACnD,UAAI,WAAW;AACb,gBAAQ,IAAI,KAAK,SAAS;AAAA,MAC5B;AAAA,IACF,OAAO;AACL;AAAA,QACE,2BAA2B,GAAG,YAAY,KAAK,UAAU,OAAO,UAAU,MAAM,SAAS,OAAO,KAAK,CAAC;AAAA,MACxG;AAAA,IACF;AAAA,EACF;AAEA,QAAM,WAA4B,OAAO,IAAI,CAAC,OAAO,OAAO;AAAA,IAC1D,GAAG;AAAA,IACH,QAAQ,QAAQ,IAAI,CAAC,KAAK,CAAC;AAAA,EAC7B,EAAE;AAEF,QAAM,UAAU,SAAS,OAAO,CAAC,MAAM,EAAE,OAAO,WAAW,CAAC,EAAE;AAC9D,MAAI,UAAU,GAAG;AACf,YAAQ,mBAAmB,OAAO,qDAAqD;AAAA,EACzF;AAEA,UAAQ,0CAA0C,SAAS,MAAM,uBAAuB;AACxF,SAAO;AACT;AAEO,IAAM,oBAAoB,OAAO,aAAqB,iBAAiC;AAC5F,QAAM,SAAS,IAAI,OAAO;AAC1B,QAAM,OAAO,MAAM,IAAI,WAAW,EAAE,MAAM,MAAM,MAAS;AACzD,MAAI,cAAc;AAChB,UAAM,OAAO,MAAM,IAAI,YAAY,EAAE,MAAM,MAAM,MAAS;AAAA,EAC5D;AACF;","names":["mkdir","configPath","mkdir"]}

package/dist/chunk-7CO4PMU5.js

@@ -0,0 +1,92 @@
+import {
+  getModels,
+  logInfo,
+  logWarn
+} from "./chunk-KGG7WEYE.js";
+
+// src/services/batch-embedder.ts
+import OpenAI from "openai";
+var buildJsonl = (chunks, model) => chunks.map(
+  (chunk, i) => ({
+    custom_id: String(i),
+    method: "POST",
+    url: "/v1/embeddings",
+    body: { model, input: chunk.content }
+  })
+).map((line) => JSON.stringify(line)).join("\n");
+var submitBatchEmbeddings = async (chunks) => {
+  const models = await getModels();
+  const client = new OpenAI();
+  logInfo(`[BatchEmbedder] Preparing batch request for ${chunks.length} chunks`);
+  const jsonl = buildJsonl(chunks, models.embedding);
+  const blob = new Blob([jsonl], { type: "application/jsonl" });
+  const file = await client.files.create({
+    file: new File([blob], "embeddings.jsonl", { type: "application/jsonl" }),
+    purpose: "batch"
+  });
+  logInfo(`[BatchEmbedder] Uploaded input file ${file.id}`);
+  const batch = await client.batches.create({
+    input_file_id: file.id,
+    endpoint: "/v1/embeddings",
+    completion_window: "24h"
+  });
+  logInfo(`[BatchEmbedder] Created batch ${batch.id} \u2014 status: ${batch.status}`);
+  return { batchId: batch.id, inputFileId: file.id };
+};
+var checkBatchStatus = async (batchId) => {
+  const client = new OpenAI();
+  const batch = await client.batches.retrieve(batchId);
+  return {
+    status: batch.status,
+    completed: batch.request_counts?.completed ?? 0,
+    total: batch.request_counts?.total ?? 0,
+    outputFileId: batch.output_file_id ?? null
+  };
+};
+var downloadBatchResults = async (outputFileId, chunks) => {
+  const client = new OpenAI();
+  logInfo(`[BatchEmbedder] Downloading results from ${outputFileId}`);
+  const response = await client.files.content(outputFileId);
+  const text = await response.text();
+  const lines = text.trim().split("\n");
+  const vectors = /* @__PURE__ */ new Map();
+  for (const line of lines) {
+    const result = JSON.parse(line);
+    const idx = Number(result.custom_id);
+    if (result.response?.status_code === 200) {
+      const embedding = result.response.body?.data?.[0]?.embedding;
+      if (embedding) {
+        vectors.set(idx, embedding);
+      }
+    } else {
+      logWarn(
+        `[BatchEmbedder] Request ${idx} failed: ${JSON.stringify(result.response?.body?.error ?? result.error)}`
+      );
+    }
+  }
+  const embedded = chunks.map((chunk, i) => ({
+    ...chunk,
+    vector: vectors.get(i) ?? []
+  }));
+  const missing = embedded.filter((e) => e.vector.length === 0).length;
+  if (missing > 0) {
+    logWarn(`[BatchEmbedder] ${missing} chunk(s) have empty embeddings due to batch errors`);
+  }
+  logInfo(`[BatchEmbedder] Successfully processed ${embedded.length} chunks via batch API`);
+  return embedded;
+};
+var cleanupBatchFiles = async (inputFileId, outputFileId) => {
+  const client = new OpenAI();
+  await client.files.del(inputFileId).catch(() => void 0);
+  if (outputFileId) {
+    await client.files.del(outputFileId).catch(() => void 0);
+  }
+};
+
+export {
+  submitBatchEmbeddings,
+  checkBatchStatus,
+  downloadBatchResults,
+  cleanupBatchFiles
+};
+//# sourceMappingURL=chunk-7CO4PMU5.js.map

package/dist/chunk-7CO4PMU5.js.map

@@ -0,0 +1 @@
+{"version":3,"sources":["../src/services/batch-embedder.ts"],"sourcesContent":["import OpenAI from \"openai\";\nimport type { BookChunk } from \"../shared/types.js\";\nimport type { EmbeddedChunk } from \"./embedder.js\";\nimport { getModels, logInfo, logWarn } from \"./constants.js\";\n\ntype BatchRequestLine = {\n  custom_id: string;\n  method: \"POST\";\n  url: \"/v1/embeddings\";\n  body: { model: string; input: string };\n};\n\nconst buildJsonl = (chunks: BookChunk[], model: string): string =>\n  chunks\n    .map(\n      (chunk, i): BatchRequestLine => ({\n        custom_id: String(i),\n        method: \"POST\",\n        url: \"/v1/embeddings\",\n        body: { model, input: chunk.content },\n      })\n    )\n    .map((line) => JSON.stringify(line))\n    .join(\"\\n\");\n\nexport type BatchSubmitResult = {\n  batchId: string;\n  inputFileId: string;\n};\n\nexport const submitBatchEmbeddings = async (chunks: BookChunk[]): Promise<BatchSubmitResult> => {\n  const models = await getModels();\n  const client = new OpenAI();\n\n  logInfo(`[BatchEmbedder] Preparing batch request for ${chunks.length} chunks`);\n\n  const jsonl = buildJsonl(chunks, models.embedding);\n  const blob = new Blob([jsonl], { type: \"application/jsonl\" });\n  const file = await client.files.create({\n    file: new File([blob], \"embeddings.jsonl\", { type: \"application/jsonl\" }),\n    purpose: \"batch\",\n  });\n  logInfo(`[BatchEmbedder] Uploaded input file ${file.id}`);\n\n  const batch = await client.batches.create({\n    input_file_id: file.id,\n    endpoint: \"/v1/embeddings\",\n    completion_window: \"24h\",\n  });\n  logInfo(`[BatchEmbedder] Created batch ${batch.id} — status: ${batch.status}`);\n\n  return { batchId: batch.id, inputFileId: file.id };\n};\n\nexport type BatchStatus = {\n  status: string;\n  completed: number;\n  total: number;\n  outputFileId: string | null;\n};\n\nexport const checkBatchStatus = async (batchId: string): Promise<BatchStatus> => {\n  const client = new OpenAI();\n  const batch = await client.batches.retrieve(batchId);\n  return {\n    status: batch.status,\n    completed: batch.request_counts?.completed ?? 0,\n    total: batch.request_counts?.total ?? 0,\n    outputFileId: batch.output_file_id ?? null,\n  };\n};\n\nexport const downloadBatchResults = async (\n  outputFileId: string,\n  chunks: BookChunk[],\n): Promise<EmbeddedChunk[]> => {\n  const client = new OpenAI();\n\n  logInfo(`[BatchEmbedder] Downloading results from ${outputFileId}`);\n  const response = await client.files.content(outputFileId);\n  const text = await response.text();\n  const lines = text.trim().split(\"\\n\");\n\n  const vectors = new Map<number, number[]>();\n  for (const line of lines) {\n    const result = JSON.parse(line);\n    const idx = Number(result.custom_id);\n    if (result.response?.status_code === 200) {\n      const embedding = result.response.body?.data?.[0]?.embedding;\n      if (embedding) {\n        vectors.set(idx, embedding);\n      }\n    } else {\n      logWarn(\n        `[BatchEmbedder] Request ${idx} failed: ${JSON.stringify(result.response?.body?.error ?? result.error)}`\n      );\n    }\n  }\n\n  const embedded: EmbeddedChunk[] = chunks.map((chunk, i) => ({\n    ...chunk,\n    vector: vectors.get(i) ?? [],\n  }));\n\n  const missing = embedded.filter((e) => e.vector.length === 0).length;\n  if (missing > 0) {\n    logWarn(`[BatchEmbedder] ${missing} chunk(s) have empty embeddings due to batch errors`);\n  }\n\n  logInfo(`[BatchEmbedder] Successfully processed ${embedded.length} chunks via batch API`);\n  return embedded;\n};\n\nexport const cleanupBatchFiles = async (inputFileId: string, outputFileId?: string | null) => {\n  const client = new OpenAI();\n  await client.files.del(inputFileId).catch(() => undefined);\n  if (outputFileId) {\n    await client.files.del(outputFileId).catch(() => undefined);\n  }\n};\n"],"mappings":";;;;;;;AAAA,OAAO,YAAY;AAYnB,IAAM,aAAa,CAAC,QAAqB,UACvC,OACG;AAAA,EACC,CAAC,OAAO,OAAyB;AAAA,IAC/B,WAAW,OAAO,CAAC;AAAA,IACnB,QAAQ;AAAA,IACR,KAAK;AAAA,IACL,MAAM,EAAE,OAAO,OAAO,MAAM,QAAQ;AAAA,EACtC;AACF,EACC,IAAI,CAAC,SAAS,KAAK,UAAU,IAAI,CAAC,EAClC,KAAK,IAAI;AAOP,IAAM,wBAAwB,OAAO,WAAoD;AAC9F,QAAM,SAAS,MAAM,UAAU;AAC/B,QAAM,SAAS,IAAI,OAAO;AAE1B,UAAQ,+CAA+C,OAAO,MAAM,SAAS;AAE7E,QAAM,QAAQ,WAAW,QAAQ,OAAO,SAAS;AACjD,QAAM,OAAO,IAAI,KAAK,CAAC,KAAK,GAAG,EAAE,MAAM,oBAAoB,CAAC;AAC5D,QAAM,OAAO,MAAM,OAAO,MAAM,OAAO;AAAA,IACrC,MAAM,IAAI,KAAK,CAAC,IAAI,GAAG,oBAAoB,EAAE,MAAM,oBAAoB,CAAC;AAAA,IACxE,SAAS;AAAA,EACX,CAAC;AACD,UAAQ,uCAAuC,KAAK,EAAE,EAAE;AAExD,QAAM,QAAQ,MAAM,OAAO,QAAQ,OAAO;AAAA,IACxC,eAAe,KAAK;AAAA,IACpB,UAAU;AAAA,IACV,mBAAmB;AAAA,EACrB,CAAC;AACD,UAAQ,iCAAiC,MAAM,EAAE,mBAAc,MAAM,MAAM,EAAE;AAE7E,SAAO,EAAE,SAAS,MAAM,IAAI,aAAa,KAAK,GAAG;AACnD;AASO,IAAM,mBAAmB,OAAO,YAA0C;AAC/E,QAAM,SAAS,IAAI,OAAO;AAC1B,QAAM,QAAQ,MAAM,OAAO,QAAQ,SAAS,OAAO;AACnD,SAAO;AAAA,IACL,QAAQ,MAAM;AAAA,IACd,WAAW,MAAM,gBAAgB,aAAa;AAAA,IAC9C,OAAO,MAAM,gBAAgB,SAAS;AAAA,IACtC,cAAc,MAAM,kBAAkB;AAAA,EACxC;AACF;AAEO,IAAM,uBAAuB,OAClC,cACA,WAC6B;AAC7B,QAAM,SAAS,IAAI,OAAO;AAE1B,UAAQ,4CAA4C,YAAY,EAAE;AAClE,QAAM,WAAW,MAAM,OAAO,MAAM,QAAQ,YAAY;AACxD,QAAM,OAAO,MAAM,SAAS,KAAK;AACjC,QAAM,QAAQ,KAAK,KAAK,EAAE,MAAM,IAAI;AAEpC,QAAM,UAAU,oBAAI,IAAsB;AAC1C,aAAW,QAAQ,OAAO;AACxB,UAAM,SAAS,KAAK,MAAM,IAAI;AAC9B,UAAM,MAAM,OAAO,OAAO,SAAS;AACnC,QAAI,OAAO,UAAU,gBAAgB,KAAK;AACxC,YAAM,YAAY,OAAO,SAAS,MAAM,OAAO,CAAC,GAAG;AACnD,UAAI,WAAW;AACb,gBAAQ,IAAI,KAAK,SAAS;AAAA,MAC5B;AAAA,IACF,OAAO;AACL;AAAA,QACE,2BAA2B,GAAG,YAAY,KAAK,UAAU,OAAO,UAAU,MAAM,SAAS,OAAO,KAAK,CAAC;AAAA,MACxG;AAAA,IACF;AAAA,EACF;AAEA,QAAM,WAA4B,OAAO,IAAI,CAAC,OAAO,OAAO;AAAA,IAC1D,GAAG;AAAA,IACH,QAAQ,QAAQ,IAAI,CAAC,KAAK,CAAC;AAAA,EAC7B,EAAE;AAEF,QAAM,UAAU,SAAS,OAAO,CAAC,MAAM,EAAE,OAAO,WAAW,CAAC,EAAE;AAC9D,MAAI,UAAU,GAAG;AACf,YAAQ,mBAAmB,OAAO,qDAAqD;AAAA,EACzF;AAEA,UAAQ,0CAA0C,SAAS,MAAM,uBAAuB;AACxF,SAAO;AACT;AAEO,IAAM,oBAAoB,OAAO,aAAqB,iBAAiC;AAC5F,QAAM,SAAS,IAAI,OAAO;AAC1B,QAAM,OAAO,MAAM,IAAI,WAAW,EAAE,MAAM,MAAM,MAAS;AACzD,MAAI,cAAc;AAChB,UAAM,OAAO,MAAM,IAAI,YAAY,EAAE,MAAM,MAAM,MAAS;AAAA,EAC5D;AACF;","names":[]}

package/dist/chunk-7DUQNGEK.js

@@ -0,0 +1,253 @@
+import {
+  SUMMARY_MAX_TOKENS,
+  SUMMARY_TARGET_WORDS,
+  getModels,
+  logInfo,
+  logWarn
+} from "./chunk-KGG7WEYE.js";
+
+// src/services/batch-summarizer.ts
+import OpenAI from "openai";
+var CHARS_PER_TOKEN = 4;
+var estimateTokens = (text) => Math.ceil(text.length / CHARS_PER_TOKEN);
+var SUMMARY_PROMPT = (title, chapterNum, content) => `You are analyzing a chapter from a book (fiction or nonfiction). Extract key information to help readers understand the chapter's content.
+
+Chapter Title: ${title}
+Chapter Number: ${chapterNum}
+
+---
+${content}
+---
+
+Extract the following information and respond ONLY with valid JSON (no markdown, no code blocks):
+
+{
+  "characters": ["Name - brief description (role, traits, first appearance)", ...],
+  "events": "What happens in this chapter? (2-3 sentences)",
+  "setting": "Where does this chapter take place?",
+  "revelations": "Any important information revealed? (secrets, backstory, foreshadowing)"
+}
+
+Keep the total response around ${SUMMARY_TARGET_WORDS} words.`;
+var splitIntoSections = (text, maxTokens) => {
+  const estimatedTokens = estimateTokens(text);
+  if (estimatedTokens <= maxTokens) return [text];
+  const numSections = Math.ceil(estimatedTokens / maxTokens);
+  const charsPerSection = Math.floor(text.length / numSections);
+  const sections = [];
+  for (let i = 0; i < numSections; i++) {
+    const start = i * charsPerSection;
+    const end = i === numSections - 1 ? text.length : (i + 1) * charsPerSection;
+    sections.push(text.slice(start, end));
+  }
+  return sections;
+};
+var buildJsonl = (chapters, model) => {
+  const lines = [];
+  const metadata = [];
+  for (let i = 0; i < chapters.length; i++) {
+    const chapter = chapters[i];
+    const tokens = estimateTokens(chapter.content);
+    if (tokens <= SUMMARY_MAX_TOKENS) {
+      const line = {
+        custom_id: `summary-${i}`,
+        method: "POST",
+        url: "/v1/chat/completions",
+        body: {
+          model,
+          messages: [{ role: "user", content: SUMMARY_PROMPT(chapter.title, i + 1, chapter.content) }]
+        }
+      };
+      lines.push(JSON.stringify(line));
+      metadata.push({ chapterIndex: i, title: chapter.title, needsTwoPass: false, sectionCount: 1 });
+    } else {
+      const sections = splitIntoSections(chapter.content, SUMMARY_MAX_TOKENS);
+      for (let s = 0; s < sections.length; s++) {
+        const line = {
+          custom_id: `section-${i}-${s}`,
+          method: "POST",
+          url: "/v1/chat/completions",
+          body: {
+            model,
+            messages: [{
+              role: "user",
+              content: `Summarize this section from chapter "${chapter.title}" (Part ${s + 1}). Focus on key events, characters, and revelations. Keep it concise (100-150 words):
+
+${sections[s]}`
+            }]
+          }
+        };
+        lines.push(JSON.stringify(line));
+      }
+      metadata.push({ chapterIndex: i, title: chapter.title, needsTwoPass: true, sectionCount: sections.length });
+    }
+  }
+  return { jsonl: lines.join("\n"), metadata };
+};
+var submitBatchSummaries = async (chapters) => {
+  const models = await getModels();
+  const client = new OpenAI();
+  logInfo(`[BatchSummarizer] Preparing batch request for ${chapters.length} chapters`);
+  const { jsonl, metadata } = buildJsonl(chapters, models.summary);
+  const blob = new Blob([jsonl], { type: "application/jsonl" });
+  const file = await client.files.create({
+    file: new File([blob], "summaries.jsonl", { type: "application/jsonl" }),
+    purpose: "batch"
+  });
+  logInfo(`[BatchSummarizer] Uploaded input file ${file.id}`);
+  const batch = await client.batches.create({
+    input_file_id: file.id,
+    endpoint: "/v1/chat/completions",
+    completion_window: "24h"
+  });
+  logInfo(`[BatchSummarizer] Created batch ${batch.id} \u2014 status: ${batch.status}`);
+  return { batchId: batch.id, inputFileId: file.id, metadata };
+};
+var parseStructuredSummary = (text, chapterIndex, title) => {
+  try {
+    let jsonText = text.trim();
+    if (jsonText.startsWith("```json")) {
+      jsonText = jsonText.slice(7, -3).trim();
+    } else if (jsonText.startsWith("```")) {
+      jsonText = jsonText.slice(3, -3).trim();
+    }
+    const parsed = JSON.parse(jsonText);
+    const fullSummary = `Chapter ${chapterIndex + 1}: ${title}
+
+Characters: ${parsed.characters.join(", ")}
+
+Events: ${parsed.events}
+
+Setting: ${parsed.setting}
+
+Revelations: ${parsed.revelations}`;
+    return {
+      chapterIndex,
+      chapterTitle: title,
+      characters: parsed.characters,
+      events: parsed.events,
+      setting: parsed.setting,
+      revelations: parsed.revelations,
+      fullSummary
+    };
+  } catch (error) {
+    logWarn(`[BatchSummarizer] Failed to parse summary JSON for "${title}": ${error instanceof Error ? error.message : String(error)}`);
+    return null;
+  }
+};
+var downloadBatchSummaryResults = async (outputFileId, chapters, metadata) => {
+  const client = new OpenAI();
+  logInfo(`[BatchSummarizer] Downloading results from ${outputFileId}`);
+  const response = await client.files.content(outputFileId);
+  const text = await response.text();
+  const lines = text.trim().split("\n");
+  const results = /* @__PURE__ */ new Map();
+  for (const line of lines) {
+    const result = JSON.parse(line);
+    if (result.response?.status_code === 200) {
+      const content = result.response.body?.choices?.[0]?.message?.content;
+      if (content) {
+        results.set(result.custom_id, content);
+      }
+    } else {
+      logWarn(`[BatchSummarizer] Request ${result.custom_id} failed: ${JSON.stringify(result.response?.body?.error ?? result.error)}`);
+    }
+  }
+  const summaries = [];
+  const needsMergePass = [];
+  for (const meta of metadata) {
+    if (!meta.needsTwoPass) {
+      const content = results.get(`summary-${meta.chapterIndex}`);
+      if (content) {
+        const summary = parseStructuredSummary(content, meta.chapterIndex, meta.title);
+        if (summary) summaries.push(summary);
+      }
+    } else {
+      const sectionSummaries = [];
+      let allPresent = true;
+      for (let s = 0; s < meta.sectionCount; s++) {
+        const content = results.get(`section-${meta.chapterIndex}-${s}`);
+        if (content) {
+          sectionSummaries.push(content);
+        } else {
+          allPresent = false;
+        }
+      }
+      if (allPresent && sectionSummaries.length > 0) {
+        needsMergePass.push({ chapterIndex: meta.chapterIndex, title: meta.title, sectionSummaries });
+      } else {
+        logWarn(`[BatchSummarizer] Missing section results for chapter ${meta.chapterIndex + 1} "${meta.title}"`);
+      }
+    }
+  }
+  logInfo(`[BatchSummarizer] Parsed ${summaries.length} direct summaries, ${needsMergePass.length} chapters need merge pass`);
+  return { summaries, needsMergePass };
+};
+var submitMergePass = async (mergeChapters) => {
+  const models = await getModels();
+  const client = new OpenAI();
+  const lines = [];
+  const metadata = [];
+  for (const ch of mergeChapters) {
+    const combined = ch.sectionSummaries.join("\n\n");
+    const line = {
+      custom_id: `summary-${ch.chapterIndex}`,
+      method: "POST",
+      url: "/v1/chat/completions",
+      body: {
+        model: models.summary,
+        messages: [{ role: "user", content: SUMMARY_PROMPT(ch.title, ch.chapterIndex + 1, combined) }]
+      }
+    };
+    lines.push(JSON.stringify(line));
+    metadata.push({ chapterIndex: ch.chapterIndex, title: ch.title, needsTwoPass: false, sectionCount: 1 });
+  }
+  const jsonl = lines.join("\n");
+  const blob = new Blob([jsonl], { type: "application/jsonl" });
+  const file = await client.files.create({
+    file: new File([blob], "summaries-merge.jsonl", { type: "application/jsonl" }),
+    purpose: "batch"
+  });
+  logInfo(`[BatchSummarizer] Uploaded merge input file ${file.id} (${mergeChapters.length} chapters)`);
+  const batch = await client.batches.create({
+    input_file_id: file.id,
+    endpoint: "/v1/chat/completions",
+    completion_window: "24h"
+  });
+  logInfo(`[BatchSummarizer] Created merge batch ${batch.id} \u2014 status: ${batch.status}`);
+  return { batchId: batch.id, inputFileId: file.id, metadata };
+};
+var downloadMergeResults = async (outputFileId, mergeChapters) => {
+  const client = new OpenAI();
+  logInfo(`[BatchSummarizer] Downloading merge results from ${outputFileId}`);
+  const response = await client.files.content(outputFileId);
+  const text = await response.text();
+  const lines = text.trim().split("\n");
+  const summaries = [];
+  for (const line of lines) {
+    const result = JSON.parse(line);
+    if (result.response?.status_code === 200) {
+      const content = result.response.body?.choices?.[0]?.message?.content;
+      if (content) {
+        const idx = Number(result.custom_id.replace("summary-", ""));
+        const meta = mergeChapters.find((ch) => ch.chapterIndex === idx);
+        if (meta) {
+          const summary = parseStructuredSummary(content, idx, meta.title);
+          if (summary) summaries.push(summary);
+        }
+      }
+    } else {
+      logWarn(`[BatchSummarizer] Merge request ${result.custom_id} failed: ${JSON.stringify(result.response?.body?.error ?? result.error)}`);
+    }
+  }
+  logInfo(`[BatchSummarizer] Parsed ${summaries.length} merged summaries`);
+  return summaries;
+};
+
+export {
+  submitBatchSummaries,
+  downloadBatchSummaryResults,
+  submitMergePass,
+  downloadMergeResults
+};
+//# sourceMappingURL=chunk-7DUQNGEK.js.map