wiki-graph-core 0.3.1 → 0.4.0

package/dist/document/text-streams/index.js.map

@@ -0,0 +1 @@
+{"version":3,"sources":["../../../src/document/text-streams/index.ts"],"sourcesContent":["export * from \"./core.js\";\nexport * from \"./draft.js\";\nexport * from \"./serial.js\";\nexport * from \"./types.js\";\n"],"mappings":"AAAA,cAAc;AACd,cAAc;AACd,cAAc;AACd,cAAc;","names":[]}

package/dist/document/text-streams/sentence.d.ts

@@ -0,0 +1,13 @@
+import { SentenceRecord } from '../types.js';
+import { TextStreamSentenceSegmenter } from './types.js';
+
+declare function splitTextIntoSentenceSpans(text: string, segmenter: TextStreamSentenceSegmenter | undefined): Promise<ReadonlyArray<SentenceRecord & {
+    readonly byteOffset: number;
+    readonly byteLength: number;
+}>>;
+declare function getSentenceByteOffset(sentence: SentenceRecord): number;
+declare function hasSentenceByteOffset(sentence: SentenceRecord): boolean;
+declare function getSentenceByteLength(sentence: SentenceRecord): number;
+declare function getSentenceRawText(sentence: SentenceRecord): string;
+
+export { getSentenceByteLength, getSentenceByteOffset, getSentenceRawText, hasSentenceByteOffset, splitTextIntoSentenceSpans };

package/dist/document/text-streams/sentence.js

@@ -0,0 +1,77 @@
+import { Sentence } from "../types.js";
+async function splitTextIntoSentenceSpans(text, segmenter) {
+  if (segmenter !== void 0) {
+    return await splitTextIntoCustomSentenceSpans(text, segmenter);
+  }
+  const spans = [];
+  for (const segment of createSentenceSegmenter().segment(text)) {
+    const rawText = segment.segment;
+    if (rawText.trim() === "") {
+      continue;
+    }
+    const sentence = new Sentence(rawText, countWords(rawText));
+    Object.assign(sentence, {
+      byteLength: Buffer.byteLength(rawText, "utf8"),
+      byteOffset: Buffer.byteLength(text.slice(0, segment.index), "utf8")
+    });
+    spans.push(
+      sentence
+    );
+  }
+  return spans;
+}
+async function splitTextIntoCustomSentenceSpans(text, segmenter) {
+  const spans = [];
+  for await (const segment of segmenter.pipe([text])) {
+    const rawText = text.slice(
+      segment.offset,
+      segment.offset + segment.text.length
+    );
+    if (rawText.trim() === "") {
+      continue;
+    }
+    const sentence = new Sentence(rawText, segment.wordsCount);
+    Object.assign(sentence, {
+      byteLength: Buffer.byteLength(rawText, "utf8"),
+      byteOffset: Buffer.byteLength(text.slice(0, segment.offset), "utf8")
+    });
+    spans.push(
+      sentence
+    );
+  }
+  return spans;
+}
+let SENTENCE_SEGMENTER;
+function createSentenceSegmenter() {
+  SENTENCE_SEGMENTER ??= new Intl.Segmenter(void 0, {
+    granularity: "sentence"
+  });
+  return SENTENCE_SEGMENTER;
+}
+function countWords(text) {
+  const trimmed = text.trim();
+  return trimmed === "" ? 0 : trimmed.split(/\s+/u).length;
+}
+function getSentenceByteOffset(sentence) {
+  const value = sentence.byteOffset;
+  return typeof value === "number" ? value : 0;
+}
+function hasSentenceByteOffset(sentence) {
+  return typeof sentence.byteOffset === "number";
+}
+function getSentenceByteLength(sentence) {
+  const value = sentence.byteLength;
+  return typeof value === "number" ? value : Buffer.byteLength(getSentenceRawText(sentence), "utf8");
+}
+function getSentenceRawText(sentence) {
+  const value = sentence.rawText;
+  return typeof value === "string" ? value : sentence.text;
+}
+export {
+  getSentenceByteLength,
+  getSentenceByteOffset,
+  getSentenceRawText,
+  hasSentenceByteOffset,
+  splitTextIntoSentenceSpans
+};
+//# sourceMappingURL=sentence.js.map

package/dist/document/text-streams/sentence.js.map

@@ -0,0 +1 @@
+{"version":3,"sources":["../../../src/document/text-streams/sentence.ts"],"sourcesContent":["import { Sentence, type SentenceRecord } from \"../types.js\";\nimport type { TextStreamSentenceSegmenter } from \"./types.js\";\n\nexport async function splitTextIntoSentenceSpans(\n  text: string,\n  segmenter: TextStreamSentenceSegmenter | undefined,\n): Promise<\n  ReadonlyArray<\n    SentenceRecord & {\n      readonly byteOffset: number;\n      readonly byteLength: number;\n    }\n  >\n> {\n  if (segmenter !== undefined) {\n    return await splitTextIntoCustomSentenceSpans(text, segmenter);\n  }\n\n  const spans: Array<\n    SentenceRecord & {\n      readonly byteOffset: number;\n      readonly byteLength: number;\n    }\n  > = [];\n\n  for (const segment of createSentenceSegmenter().segment(text)) {\n    const rawText = segment.segment;\n\n    if (rawText.trim() === \"\") {\n      continue;\n    }\n    const sentence = new Sentence(rawText, countWords(rawText));\n\n    Object.assign(sentence, {\n      byteLength: Buffer.byteLength(rawText, \"utf8\"),\n      byteOffset: Buffer.byteLength(text.slice(0, segment.index), \"utf8\"),\n    });\n    spans.push(\n      sentence as unknown as SentenceRecord & {\n        readonly byteOffset: number;\n        readonly byteLength: number;\n      },\n    );\n  }\n\n  return spans;\n}\n\nasync function splitTextIntoCustomSentenceSpans(\n  text: string,\n  segmenter: TextStreamSentenceSegmenter,\n): Promise<\n  ReadonlyArray<\n    SentenceRecord & {\n      readonly byteOffset: number;\n      readonly byteLength: number;\n    }\n  >\n> {\n  const spans: Array<\n    SentenceRecord & {\n      readonly byteOffset: number;\n      readonly byteLength: number;\n    }\n  > = [];\n\n  for await (const segment of segmenter.pipe([text])) {\n    const rawText = text.slice(\n      segment.offset,\n      segment.offset + segment.text.length,\n    );\n\n    if (rawText.trim() === \"\") {\n      continue;\n    }\n    const sentence = new Sentence(rawText, segment.wordsCount);\n\n    Object.assign(sentence, {\n      byteLength: Buffer.byteLength(rawText, \"utf8\"),\n      byteOffset: Buffer.byteLength(text.slice(0, segment.offset), \"utf8\"),\n    });\n    spans.push(\n      sentence as unknown as SentenceRecord & {\n        readonly byteOffset: number;\n        readonly byteLength: number;\n      },\n    );\n  }\n\n  return spans;\n}\n\nlet SENTENCE_SEGMENTER: Intl.Segmenter | undefined;\n\nfunction createSentenceSegmenter(): Intl.Segmenter {\n  SENTENCE_SEGMENTER ??= new Intl.Segmenter(undefined, {\n    granularity: \"sentence\",\n  });\n\n  return SENTENCE_SEGMENTER;\n}\n\nfunction countWords(text: string): number {\n  const trimmed = text.trim();\n\n  return trimmed === \"\" ? 0 : trimmed.split(/\\s+/u).length;\n}\n\nexport function getSentenceByteOffset(sentence: SentenceRecord): number {\n  const value = (sentence as { readonly byteOffset?: unknown }).byteOffset;\n\n  return typeof value === \"number\" ? value : 0;\n}\n\nexport function hasSentenceByteOffset(sentence: SentenceRecord): boolean {\n  return (\n    typeof (sentence as { readonly byteOffset?: unknown }).byteOffset ===\n    \"number\"\n  );\n}\n\nexport function getSentenceByteLength(sentence: SentenceRecord): number {\n  const value = (sentence as { readonly byteLength?: unknown }).byteLength;\n\n  return typeof value === \"number\"\n    ? value\n    : Buffer.byteLength(getSentenceRawText(sentence), \"utf8\");\n}\n\nexport function getSentenceRawText(sentence: SentenceRecord): string {\n  const value = (sentence as { readonly rawText?: unknown }).rawText;\n\n  return typeof value === \"string\" ? value : sentence.text;\n}\n"],"mappings":"AAAA,SAAS,gBAAqC;AAG9C,eAAsB,2BACpB,MACA,WAQA;AACA,MAAI,cAAc,QAAW;AAC3B,WAAO,MAAM,iCAAiC,MAAM,SAAS;AAAA,EAC/D;AAEA,QAAM,QAKF,CAAC;AAEL,aAAW,WAAW,wBAAwB,EAAE,QAAQ,IAAI,GAAG;AAC7D,UAAM,UAAU,QAAQ;AAExB,QAAI,QAAQ,KAAK,MAAM,IAAI;AACzB;AAAA,IACF;AACA,UAAM,WAAW,IAAI,SAAS,SAAS,WAAW,OAAO,CAAC;AAE1D,WAAO,OAAO,UAAU;AAAA,MACtB,YAAY,OAAO,WAAW,SAAS,MAAM;AAAA,MAC7C,YAAY,OAAO,WAAW,KAAK,MAAM,GAAG,QAAQ,KAAK,GAAG,MAAM;AAAA,IACpE,CAAC;AACD,UAAM;AAAA,MACJ;AAAA,IAIF;AAAA,EACF;AAEA,SAAO;AACT;AAEA,eAAe,iCACb,MACA,WAQA;AACA,QAAM,QAKF,CAAC;AAEL,mBAAiB,WAAW,UAAU,KAAK,CAAC,IAAI,CAAC,GAAG;AAClD,UAAM,UAAU,KAAK;AAAA,MACnB,QAAQ;AAAA,MACR,QAAQ,SAAS,QAAQ,KAAK;AAAA,IAChC;AAEA,QAAI,QAAQ,KAAK,MAAM,IAAI;AACzB;AAAA,IACF;AACA,UAAM,WAAW,IAAI,SAAS,SAAS,QAAQ,UAAU;AAEzD,WAAO,OAAO,UAAU;AAAA,MACtB,YAAY,OAAO,WAAW,SAAS,MAAM;AAAA,MAC7C,YAAY,OAAO,WAAW,KAAK,MAAM,GAAG,QAAQ,MAAM,GAAG,MAAM;AAAA,IACrE,CAAC;AACD,UAAM;AAAA,MACJ;AAAA,IAIF;AAAA,EACF;AAEA,SAAO;AACT;AAEA,IAAI;AAEJ,SAAS,0BAA0C;AACjD,yBAAuB,IAAI,KAAK,UAAU,QAAW;AAAA,IACnD,aAAa;AAAA,EACf,CAAC;AAED,SAAO;AACT;AAEA,SAAS,WAAW,MAAsB;AACxC,QAAM,UAAU,KAAK,KAAK;AAE1B,SAAO,YAAY,KAAK,IAAI,QAAQ,MAAM,MAAM,EAAE;AACpD;AAEO,SAAS,sBAAsB,UAAkC;AACtE,QAAM,QAAS,SAA+C;AAE9D,SAAO,OAAO,UAAU,WAAW,QAAQ;AAC7C;AAEO,SAAS,sBAAsB,UAAmC;AACvE,SACE,OAAQ,SAA+C,eACvD;AAEJ;AAEO,SAAS,sBAAsB,UAAkC;AACtE,QAAM,QAAS,SAA+C;AAE9D,SAAO,OAAO,UAAU,WACpB,QACA,OAAO,WAAW,mBAAmB,QAAQ,GAAG,MAAM;AAC5D;AAEO,SAAS,mBAAmB,UAAkC;AACnE,QAAM,QAAS,SAA4C;AAE3D,SAAO,OAAO,UAAU,WAAW,QAAQ,SAAS;AACtD;","names":[]}

package/dist/document/text-streams/serial.d.ts

@@ -0,0 +1,24 @@
+import { Database } from '../database.js';
+import { FragmentRecord, SentenceRecord } from '../types.js';
+import { TextStreamDraft } from './draft.js';
+import { ReadonlySerialTextStream, TextStreamFileAccess, TextStreamName, WriteTextStreamOptions } from './types.js';
+import 'sqlite3';
+
+declare class SerialTextStream implements ReadonlySerialTextStream {
+    #private;
+    constructor(documentPath: string, database: Database, fileAccess: TextStreamFileAccess, stream: TextStreamName, serialId: number);
+    createDraft(): Promise<TextStreamDraft>;
+    getFragment(fragmentId: number): Promise<FragmentRecord>;
+    getSentence(sentenceIndex: number): Promise<SentenceRecord>;
+    listFragmentIds(): Promise<readonly number[]>;
+    listSentences(): Promise<readonly SentenceRecord[]>;
+    listSentencesInRange(startSentenceIndex: number, endSentenceIndex: number): Promise<readonly SentenceRecord[]>;
+    readTextInRange(startSentenceIndex: number, endSentenceIndex: number): Promise<string | undefined>;
+    readText(): Promise<string | undefined>;
+    writeTextStream(text: string, options?: WriteTextStreamOptions): Promise<void>;
+    delete(): Promise<void>;
+    get path(): string;
+    get serialId(): number;
+}
+
+export { SerialTextStream };

package/dist/document/text-streams/serial.js

@@ -0,0 +1,324 @@
+import { join, resolve } from "path";
+import {
+  Sentence
+} from "../types.js";
+import { TextStreamDraft } from "./draft.js";
+import {
+  getSentenceByteLength,
+  getSentenceByteOffset,
+  getSentenceRawText,
+  hasSentenceByteOffset,
+  splitTextIntoSentenceSpans
+} from "./sentence.js";
+import {
+  DEFAULT_FRAGMENT_WORDS_COUNT,
+  TEXT_STREAM_KIND
+} from "./types.js";
+class SerialTextStream {
+  static #draftStates = /* @__PURE__ */ new Map();
+  #database;
+  #documentPath;
+  #fileAccess;
+  #stream;
+  #serialId;
+  constructor(documentPath, database, fileAccess, stream, serialId) {
+    this.#database = database;
+    this.#documentPath = resolve(documentPath);
+    this.#fileAccess = fileAccess;
+    this.#stream = stream;
+    this.#serialId = serialId;
+  }
+  async createDraft() {
+    const draftState = this.#getDraftState();
+    if (draftState.draftOpen) {
+      throw new Error("Only one text stream draft can be open at a time");
+    }
+    await this.#fileAccess.ensureDirectory(this.#getDirectoryPath());
+    draftState.draftOpen = true;
+    return new TextStreamDraft(this.#serialId, await this.#peekNextIndex(), {
+      discard: () => {
+        draftState.draftOpen = false;
+      },
+      finalize: async (startIndex, summary, sentences) => await this.#commitDraft(startIndex, summary, sentences)
+    });
+  }
+  async getFragment(fragmentId) {
+    const fragments = await this.#listFragments();
+    const fragment = fragments.find((item) => item.fragmentId === fragmentId);
+    if (fragment === void 0) {
+      throw new Error(`Fragment ${fragmentId} does not exist`);
+    }
+    return fragment;
+  }
+  async getSentence(sentenceIndex) {
+    const location = await this.#getSentenceLocation(sentenceIndex);
+    if (location === void 0) {
+      throw new RangeError(`Sentence ${sentenceIndex} does not exist`);
+    }
+    return this.#readSentenceLocation(location, await this.#readContent());
+  }
+  async listFragmentIds() {
+    return (await this.#listFragments()).map((fragment) => fragment.fragmentId);
+  }
+  async listSentences() {
+    const rows = await this.#listSentenceLocations();
+    const content = await this.#readContent();
+    return rows.map((row) => this.#readSentenceLocation(row, content));
+  }
+  async listSentencesInRange(startSentenceIndex, endSentenceIndex) {
+    if (endSentenceIndex < startSentenceIndex) {
+      return [];
+    }
+    const rows = await this.#database.queryAll(
+      `
+        SELECT sentence_index, byte_offset, byte_length, words_count
+        FROM text_sentence_records
+        WHERE kind = ? AND chapter_id = ?
+          AND sentence_index BETWEEN ? AND ?
+        ORDER BY sentence_index
+      `,
+      [
+        TEXT_STREAM_KIND[this.#stream],
+        this.#serialId,
+        startSentenceIndex,
+        endSentenceIndex
+      ],
+      mapTextSentenceLocation
+    );
+    const content = await this.#readContent();
+    return rows.map((row) => this.#readSentenceLocation(row, content));
+  }
+  async readTextInRange(startSentenceIndex, endSentenceIndex) {
+    if (endSentenceIndex < startSentenceIndex) {
+      return "";
+    }
+    const rows = await this.#database.queryAll(
+      `
+        SELECT sentence_index, byte_offset, byte_length, words_count
+        FROM text_sentence_records
+        WHERE kind = ? AND chapter_id = ?
+          AND sentence_index BETWEEN ? AND ?
+        ORDER BY sentence_index
+      `,
+      [
+        TEXT_STREAM_KIND[this.#stream],
+        this.#serialId,
+        startSentenceIndex,
+        endSentenceIndex
+      ],
+      mapTextSentenceLocation
+    );
+    const first = rows[0];
+    const last = rows[rows.length - 1];
+    if (first === void 0 || last === void 0) {
+      return void 0;
+    }
+    const content = await this.#readContent();
+    return content.subarray(first.byteOffset, last.byteOffset + last.byteLength).toString("utf8");
+  }
+  async #getSentenceLocation(sentenceIndex) {
+    return await this.#database.queryOne(
+      `
+        SELECT sentence_index, byte_offset, byte_length, words_count
+        FROM text_sentence_records
+        WHERE kind = ? AND chapter_id = ? AND sentence_index = ?
+      `,
+      [TEXT_STREAM_KIND[this.#stream], this.#serialId, sentenceIndex],
+      mapTextSentenceLocation
+    );
+  }
+  async #listSentenceLocations() {
+    return await this.#database.queryAll(
+      `
+        SELECT sentence_index, byte_offset, byte_length, words_count
+        FROM text_sentence_records
+        WHERE kind = ? AND chapter_id = ?
+        ORDER BY sentence_index
+      `,
+      [TEXT_STREAM_KIND[this.#stream], this.#serialId],
+      mapTextSentenceLocation
+    );
+  }
+  #readSentenceLocation(location, content) {
+    return new Sentence(
+      content.subarray(
+        location.byteOffset,
+        location.byteOffset + location.byteLength
+      ).toString("utf8"),
+      location.wordsCount
+    );
+  }
+  async readText() {
+    const content = await this.#fileAccess.readFile(this.#getTextPath());
+    return content === void 0 ? void 0 : Buffer.from(content).toString("utf8");
+  }
+  async writeTextStream(text, options = {}) {
+    await this.delete();
+    const sentences = await splitTextIntoSentenceSpans(text, options.segmenter);
+    const draft = await this.createDraft();
+    for (const sentence of sentences) {
+      draft.addSentence(sentence.text, sentence.wordsCount, {
+        byteOffset: sentence.byteOffset,
+        byteLength: sentence.byteLength
+      });
+    }
+    await draft.commitWithText(text);
+  }
+  async delete() {
+    await this.#fileAccess.deleteTree(this.#getTextPath());
+    await this.#database.run(
+      `
+        DELETE FROM text_sentence_records
+        WHERE kind = ? AND chapter_id = ?
+      `,
+      [TEXT_STREAM_KIND[this.#stream], this.#serialId]
+    );
+    delete this.#getDraftState().nextSentenceIndex;
+  }
+  get path() {
+    return this.#getTextPath();
+  }
+  get serialId() {
+    return this.#serialId;
+  }
+  async #commitDraft(startIndex, textOverride, sentences) {
+    const draftState = this.#getDraftState();
+    draftState.draftOpen = false;
+    const existing = await this.#fileAccess.readFile(this.#getTextPath());
+    const existingBuffer = existing === void 0 ? Buffer.alloc(0) : Buffer.from(existing);
+    const text = textOverride === "" ? sentences.map(getSentenceRawText).join("") : textOverride;
+    const appendBuffer = Buffer.from(text, "utf8");
+    let offset = existingBuffer.length;
+    await this.#fileAccess.ensureDirectory(this.#getDirectoryPath());
+    await this.#fileAccess.writeFile(
+      this.#getTextPath(),
+      Buffer.concat([existingBuffer, appendBuffer]),
+      { overwrite: true }
+    );
+    if (sentences.length === 0) {
+      return void 0;
+    }
+    for (let index = 0; index < sentences.length; index += 1) {
+      const sentence = sentences[index];
+      if (sentence === void 0) {
+        continue;
+      }
+      const length = getSentenceByteLength(sentence);
+      const explicitOffset = getSentenceByteOffset(sentence);
+      const sentenceOffset = offset + explicitOffset;
+      await this.#database.run(
+        `
+          INSERT OR REPLACE INTO text_sentence_records (
+            kind,
+            chapter_id,
+            sentence_index,
+            words_count,
+            byte_offset,
+            byte_length
+          )
+          VALUES (?, ?, ?, ?, ?, ?)
+        `,
+        [
+          TEXT_STREAM_KIND[this.#stream],
+          this.#serialId,
+          startIndex + index,
+          sentence.wordsCount,
+          sentenceOffset,
+          length
+        ]
+      );
+      if (!hasSentenceByteOffset(sentence)) {
+        offset += length;
+      }
+    }
+    draftState.nextSentenceIndex = startIndex + sentences.length;
+    return {
+      fragmentId: startIndex,
+      sentences,
+      serialId: this.#serialId,
+      summary: ""
+    };
+  }
+  async #listFragments() {
+    const sentences = await this.listSentences();
+    const fragments = [];
+    let current = [];
+    let currentWords = 0;
+    let fragmentStart = 0;
+    for (let index = 0; index < sentences.length; index += 1) {
+      const sentence = sentences[index];
+      if (sentence === void 0) {
+        continue;
+      }
+      if (current.length > 0 && currentWords + sentence.wordsCount > DEFAULT_FRAGMENT_WORDS_COUNT) {
+        fragments.push({
+          fragmentId: fragmentStart,
+          sentences: current,
+          serialId: this.#serialId,
+          summary: ""
+        });
+        fragmentStart = index;
+        current = [];
+        currentWords = 0;
+      }
+      current.push(sentence);
+      currentWords += sentence.wordsCount;
+    }
+    if (current.length > 0) {
+      fragments.push({
+        fragmentId: fragmentStart,
+        sentences: current,
+        serialId: this.#serialId,
+        summary: ""
+      });
+    }
+    return fragments;
+  }
+  async #peekNextIndex() {
+    const draftState = this.#getDraftState();
+    if (draftState.nextSentenceIndex !== void 0) {
+      return draftState.nextSentenceIndex;
+    }
+    draftState.nextSentenceIndex = await this.#database.queryOne(
+      `
+          SELECT COALESCE(MAX(sentence_index), -1) + 1 AS next_index
+          FROM text_sentence_records
+          WHERE kind = ? AND chapter_id = ?
+        `,
+      [TEXT_STREAM_KIND[this.#stream], this.#serialId],
+      (row) => Number(row.next_index)
+    ) ?? 0;
+    return draftState.nextSentenceIndex;
+  }
+  async #readContent() {
+    const content = await this.#fileAccess.readFile(this.#getTextPath());
+    return Buffer.from(content ?? new Uint8Array());
+  }
+  #getDirectoryPath() {
+    return join(this.#documentPath, "texts", this.#stream);
+  }
+  #getTextPath() {
+    return join(this.#getDirectoryPath(), `${this.#serialId}.txt`);
+  }
+  #getDraftState() {
+    const key = `${this.#documentPath}\0${this.#stream}\0${this.#serialId}`;
+    let state = SerialTextStream.#draftStates.get(key);
+    if (state === void 0) {
+      state = { draftOpen: false };
+      SerialTextStream.#draftStates.set(key, state);
+    }
+    return state;
+  }
+}
+function mapTextSentenceLocation(row) {
+  return {
+    byteLength: Number(row.byte_length),
+    byteOffset: Number(row.byte_offset),
+    sentenceIndex: Number(row.sentence_index),
+    wordsCount: Number(row.words_count)
+  };
+}
+export {
+  SerialTextStream
+};
+//# sourceMappingURL=serial.js.map

package/dist/document/text-streams/serial.js.map

@@ -0,0 +1 @@
+{"version":3,"sources":["../../../src/document/text-streams/serial.ts"],"sourcesContent":["import { join, resolve } from \"path\";\n\nimport type { Database } from \"../database.js\";\nimport {\n  Sentence,\n  type FragmentRecord,\n  type SentenceRecord,\n} from \"../types.js\";\nimport { TextStreamDraft } from \"./draft.js\";\nimport {\n  getSentenceByteLength,\n  getSentenceByteOffset,\n  getSentenceRawText,\n  hasSentenceByteOffset,\n  splitTextIntoSentenceSpans,\n} from \"./sentence.js\";\nimport {\n  DEFAULT_FRAGMENT_WORDS_COUNT,\n  TEXT_STREAM_KIND,\n  type ReadonlySerialTextStream,\n  type TextSentenceLocation,\n  type TextStreamDraftState,\n  type TextStreamFileAccess,\n  type TextStreamName,\n  type WriteTextStreamOptions,\n} from \"./types.js\";\n\nexport class SerialTextStream implements ReadonlySerialTextStream {\n  static readonly #draftStates = new Map<string, TextStreamDraftState>();\n\n  readonly #database: Database;\n  readonly #documentPath: string;\n  readonly #fileAccess: TextStreamFileAccess;\n  readonly #stream: TextStreamName;\n  readonly #serialId: number;\n\n  public constructor(\n    documentPath: string,\n    database: Database,\n    fileAccess: TextStreamFileAccess,\n    stream: TextStreamName,\n    serialId: number,\n  ) {\n    this.#database = database;\n    this.#documentPath = resolve(documentPath);\n    this.#fileAccess = fileAccess;\n    this.#stream = stream;\n    this.#serialId = serialId;\n  }\n\n  public async createDraft(): Promise<TextStreamDraft> {\n    const draftState = this.#getDraftState();\n\n    if (draftState.draftOpen) {\n      throw new Error(\"Only one text stream draft can be open at a time\");\n    }\n\n    await this.#fileAccess.ensureDirectory(this.#getDirectoryPath());\n    draftState.draftOpen = true;\n\n    return new TextStreamDraft(this.#serialId, await this.#peekNextIndex(), {\n      discard: () => {\n        draftState.draftOpen = false;\n      },\n      finalize: async (startIndex, summary, sentences) =>\n        await this.#commitDraft(startIndex, summary, sentences),\n    });\n  }\n\n  public async getFragment(fragmentId: number): Promise<FragmentRecord> {\n    const fragments = await this.#listFragments();\n    const fragment = fragments.find((item) => item.fragmentId === fragmentId);\n\n    if (fragment === undefined) {\n      throw new Error(`Fragment ${fragmentId} does not exist`);\n    }\n\n    return fragment;\n  }\n\n  public async getSentence(sentenceIndex: number): Promise<SentenceRecord> {\n    const location = await this.#getSentenceLocation(sentenceIndex);\n\n    if (location === undefined) {\n      throw new RangeError(`Sentence ${sentenceIndex} does not exist`);\n    }\n\n    return this.#readSentenceLocation(location, await this.#readContent());\n  }\n\n  public async listFragmentIds(): Promise<readonly number[]> {\n    return (await this.#listFragments()).map((fragment) => fragment.fragmentId);\n  }\n\n  public async listSentences(): Promise<readonly SentenceRecord[]> {\n    const rows = await this.#listSentenceLocations();\n    const content = await this.#readContent();\n\n    return rows.map((row) => this.#readSentenceLocation(row, content));\n  }\n\n  public async listSentencesInRange(\n    startSentenceIndex: number,\n    endSentenceIndex: number,\n  ): Promise<readonly SentenceRecord[]> {\n    if (endSentenceIndex < startSentenceIndex) {\n      return [];\n    }\n\n    const rows = await this.#database.queryAll(\n      `\n        SELECT sentence_index, byte_offset, byte_length, words_count\n        FROM text_sentence_records\n        WHERE kind = ? AND chapter_id = ?\n          AND sentence_index BETWEEN ? AND ?\n        ORDER BY sentence_index\n      `,\n      [\n        TEXT_STREAM_KIND[this.#stream],\n        this.#serialId,\n        startSentenceIndex,\n        endSentenceIndex,\n      ],\n      mapTextSentenceLocation,\n    );\n    const content = await this.#readContent();\n\n    return rows.map((row) => this.#readSentenceLocation(row, content));\n  }\n\n  public async readTextInRange(\n    startSentenceIndex: number,\n    endSentenceIndex: number,\n  ): Promise<string | undefined> {\n    if (endSentenceIndex < startSentenceIndex) {\n      return \"\";\n    }\n\n    const rows = await this.#database.queryAll(\n      `\n        SELECT sentence_index, byte_offset, byte_length, words_count\n        FROM text_sentence_records\n        WHERE kind = ? AND chapter_id = ?\n          AND sentence_index BETWEEN ? AND ?\n        ORDER BY sentence_index\n      `,\n      [\n        TEXT_STREAM_KIND[this.#stream],\n        this.#serialId,\n        startSentenceIndex,\n        endSentenceIndex,\n      ],\n      mapTextSentenceLocation,\n    );\n    const first = rows[0];\n    const last = rows[rows.length - 1];\n\n    if (first === undefined || last === undefined) {\n      return undefined;\n    }\n\n    const content = await this.#readContent();\n\n    return content\n      .subarray(first.byteOffset, last.byteOffset + last.byteLength)\n      .toString(\"utf8\");\n  }\n\n  async #getSentenceLocation(\n    sentenceIndex: number,\n  ): Promise<TextSentenceLocation | undefined> {\n    return await this.#database.queryOne(\n      `\n        SELECT sentence_index, byte_offset, byte_length, words_count\n        FROM text_sentence_records\n        WHERE kind = ? AND chapter_id = ? AND sentence_index = ?\n      `,\n      [TEXT_STREAM_KIND[this.#stream], this.#serialId, sentenceIndex],\n      mapTextSentenceLocation,\n    );\n  }\n\n  async #listSentenceLocations(): Promise<readonly TextSentenceLocation[]> {\n    return await this.#database.queryAll(\n      `\n        SELECT sentence_index, byte_offset, byte_length, words_count\n        FROM text_sentence_records\n        WHERE kind = ? AND chapter_id = ?\n        ORDER BY sentence_index\n      `,\n      [TEXT_STREAM_KIND[this.#stream], this.#serialId],\n      mapTextSentenceLocation,\n    );\n  }\n\n  #readSentenceLocation(\n    location: TextSentenceLocation,\n    content: Buffer,\n  ): SentenceRecord {\n    return new Sentence(\n      content\n        .subarray(\n          location.byteOffset,\n          location.byteOffset + location.byteLength,\n        )\n        .toString(\"utf8\"),\n      location.wordsCount,\n    );\n  }\n\n  public async readText(): Promise<string | undefined> {\n    const content = await this.#fileAccess.readFile(this.#getTextPath());\n\n    return content === undefined\n      ? undefined\n      : Buffer.from(content).toString(\"utf8\");\n  }\n\n  public async writeTextStream(\n    text: string,\n    options: WriteTextStreamOptions = {},\n  ): Promise<void> {\n    await this.delete();\n    const sentences = await splitTextIntoSentenceSpans(text, options.segmenter);\n    const draft = await this.createDraft();\n\n    for (const sentence of sentences) {\n      draft.addSentence(sentence.text, sentence.wordsCount, {\n        byteOffset: sentence.byteOffset,\n        byteLength: sentence.byteLength,\n      });\n    }\n\n    await draft.commitWithText(text);\n  }\n\n  public async delete(): Promise<void> {\n    await this.#fileAccess.deleteTree(this.#getTextPath());\n    await this.#database.run(\n      `\n        DELETE FROM text_sentence_records\n        WHERE kind = ? AND chapter_id = ?\n      `,\n      [TEXT_STREAM_KIND[this.#stream], this.#serialId],\n    );\n    delete this.#getDraftState().nextSentenceIndex;\n  }\n\n  public get path(): string {\n    return this.#getTextPath();\n  }\n\n  public get serialId(): number {\n    return this.#serialId;\n  }\n\n  async #commitDraft(\n    startIndex: number,\n    textOverride: string,\n    sentences: readonly SentenceRecord[],\n  ): Promise<FragmentRecord | undefined> {\n    const draftState = this.#getDraftState();\n\n    draftState.draftOpen = false;\n\n    const existing = await this.#fileAccess.readFile(this.#getTextPath());\n    const existingBuffer =\n      existing === undefined ? Buffer.alloc(0) : Buffer.from(existing);\n    const text =\n      textOverride === \"\"\n        ? sentences.map(getSentenceRawText).join(\"\")\n        : textOverride;\n    const appendBuffer = Buffer.from(text, \"utf8\");\n    let offset = existingBuffer.length;\n\n    await this.#fileAccess.ensureDirectory(this.#getDirectoryPath());\n    await this.#fileAccess.writeFile(\n      this.#getTextPath(),\n      Buffer.concat([existingBuffer, appendBuffer]),\n      { overwrite: true },\n    );\n\n    if (sentences.length === 0) {\n      return undefined;\n    }\n\n    for (let index = 0; index < sentences.length; index += 1) {\n      const sentence = sentences[index];\n\n      if (sentence === undefined) {\n        continue;\n      }\n\n      const length = getSentenceByteLength(sentence);\n      const explicitOffset = getSentenceByteOffset(sentence);\n      const sentenceOffset = offset + explicitOffset;\n\n      await this.#database.run(\n        `\n          INSERT OR REPLACE INTO text_sentence_records (\n            kind,\n            chapter_id,\n            sentence_index,\n            words_count,\n            byte_offset,\n            byte_length\n          )\n          VALUES (?, ?, ?, ?, ?, ?)\n        `,\n        [\n          TEXT_STREAM_KIND[this.#stream],\n          this.#serialId,\n          startIndex + index,\n          sentence.wordsCount,\n          sentenceOffset,\n          length,\n        ],\n      );\n      if (!hasSentenceByteOffset(sentence)) {\n        offset += length;\n      }\n    }\n\n    draftState.nextSentenceIndex = startIndex + sentences.length;\n\n    return {\n      fragmentId: startIndex,\n      sentences,\n      serialId: this.#serialId,\n      summary: \"\",\n    };\n  }\n\n  async #listFragments(): Promise<readonly FragmentRecord[]> {\n    const sentences = await this.listSentences();\n    const fragments: FragmentRecord[] = [];\n    let current: SentenceRecord[] = [];\n    let currentWords = 0;\n    let fragmentStart = 0;\n\n    for (let index = 0; index < sentences.length; index += 1) {\n      const sentence = sentences[index];\n\n      if (sentence === undefined) {\n        continue;\n      }\n      if (\n        current.length > 0 &&\n        currentWords + sentence.wordsCount > DEFAULT_FRAGMENT_WORDS_COUNT\n      ) {\n        fragments.push({\n          fragmentId: fragmentStart,\n          sentences: current,\n          serialId: this.#serialId,\n          summary: \"\",\n        });\n        fragmentStart = index;\n        current = [];\n        currentWords = 0;\n      }\n\n      current.push(sentence);\n      currentWords += sentence.wordsCount;\n    }\n\n    if (current.length > 0) {\n      fragments.push({\n        fragmentId: fragmentStart,\n        sentences: current,\n        serialId: this.#serialId,\n        summary: \"\",\n      });\n    }\n\n    return fragments;\n  }\n\n  async #peekNextIndex(): Promise<number> {\n    const draftState = this.#getDraftState();\n\n    if (draftState.nextSentenceIndex !== undefined) {\n      return draftState.nextSentenceIndex;\n    }\n\n    draftState.nextSentenceIndex =\n      (await this.#database.queryOne(\n        `\n          SELECT COALESCE(MAX(sentence_index), -1) + 1 AS next_index\n          FROM text_sentence_records\n          WHERE kind = ? AND chapter_id = ?\n        `,\n        [TEXT_STREAM_KIND[this.#stream], this.#serialId],\n        (row) => Number(row.next_index),\n      )) ?? 0;\n\n    return draftState.nextSentenceIndex;\n  }\n\n  async #readContent(): Promise<Buffer> {\n    const content = await this.#fileAccess.readFile(this.#getTextPath());\n\n    return Buffer.from(content ?? new Uint8Array());\n  }\n\n  #getDirectoryPath(): string {\n    return join(this.#documentPath, \"texts\", this.#stream);\n  }\n\n  #getTextPath(): string {\n    return join(this.#getDirectoryPath(), `${this.#serialId}.txt`);\n  }\n\n  #getDraftState(): TextStreamDraftState {\n    const key = `${this.#documentPath}\\0${this.#stream}\\0${this.#serialId}`;\n    let state = SerialTextStream.#draftStates.get(key);\n\n    if (state === undefined) {\n      state = { draftOpen: false };\n      SerialTextStream.#draftStates.set(key, state);\n    }\n\n    return state;\n  }\n}\n\nfunction mapTextSentenceLocation(\n  row: Record<string, unknown>,\n): TextSentenceLocation {\n  return {\n    byteLength: Number(row.byte_length),\n    byteOffset: Number(row.byte_offset),\n    sentenceIndex: Number(row.sentence_index),\n    wordsCount: Number(row.words_count),\n  };\n}\n"],"mappings":"AAAA,SAAS,MAAM,eAAe;AAG9B;AAAA,EACE;AAAA,OAGK;AACP,SAAS,uBAAuB;AAChC;AAAA,EACE;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,OACK;AACP;AAAA,EACE;AAAA,EACA;AAAA,OAOK;AAEA,MAAM,iBAAqD;AAAA,EAChE,OAAgB,eAAe,oBAAI,IAAkC;AAAA,EAE5D;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EAEF,YACL,cACA,UACA,YACA,QACA,UACA;AACA,SAAK,YAAY;AACjB,SAAK,gBAAgB,QAAQ,YAAY;AACzC,SAAK,cAAc;AACnB,SAAK,UAAU;AACf,SAAK,YAAY;AAAA,EACnB;AAAA,EAEA,MAAa,cAAwC;AACnD,UAAM,aAAa,KAAK,eAAe;AAEvC,QAAI,WAAW,WAAW;AACxB,YAAM,IAAI,MAAM,kDAAkD;AAAA,IACpE;AAEA,UAAM,KAAK,YAAY,gBAAgB,KAAK,kBAAkB,CAAC;AAC/D,eAAW,YAAY;AAEvB,WAAO,IAAI,gBAAgB,KAAK,WAAW,MAAM,KAAK,eAAe,GAAG;AAAA,MACtE,SAAS,MAAM;AACb,mBAAW,YAAY;AAAA,MACzB;AAAA,MACA,UAAU,OAAO,YAAY,SAAS,cACpC,MAAM,KAAK,aAAa,YAAY,SAAS,SAAS;AAAA,IAC1D,CAAC;AAAA,EACH;AAAA,EAEA,MAAa,YAAY,YAA6C;AACpE,UAAM,YAAY,MAAM,KAAK,eAAe;AAC5C,UAAM,WAAW,UAAU,KAAK,CAAC,SAAS,KAAK,eAAe,UAAU;AAExE,QAAI,aAAa,QAAW;AAC1B,YAAM,IAAI,MAAM,YAAY,UAAU,iBAAiB;AAAA,IACzD;AAEA,WAAO;AAAA,EACT;AAAA,EAEA,MAAa,YAAY,eAAgD;AACvE,UAAM,WAAW,MAAM,KAAK,qBAAqB,aAAa;AAE9D,QAAI,aAAa,QAAW;AAC1B,YAAM,IAAI,WAAW,YAAY,aAAa,iBAAiB;AAAA,IACjE;AAEA,WAAO,KAAK,sBAAsB,UAAU,MAAM,KAAK,aAAa,CAAC;AAAA,EACvE;AAAA,EAEA,MAAa,kBAA8C;AACzD,YAAQ,MAAM,KAAK,eAAe,GAAG,IAAI,CAAC,aAAa,SAAS,UAAU;AAAA,EAC5E;AAAA,EAEA,MAAa,gBAAoD;AAC/D,UAAM,OAAO,MAAM,KAAK,uBAAuB;AAC/C,UAAM,UAAU,MAAM,KAAK,aAAa;AAExC,WAAO,KAAK,IAAI,CAAC,QAAQ,KAAK,sBAAsB,KAAK,OAAO,CAAC;AAAA,EACnE;AAAA,EAEA,MAAa,qBACX,oBACA,kBACoC;AACpC,QAAI,mBAAmB,oBAAoB;AACzC,aAAO,CAAC;AAAA,IACV;AAEA,UAAM,OAAO,MAAM,KAAK,UAAU;AAAA,MAChC;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,MAOA;AAAA,QACE,iBAAiB,KAAK,OAAO;AAAA,QAC7B,KAAK;AAAA,QACL;AAAA,QACA;AAAA,MACF;AAAA,MACA;AAAA,IACF;AACA,UAAM,UAAU,MAAM,KAAK,aAAa;AAExC,WAAO,KAAK,IAAI,CAAC,QAAQ,KAAK,sBAAsB,KAAK,OAAO,CAAC;AAAA,EACnE;AAAA,EAEA,MAAa,gBACX,oBACA,kBAC6B;AAC7B,QAAI,mBAAmB,oBAAoB;AACzC,aAAO;AAAA,IACT;AAEA,UAAM,OAAO,MAAM,KAAK,UAAU;AAAA,MAChC;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,MAOA;AAAA,QACE,iBAAiB,KAAK,OAAO;AAAA,QAC7B,KAAK;AAAA,QACL;AAAA,QACA;AAAA,MACF;AAAA,MACA;AAAA,IACF;AACA,UAAM,QAAQ,KAAK,CAAC;AACpB,UAAM,OAAO,KAAK,KAAK,SAAS,CAAC;AAEjC,QAAI,UAAU,UAAa,SAAS,QAAW;AAC7C,aAAO;AAAA,IACT;AAEA,UAAM,UAAU,MAAM,KAAK,aAAa;AAExC,WAAO,QACJ,SAAS,MAAM,YAAY,KAAK,aAAa,KAAK,UAAU,EAC5D,SAAS,MAAM;AAAA,EACpB;AAAA,EAEA,MAAM,qBACJ,eAC2C;AAC3C,WAAO,MAAM,KAAK,UAAU;AAAA,MAC1B;AAAA;AAAA;AAAA;AAAA;AAAA,MAKA,CAAC,iBAAiB,KAAK,OAAO,GAAG,KAAK,WAAW,aAAa;AAAA,MAC9D;AAAA,IACF;AAAA,EACF;AAAA,EAEA,MAAM,yBAAmE;AACvE,WAAO,MAAM,KAAK,UAAU;AAAA,MAC1B;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,MAMA,CAAC,iBAAiB,KAAK,OAAO,GAAG,KAAK,SAAS;AAAA,MAC/C;AAAA,IACF;AAAA,EACF;AAAA,EAEA,sBACE,UACA,SACgB;AAChB,WAAO,IAAI;AAAA,MACT,QACG;AAAA,QACC,SAAS;AAAA,QACT,SAAS,aAAa,SAAS;AAAA,MACjC,EACC,SAAS,MAAM;AAAA,MAClB,SAAS;AAAA,IACX;AAAA,EACF;AAAA,EAEA,MAAa,WAAwC;AACnD,UAAM,UAAU,MAAM,KAAK,YAAY,SAAS,KAAK,aAAa,CAAC;AAEnE,WAAO,YAAY,SACf,SACA,OAAO,KAAK,OAAO,EAAE,SAAS,MAAM;AAAA,EAC1C;AAAA,EAEA,MAAa,gBACX,MACA,UAAkC,CAAC,GACpB;AACf,UAAM,KAAK,OAAO;AAClB,UAAM,YAAY,MAAM,2BAA2B,MAAM,QAAQ,SAAS;AAC1E,UAAM,QAAQ,MAAM,KAAK,YAAY;AAErC,eAAW,YAAY,WAAW;AAChC,YAAM,YAAY,SAAS,MAAM,SAAS,YAAY;AAAA,QACpD,YAAY,SAAS;AAAA,QACrB,YAAY,SAAS;AAAA,MACvB,CAAC;AAAA,IACH;AAEA,UAAM,MAAM,eAAe,IAAI;AAAA,EACjC;AAAA,EAEA,MAAa,SAAwB;AACnC,UAAM,KAAK,YAAY,WAAW,KAAK,aAAa,CAAC;AACrD,UAAM,KAAK,UAAU;AAAA,MACnB;AAAA;AAAA;AAAA;AAAA,MAIA,CAAC,iBAAiB,KAAK,OAAO,GAAG,KAAK,SAAS;AAAA,IACjD;AACA,WAAO,KAAK,eAAe,EAAE;AAAA,EAC/B;AAAA,EAEA,IAAW,OAAe;AACxB,WAAO,KAAK,aAAa;AAAA,EAC3B;AAAA,EAEA,IAAW,WAAmB;AAC5B,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,MAAM,aACJ,YACA,cACA,WACqC;AACrC,UAAM,aAAa,KAAK,eAAe;AAEvC,eAAW,YAAY;AAEvB,UAAM,WAAW,MAAM,KAAK,YAAY,SAAS,KAAK,aAAa,CAAC;AACpE,UAAM,iBACJ,aAAa,SAAY,OAAO,MAAM,CAAC,IAAI,OAAO,KAAK,QAAQ;AACjE,UAAM,OACJ,iBAAiB,KACb,UAAU,IAAI,kBAAkB,EAAE,KAAK,EAAE,IACzC;AACN,UAAM,eAAe,OAAO,KAAK,MAAM,MAAM;AAC7C,QAAI,SAAS,eAAe;AAE5B,UAAM,KAAK,YAAY,gBAAgB,KAAK,kBAAkB,CAAC;AAC/D,UAAM,KAAK,YAAY;AAAA,MACrB,KAAK,aAAa;AAAA,MAClB,OAAO,OAAO,CAAC,gBAAgB,YAAY,CAAC;AAAA,MAC5C,EAAE,WAAW,KAAK;AAAA,IACpB;AAEA,QAAI,UAAU,WAAW,GAAG;AAC1B,aAAO;AAAA,IACT;AAEA,aAAS,QAAQ,GAAG,QAAQ,UAAU,QAAQ,SAAS,GAAG;AACxD,YAAM,WAAW,UAAU,KAAK;AAEhC,UAAI,aAAa,QAAW;AAC1B;AAAA,MACF;AAEA,YAAM,SAAS,sBAAsB,QAAQ;AAC7C,YAAM,iBAAiB,sBAAsB,QAAQ;AACrD,YAAM,iBAAiB,SAAS;AAEhC,YAAM,KAAK,UAAU;AAAA,QACnB;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,QAWA;AAAA,UACE,iBAAiB,KAAK,OAAO;AAAA,UAC7B,KAAK;AAAA,UACL,aAAa;AAAA,UACb,SAAS;AAAA,UACT;AAAA,UACA;AAAA,QACF;AAAA,MACF;AACA,UAAI,CAAC,sBAAsB,QAAQ,GAAG;AACpC,kBAAU;AAAA,MACZ;AAAA,IACF;AAEA,eAAW,oBAAoB,aAAa,UAAU;AAEtD,WAAO;AAAA,MACL,YAAY;AAAA,MACZ;AAAA,MACA,UAAU,KAAK;AAAA,MACf,SAAS;AAAA,IACX;AAAA,EACF;AAAA,EAEA,MAAM,iBAAqD;AACzD,UAAM,YAAY,MAAM,KAAK,cAAc;AAC3C,UAAM,YAA8B,CAAC;AACrC,QAAI,UAA4B,CAAC;AACjC,QAAI,eAAe;AACnB,QAAI,gBAAgB;AAEpB,aAAS,QAAQ,GAAG,QAAQ,UAAU,QAAQ,SAAS,GAAG;AACxD,YAAM,WAAW,UAAU,KAAK;AAEhC,UAAI,aAAa,QAAW;AAC1B;AAAA,MACF;AACA,UACE,QAAQ,SAAS,KACjB,eAAe,SAAS,aAAa,8BACrC;AACA,kBAAU,KAAK;AAAA,UACb,YAAY;AAAA,UACZ,WAAW;AAAA,UACX,UAAU,KAAK;AAAA,UACf,SAAS;AAAA,QACX,CAAC;AACD,wBAAgB;AAChB,kBAAU,CAAC;AACX,uBAAe;AAAA,MACjB;AAEA,cAAQ,KAAK,QAAQ;AACrB,sBAAgB,SAAS;AAAA,IAC3B;AAEA,QAAI,QAAQ,SAAS,GAAG;AACtB,gBAAU,KAAK;AAAA,QACb,YAAY;AAAA,QACZ,WAAW;AAAA,QACX,UAAU,KAAK;AAAA,QACf,SAAS;AAAA,MACX,CAAC;AAAA,IACH;AAEA,WAAO;AAAA,EACT;AAAA,EAEA,MAAM,iBAAkC;AACtC,UAAM,aAAa,KAAK,eAAe;AAEvC,QAAI,WAAW,sBAAsB,QAAW;AAC9C,aAAO,WAAW;AAAA,IACpB;AAEA,eAAW,oBACR,MAAM,KAAK,UAAU;AAAA,MACpB;AAAA;AAAA;AAAA;AAAA;AAAA,MAKA,CAAC,iBAAiB,KAAK,OAAO,GAAG,KAAK,SAAS;AAAA,MAC/C,CAAC,QAAQ,OAAO,IAAI,UAAU;AAAA,IAChC,KAAM;AAER,WAAO,WAAW;AAAA,EACpB;AAAA,EAEA,MAAM,eAAgC;AACpC,UAAM,UAAU,MAAM,KAAK,YAAY,SAAS,KAAK,aAAa,CAAC;AAEnE,WAAO,OAAO,KAAK,WAAW,IAAI,WAAW,CAAC;AAAA,EAChD;AAAA,EAEA,oBAA4B;AAC1B,WAAO,KAAK,KAAK,eAAe,SAAS,KAAK,OAAO;AAAA,EACvD;AAAA,EAEA,eAAuB;AACrB,WAAO,KAAK,KAAK,kBAAkB,GAAG,GAAG,KAAK,SAAS,MAAM;AAAA,EAC/D;AAAA,EAEA,iBAAuC;AACrC,UAAM,MAAM,GAAG,KAAK,aAAa,KAAK,KAAK,OAAO,KAAK,KAAK,SAAS;AACrE,QAAI,QAAQ,iBAAiB,aAAa,IAAI,GAAG;AAEjD,QAAI,UAAU,QAAW;AACvB,cAAQ,EAAE,WAAW,MAAM;AAC3B,uBAAiB,aAAa,IAAI,KAAK,KAAK;AAAA,IAC9C;AAEA,WAAO;AAAA,EACT;AACF;AAEA,SAAS,wBACP,KACsB;AACtB,SAAO;AAAA,IACL,YAAY,OAAO,IAAI,WAAW;AAAA,IAClC,YAAY,OAAO,IAAI,WAAW;AAAA,IAClC,eAAe,OAAO,IAAI,cAAc;AAAA,IACxC,YAAY,OAAO,IAAI,WAAW;AAAA,EACpC;AACF;","names":[]}

package/dist/document/text-streams/types.d.ts

@@ -0,0 +1,55 @@
+import { FragmentRecord, SentenceRecord, SentenceId } from '../types.js';
+
+declare const DEFAULT_FRAGMENT_WORDS_COUNT = 600;
+declare const TEXT_STREAM_KIND: {
+    readonly source: 1;
+    readonly summary: 2;
+};
+type TextStreamName = keyof typeof TEXT_STREAM_KIND;
+interface TextSentenceLocation {
+    readonly byteLength: number;
+    readonly byteOffset: number;
+    readonly sentenceIndex: number;
+    readonly wordsCount: number;
+}
+interface TextStreamDraftState {
+    draftOpen: boolean;
+    nextSentenceIndex?: number;
+}
+interface TextStreamFileAccess {
+    deleteTree(path: string): Promise<void>;
+    ensureDirectory(path: string): Promise<void>;
+    listFiles(path: string): Promise<readonly string[]>;
+    readFile(path: string): Promise<Uint8Array | undefined>;
+    writeFile(path: string, content: string | Uint8Array, options: {
+        readonly overwrite?: boolean;
+    }): Promise<void>;
+}
+interface TextStreamSentenceSegmenter {
+    pipe(stream: Iterable<string>): AsyncIterable<{
+        readonly offset: number;
+        readonly text: string;
+        readonly wordsCount: number;
+    }>;
+}
+interface WriteTextStreamOptions {
+    readonly segmenter?: TextStreamSentenceSegmenter;
+}
+interface ReadonlyTextStreams {
+    getSentence(sentenceId: SentenceId): Promise<string>;
+    getSerial(serialId: number): ReadonlySerialTextStream;
+    getSummarySerial(serialId: number): ReadonlySerialTextStream;
+}
+interface ReadonlySerialTextStream {
+    getFragment(fragmentId: number): Promise<FragmentRecord>;
+    getSentence?(sentenceIndex: number): Promise<SentenceRecord>;
+    listSentencesInRange?(startSentenceIndex: number, endSentenceIndex: number): Promise<readonly SentenceRecord[]>;
+    listFragmentIds(): Promise<readonly number[]>;
+    listSentences?(): Promise<readonly SentenceRecord[]>;
+    readText?(): Promise<string | undefined>;
+    readTextInRange?(startSentenceIndex: number, endSentenceIndex: number): Promise<string | undefined>;
+    readonly path: string;
+    readonly serialId: number;
+}
+
+export { DEFAULT_FRAGMENT_WORDS_COUNT, type ReadonlySerialTextStream, type ReadonlyTextStreams, TEXT_STREAM_KIND, type TextSentenceLocation, type TextStreamDraftState, type TextStreamFileAccess, type TextStreamName, type TextStreamSentenceSegmenter, type WriteTextStreamOptions };

package/dist/document/text-streams/types.js

@@ -0,0 +1,10 @@
+const DEFAULT_FRAGMENT_WORDS_COUNT = 600;
+const TEXT_STREAM_KIND = {
+  source: 1,
+  summary: 2
+};
+export {
+  DEFAULT_FRAGMENT_WORDS_COUNT,
+  TEXT_STREAM_KIND
+};
+//# sourceMappingURL=types.js.map

package/dist/document/text-streams/types.js.map

@@ -0,0 +1 @@
+{"version":3,"sources":["../../../src/document/text-streams/types.ts"],"sourcesContent":["import type { FragmentRecord, SentenceId, SentenceRecord } from \"../types.js\";\n\nexport const DEFAULT_FRAGMENT_WORDS_COUNT = 600;\nexport const TEXT_STREAM_KIND = {\n  source: 1,\n  summary: 2,\n} as const;\n\nexport type TextStreamName = keyof typeof TEXT_STREAM_KIND;\n\nexport interface TextSentenceLocation {\n  readonly byteLength: number;\n  readonly byteOffset: number;\n  readonly sentenceIndex: number;\n  readonly wordsCount: number;\n}\n\nexport interface TextStreamDraftState {\n  draftOpen: boolean;\n  nextSentenceIndex?: number;\n}\n\nexport interface TextStreamFileAccess {\n  deleteTree(path: string): Promise<void>;\n  ensureDirectory(path: string): Promise<void>;\n  listFiles(path: string): Promise<readonly string[]>;\n  readFile(path: string): Promise<Uint8Array | undefined>;\n  writeFile(\n    path: string,\n    content: string | Uint8Array,\n    options: { readonly overwrite?: boolean },\n  ): Promise<void>;\n}\n\nexport interface TextStreamSentenceSegmenter {\n  pipe(stream: Iterable<string>): AsyncIterable<{\n    readonly offset: number;\n    readonly text: string;\n    readonly wordsCount: number;\n  }>;\n}\n\nexport interface WriteTextStreamOptions {\n  readonly segmenter?: TextStreamSentenceSegmenter;\n}\nexport interface ReadonlyTextStreams {\n  getSentence(sentenceId: SentenceId): Promise<string>;\n  getSerial(serialId: number): ReadonlySerialTextStream;\n  getSummarySerial(serialId: number): ReadonlySerialTextStream;\n}\n\nexport interface ReadonlySerialTextStream {\n  getFragment(fragmentId: number): Promise<FragmentRecord>;\n  getSentence?(sentenceIndex: number): Promise<SentenceRecord>;\n  listSentencesInRange?(\n    startSentenceIndex: number,\n    endSentenceIndex: number,\n  ): Promise<readonly SentenceRecord[]>;\n  listFragmentIds(): Promise<readonly number[]>;\n  listSentences?(): Promise<readonly SentenceRecord[]>;\n  readText?(): Promise<string | undefined>;\n  readTextInRange?(\n    startSentenceIndex: number,\n    endSentenceIndex: number,\n  ): Promise<string | undefined>;\n  readonly path: string;\n  readonly serialId: number;\n}\n"],"mappings":"AAEO,MAAM,+BAA+B;AACrC,MAAM,mBAAmB;AAAA,EAC9B,QAAQ;AAAA,EACR,SAAS;AACX;","names":[]}

package/dist/external/guaranteed/classifier.js.map

@@ -0,0 +1 @@
+{"version":3,"sources":["../../../src/external/guaranteed/classifier.ts"],"sourcesContent":["import type { LLMessage } from \"../llm/index.js\";\n\nexport const RESPONSE_INTENT_CLASSIFIER_PROMPT_TEMPLATE =\n  \"guaranteed/response_intent_classifier\";\n\nexport type GuaranteedResponseIntent =\n  | \"ambiguous\"\n  | \"malformed_json\"\n  | \"natural_language\";\n\nconst NATURAL_LANGUAGE_PATTERNS = [\n  /\\b(?:sorry|apologies)\\b/i,\n  /\\b(?:i\\s+can(?:not|'t)|unable\\s+to|do\\s+not\\s+understand)\\b/i,\n  /\\b(?:cannot\\s+answer|can't\\s+answer|can't\\s+help)\\b/i,\n  /\\bno\\s+relevant\\s+result/i,\n  /抱歉/,\n  /无法回答/,\n  /无法识别/,\n  /无法给到/,\n  /没有找到相关的结果/,\n];\n\nexport function buildResponseIntentClassificationMessages(\n  systemPrompt: string,\n  response: string,\n): readonly LLMessage[] {\n  return [\n    {\n      role: \"system\",\n      content: systemPrompt,\n    },\n    {\n      role: \"user\",\n      content: response,\n    },\n  ];\n}\n\nexport function classifyResponseIntentLocally(\n  response: string,\n): GuaranteedResponseIntent {\n  const trimmed = response.trim();\n\n  if (trimmed === \"\") {\n    return \"ambiguous\";\n  }\n\n  if (hasStrongMalformedJsonSignal(trimmed)) {\n    return \"malformed_json\";\n  }\n\n  if (looksLikeNaturalLanguage(trimmed)) {\n    return \"natural_language\";\n  }\n\n  if (hasWeakMalformedJsonSignal(trimmed)) {\n    return \"ambiguous\";\n  }\n\n  if (looksLikePlainSentence(trimmed)) {\n    return \"natural_language\";\n  }\n\n  return \"ambiguous\";\n}\n\nexport function parseResponseIntentClassification(\n  response: string,\n): GuaranteedResponseIntent {\n  const normalized = response.trim().toLowerCase();\n\n  if (normalized.includes(\"malformed_json\")) {\n    return \"malformed_json\";\n  }\n\n  if (normalized.includes(\"natural_language\")) {\n    return \"natural_language\";\n  }\n\n  if (normalized.includes(\"ambiguous\")) {\n    return \"ambiguous\";\n  }\n\n  return \"ambiguous\";\n}\n\nfunction hasStrongMalformedJsonSignal(text: string): boolean {\n  return (\n    /^(?:```json\\b|```|{|\\[)/i.test(text) ||\n    /\"[^\"\\r\\n]+\"\\s*:/.test(text) ||\n    (hasWeakMalformedJsonSignal(text) && endsWithLikelyTruncationToken(text))\n  );\n}\n\nfunction hasWeakMalformedJsonSignal(text: string): boolean {\n  const punctuationCount =\n    (text.match(/[{}[\\]:,]/g) ?? []).length + (text.match(/\"/g) ?? []).length;\n\n  return (\n    /:\\s/.test(text) ||\n    punctuationCount >= 3 ||\n    hasUnbalancedBrackets(text) ||\n    /[:,]\\s*$/.test(text)\n  );\n}\n\nfunction hasUnbalancedBrackets(text: string): boolean {\n  return count(text, \"{\") !== count(text, \"}\") ||\n    count(text, \"[\") !== count(text, \"]\")\n    ? true\n    : false;\n}\n\nfunction endsWithLikelyTruncationToken(text: string): boolean {\n  return /[:, \"{[]\\s*$/.test(text);\n}\n\nfunction looksLikeNaturalLanguage(text: string): boolean {\n  if (NATURAL_LANGUAGE_PATTERNS.some((pattern) => pattern.test(text))) {\n    return true;\n  }\n\n  return !hasWeakMalformedJsonSignal(text) && looksLikePlainSentence(text);\n}\n\nfunction looksLikePlainSentence(text: string): boolean {\n  return (\n    /[\\p{Script=Han}A-Za-z]/u.test(text) &&\n    !/^[{[]/.test(text) &&\n    !/\"[^\"\\r\\n]+\"\\s*:/.test(text)\n  );\n}\n\nfunction count(text: string, character: string): number {\n  return text.split(character).length - 1;\n}\n"],"mappings":"AAEO,MAAM,6CACX;AAOF,MAAM,4BAA4B;AAAA,EAChC;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AACF;AAEO,SAAS,0CACd,cACA,UACsB;AACtB,SAAO;AAAA,IACL;AAAA,MACE,MAAM;AAAA,MACN,SAAS;AAAA,IACX;AAAA,IACA;AAAA,MACE,MAAM;AAAA,MACN,SAAS;AAAA,IACX;AAAA,EACF;AACF;AAEO,SAAS,8BACd,UAC0B;AAC1B,QAAM,UAAU,SAAS,KAAK;AAE9B,MAAI,YAAY,IAAI;AAClB,WAAO;AAAA,EACT;AAEA,MAAI,6BAA6B,OAAO,GAAG;AACzC,WAAO;AAAA,EACT;AAEA,MAAI,yBAAyB,OAAO,GAAG;AACrC,WAAO;AAAA,EACT;AAEA,MAAI,2BAA2B,OAAO,GAAG;AACvC,WAAO;AAAA,EACT;AAEA,MAAI,uBAAuB,OAAO,GAAG;AACnC,WAAO;AAAA,EACT;AAEA,SAAO;AACT;AAEO,SAAS,kCACd,UAC0B;AAC1B,QAAM,aAAa,SAAS,KAAK,EAAE,YAAY;AAE/C,MAAI,WAAW,SAAS,gBAAgB,GAAG;AACzC,WAAO;AAAA,EACT;AAEA,MAAI,WAAW,SAAS,kBAAkB,GAAG;AAC3C,WAAO;AAAA,EACT;AAEA,MAAI,WAAW,SAAS,WAAW,GAAG;AACpC,WAAO;AAAA,EACT;AAEA,SAAO;AACT;AAEA,SAAS,6BAA6B,MAAuB;AAC3D,SACE,2BAA2B,KAAK,IAAI,KACpC,kBAAkB,KAAK,IAAI,KAC1B,2BAA2B,IAAI,KAAK,8BAA8B,IAAI;AAE3E;AAEA,SAAS,2BAA2B,MAAuB;AACzD,QAAM,oBACH,KAAK,MAAM,YAAY,KAAK,CAAC,GAAG,UAAU,KAAK,MAAM,IAAI,KAAK,CAAC,GAAG;AAErE,SACE,MAAM,KAAK,IAAI,KACf,oBAAoB,KACpB,sBAAsB,IAAI,KAC1B,WAAW,KAAK,IAAI;AAExB;AAEA,SAAS,sBAAsB,MAAuB;AACpD,SAAO,MAAM,MAAM,GAAG,MAAM,MAAM,MAAM,GAAG,KACzC,MAAM,MAAM,GAAG,MAAM,MAAM,MAAM,GAAG,IAClC,OACA;AACN;AAEA,SAAS,8BAA8B,MAAuB;AAC5D,SAAO,eAAe,KAAK,IAAI;AACjC;AAEA,SAAS,yBAAyB,MAAuB;AACvD,MAAI,0BAA0B,KAAK,CAAC,YAAY,QAAQ,KAAK,IAAI,CAAC,GAAG;AACnE,WAAO;AAAA,EACT;AAEA,SAAO,CAAC,2BAA2B,IAAI,KAAK,uBAAuB,IAAI;AACzE;AAEA,SAAS,uBAAuB,MAAuB;AACrD,SACE,0BAA0B,KAAK,IAAI,KACnC,CAAC,QAAQ,KAAK,IAAI,KAClB,CAAC,kBAAkB,KAAK,IAAI;AAEhC;AAEA,SAAS,MAAM,MAAc,WAA2B;AACtD,SAAO,KAAK,MAAM,SAAS,EAAE,SAAS;AACxC;","names":[]}

package/dist/external/guaranteed/errors.js.map

@@ -0,0 +1 @@
+{"version":3,"sources":["../../../src/external/guaranteed/errors.ts"],"sourcesContent":["export class ParsedJsonError extends Error {\n  public readonly issues: readonly string[];\n\n  public constructor(issues: readonly string[]) {\n    super(`Parse failed with ${issues.length} issue(s)`);\n    this.name = \"ParsedJsonError\";\n    this.issues = [...issues];\n  }\n}\n\nexport abstract class GuaranteedRequestFailureError extends Error {\n  public readonly attempts: number;\n  public readonly maxRetries: number;\n\n  protected constructor(message: string, attempts: number, maxRetries: number) {\n    super(message);\n    this.name = \"GuaranteedRequestFailureError\";\n    this.attempts = attempts;\n    this.maxRetries = maxRetries;\n  }\n}\n\nexport class GuaranteedEmptyResponseError extends GuaranteedRequestFailureError {\n  public constructor(attempts: number, maxRetries: number) {\n    super(\n      \"LLM returned empty response after all retries\",\n      attempts,\n      maxRetries,\n    );\n    this.name = \"GuaranteedEmptyResponseError\";\n  }\n}\n\nexport class SuspectedModelRefusalError extends GuaranteedRequestFailureError {\n  public readonly response: string;\n  public readonly reason: string;\n\n  public constructor(\n    attempts: number,\n    maxRetries: number,\n    input: {\n      response: string;\n      reason: string;\n    },\n  ) {\n    super(\n      `Suspected model refusal after ${attempts} JSON syntax error attempt(s): ${input.reason}. Last response: ${JSON.stringify(input.response)}`,\n      attempts,\n      maxRetries,\n    );\n    this.name = \"SuspectedModelRefusalError\";\n    this.response = input.response;\n    this.reason = input.reason;\n  }\n}\n\nexport class GuaranteedSchemaValidationError extends GuaranteedRequestFailureError {\n  public readonly issues: readonly string[];\n  public readonly response: string;\n\n  public constructor(\n    attempts: number,\n    maxRetries: number,\n    input: {\n      issues: readonly string[];\n      response: string;\n    },\n    cause: unknown,\n  ) {\n    super(\"Schema validation failed after all retries\", attempts, maxRetries);\n    this.name = \"GuaranteedSchemaValidationError\";\n    this.issues = [...input.issues];\n    this.response = input.response;\n    this.cause = cause;\n  }\n}\n\nexport class GuaranteedParseValidationError extends GuaranteedRequestFailureError {\n  public readonly issues: readonly string[];\n  public readonly response: string;\n\n  public constructor(\n    attempts: number,\n    maxRetries: number,\n    input: {\n      issues: readonly string[];\n      response: string;\n    },\n    cause: unknown,\n  ) {\n    super(\"Parse validation failed after all retries\", attempts, maxRetries);\n    this.name = \"GuaranteedParseValidationError\";\n    this.issues = [...input.issues];\n    this.response = input.response;\n    this.cause = cause;\n  }\n}\n"],"mappings":"AAAO,MAAM,wBAAwB,MAAM;AAAA,EACzB;AAAA,EAET,YAAY,QAA2B;AAC5C,UAAM,qBAAqB,OAAO,MAAM,WAAW;AACnD,SAAK,OAAO;AACZ,SAAK,SAAS,CAAC,GAAG,MAAM;AAAA,EAC1B;AACF;AAEO,MAAe,sCAAsC,MAAM;AAAA,EAChD;AAAA,EACA;AAAA,EAEN,YAAY,SAAiB,UAAkB,YAAoB;AAC3E,UAAM,OAAO;AACb,SAAK,OAAO;AACZ,SAAK,WAAW;AAChB,SAAK,aAAa;AAAA,EACpB;AACF;AAEO,MAAM,qCAAqC,8BAA8B;AAAA,EACvE,YAAY,UAAkB,YAAoB;AACvD;AAAA,MACE;AAAA,MACA;AAAA,MACA;AAAA,IACF;AACA,SAAK,OAAO;AAAA,EACd;AACF;AAEO,MAAM,mCAAmC,8BAA8B;AAAA,EAC5D;AAAA,EACA;AAAA,EAET,YACL,UACA,YACA,OAIA;AACA;AAAA,MACE,iCAAiC,QAAQ,kCAAkC,MAAM,MAAM,oBAAoB,KAAK,UAAU,MAAM,QAAQ,CAAC;AAAA,MACzI;AAAA,MACA;AAAA,IACF;AACA,SAAK,OAAO;AACZ,SAAK,WAAW,MAAM;AACtB,SAAK,SAAS,MAAM;AAAA,EACtB;AACF;AAEO,MAAM,wCAAwC,8BAA8B;AAAA,EACjE;AAAA,EACA;AAAA,EAET,YACL,UACA,YACA,OAIA,OACA;AACA,UAAM,8CAA8C,UAAU,UAAU;AACxE,SAAK,OAAO;AACZ,SAAK,SAAS,CAAC,GAAG,MAAM,MAAM;AAC9B,SAAK,WAAW,MAAM;AACtB,SAAK,QAAQ;AAAA,EACf;AACF;AAEO,MAAM,uCAAuC,8BAA8B;AAAA,EAChE;AAAA,EACA;AAAA,EAET,YACL,UACA,YACA,OAIA,OACA;AACA,UAAM,6CAA6C,UAAU,UAAU;AACvE,SAAK,OAAO;AACZ,SAAK,SAAS,CAAC,GAAG,MAAM,MAAM;AAC9B,SAAK,WAAW,MAAM;AACtB,SAAK,QAAQ;AAAA,EACf;AACF;","names":[]}

package/dist/external/guaranteed/index.js.map

@@ -0,0 +1 @@
+{"version":3,"sources":["../../../src/external/guaranteed/index.ts"],"sourcesContent":["export {\n  GuaranteedEmptyResponseError,\n  GuaranteedRequestFailureError,\n  GuaranteedParseValidationError,\n  GuaranteedSchemaValidationError,\n  ParsedJsonError,\n  SuspectedModelRefusalError,\n} from \"./errors.js\";\nexport { RESPONSE_INTENT_CLASSIFIER_PROMPT_TEMPLATE } from \"./classifier.js\";\nexport { requestGuaranteedJson } from \"./request.js\";\nexport type {\n  GuaranteedLazyRequest,\n  GuaranteedParser,\n  GuaranteedRequest,\n  GuaranteedRequestController,\n  GuaranteedRequestOptions,\n} from \"./types.js\";\n"],"mappings":"AAAA;AAAA,EACE;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,OACK;AACP,SAAS,kDAAkD;AAC3D,SAAS,6BAA6B;","names":[]}

package/dist/external/guaranteed/request.js.map

@@ -0,0 +1 @@
+{"version":3,"sources":["../../../src/external/guaranteed/request.ts"],"sourcesContent":["import type { LLMessage } from \"../llm/index.js\";\nimport {\n  buildResponseIntentClassificationMessages,\n  classifyResponseIntentLocally,\n  parseResponseIntentClassification,\n  type GuaranteedResponseIntent,\n} from \"./classifier.js\";\nimport {\n  GuaranteedEmptyResponseError,\n  GuaranteedParseValidationError,\n  GuaranteedSchemaValidationError,\n  ParsedJsonError,\n  SuspectedModelRefusalError,\n} from \"./errors.js\";\nimport {\n  buildMalformedJsonMessage,\n  buildNaturalLanguageMessage,\n  buildBusinessErrorMessage,\n  buildSchemaErrorMessage,\n  buildSyntaxErrorMessage,\n  extractJsonText,\n  listSchemaIssues,\n  repairJsonText,\n} from \"./response.js\";\nimport type { GuaranteedRequestOptions } from \"./types.js\";\n\nconst DEFAULT_MAX_RETRIES = 12;\nconst MAX_MALFORMED_JSON_REPAIR_HISTORY_ATTEMPTS = 3;\n\nexport async function requestGuaranteedJson<TData, TResult>(\n  options: GuaranteedRequestOptions<TData, TResult>,\n): Promise<TResult> {\n  const initialMessages = [...options.messages];\n  let currentMessages = [...options.messages];\n  let consecutiveProtocolDerailments = 0;\n  let consecutiveMalformedJsonFailures = 0;\n  const maxRetries = options.maxRetries ?? DEFAULT_MAX_RETRIES;\n\n  for (let index = 0; index <= maxRetries; index += 1) {\n    const response = await options.request(currentMessages, index, maxRetries);\n    if (response === undefined || response.trim() === \"\") {\n      if (index >= maxRetries) {\n        throw new GuaranteedEmptyResponseError(index + 1, maxRetries);\n      }\n      continue;\n    }\n    let parsedData: unknown;\n\n    try {\n      const extractedJsonText = extractJsonText(response);\n      const repairedJsonText = repairJsonText(extractedJsonText);\n\n      parsedData = JSON.parse(repairedJsonText);\n    } catch (error) {\n      const intent = await classifyResponseIntent(options, response, index);\n\n      if (intent === \"natural_language\") {\n        consecutiveProtocolDerailments += 1;\n        consecutiveMalformedJsonFailures = 0;\n\n        if (consecutiveProtocolDerailments >= 2 || index >= maxRetries) {\n          const reason =\n            index >= maxRetries\n              ? \"last retry still returned natural-language content\"\n              : \"two consecutive retries returned natural-language content\";\n          throw new SuspectedModelRefusalError(index + 1, maxRetries, {\n            response,\n            reason,\n          });\n        }\n\n        currentMessages = buildRetryMessages(\n          initialMessages,\n          response,\n          buildNaturalLanguageMessage(),\n          false,\n        );\n        continue;\n      }\n\n      consecutiveProtocolDerailments = 0;\n      consecutiveMalformedJsonFailures += 1;\n      currentMessages = buildRetryMessages(\n        initialMessages,\n        response,\n        intent === \"malformed_json\"\n          ? buildMalformedJsonMessage(asSyntaxError(error))\n          : buildSyntaxErrorMessage(asSyntaxError(error)),\n        consecutiveMalformedJsonFailures <\n          MAX_MALFORMED_JSON_REPAIR_HISTORY_ATTEMPTS,\n      );\n      continue;\n    }\n    consecutiveProtocolDerailments = 0;\n    consecutiveMalformedJsonFailures = 0;\n\n    const validation = await options.schema.safeParseAsync(parsedData);\n\n    if (!validation.success) {\n      if (\n        shouldTreatSchemaFailureAsNaturalLanguage(validation.error, parsedData)\n      ) {\n        const intent = await classifyResponseIntent(options, response, index);\n\n        if (intent === \"natural_language\") {\n          consecutiveProtocolDerailments += 1;\n          consecutiveMalformedJsonFailures = 0;\n\n          if (consecutiveProtocolDerailments >= 2 || index >= maxRetries) {\n            const reason =\n              index >= maxRetries\n                ? \"last retry still returned natural-language content\"\n                : \"two consecutive retries returned natural-language content\";\n            throw new SuspectedModelRefusalError(index + 1, maxRetries, {\n              response,\n              reason,\n            });\n          }\n\n          currentMessages = buildRetryMessages(\n            initialMessages,\n            response,\n            buildNaturalLanguageMessage(),\n            false,\n          );\n          continue;\n        }\n      }\n\n      consecutiveProtocolDerailments = 0;\n      const feedback = buildSchemaErrorMessage(validation.error);\n\n      if (index >= maxRetries) {\n        throw new GuaranteedSchemaValidationError(\n          index + 1,\n          maxRetries,\n          {\n            issues: listSchemaIssues(validation.error),\n            response,\n          },\n          validation.error,\n        );\n      }\n      currentMessages = buildRetryMessages(\n        initialMessages,\n        response,\n        feedback,\n        true,\n      );\n      continue;\n    }\n\n    try {\n      return await options.parse(validation.data, index, maxRetries);\n    } catch (error) {\n      if (!(error instanceof ParsedJsonError)) {\n        throw error;\n      }\n      const feedback = buildBusinessErrorMessage(error.issues);\n\n      if (index >= maxRetries) {\n        throw new GuaranteedParseValidationError(\n          index + 1,\n          maxRetries,\n          {\n            issues: error.issues,\n            response,\n          },\n          error,\n        );\n      }\n      currentMessages = buildRetryMessages(\n        initialMessages,\n        response,\n        feedback,\n        true,\n      );\n    }\n  }\n  throw new Error(\"requestGuaranteedJson failed unexpectedly\");\n}\n\nfunction buildRetryMessages(\n  initialMessages: readonly LLMessage[],\n  response: string,\n  feedback: string,\n  includeAssistantResponse: boolean,\n): LLMessage[] {\n  return [\n    ...initialMessages,\n    ...(includeAssistantResponse\n      ? [\n          {\n            role: \"assistant\" as const,\n            content: response,\n          },\n        ]\n      : []),\n    {\n      role: \"user\",\n      content: feedback,\n    },\n  ];\n}\n\nfunction asSyntaxError(error: unknown): SyntaxError {\n  if (error instanceof SyntaxError) {\n    return error;\n  }\n\n  return new SyntaxError(String(error));\n}\n\nasync function classifyResponseIntent<TData, TResult>(\n  options: GuaranteedRequestOptions<TData, TResult>,\n  response: string,\n  index: number,\n): Promise<GuaranteedResponseIntent> {\n  const localIntent = classifyResponseIntentLocally(response);\n\n  if (localIntent !== \"ambiguous\") {\n    return localIntent;\n  }\n\n  try {\n    const classifierResponse = await options.request(\n      buildResponseIntentClassificationMessages(\n        options.responseIntentClassifierPrompt,\n        response,\n      ),\n      index,\n      options.maxRetries ?? DEFAULT_MAX_RETRIES,\n    );\n\n    if (classifierResponse === undefined || classifierResponse.trim() === \"\") {\n      return \"natural_language\";\n    }\n\n    const classifierIntent =\n      parseResponseIntentClassification(classifierResponse);\n\n    return classifierIntent === \"ambiguous\"\n      ? \"natural_language\"\n      : classifierIntent;\n  } catch {\n    return \"natural_language\";\n  }\n}\n\nfunction shouldTreatSchemaFailureAsNaturalLanguage(\n  error: {\n    issues: readonly {\n      code?: string;\n      path: readonly PropertyKey[];\n      expected?: unknown;\n    }[];\n  },\n  parsedData: unknown,\n): boolean {\n  if (!isPrimitiveJsonValue(parsedData)) {\n    return false;\n  }\n\n  return error.issues.some(\n    (issue) =>\n      issue.code === \"invalid_type\" &&\n      issue.path.length === 0 &&\n      (issue.expected === \"object\" || issue.expected === \"array\"),\n  );\n}\n\nfunction isPrimitiveJsonValue(value: unknown): boolean {\n  return (\n    value === null || (typeof value !== \"object\" && typeof value !== \"function\")\n  );\n}\n"],"mappings":"AACA;AAAA,EACE;AAAA,EACA;AAAA,EACA;AAAA,OAEK;AACP;AAAA,EACE;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,OACK;AACP;AAAA,EACE;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,OACK;AAGP,MAAM,sBAAsB;AAC5B,MAAM,6CAA6C;AAEnD,eAAsB,sBACpB,SACkB;AAClB,QAAM,kBAAkB,CAAC,GAAG,QAAQ,QAAQ;AAC5C,MAAI,kBAAkB,CAAC,GAAG,QAAQ,QAAQ;AAC1C,MAAI,iCAAiC;AACrC,MAAI,mCAAmC;AACvC,QAAM,aAAa,QAAQ,cAAc;AAEzC,WAAS,QAAQ,GAAG,SAAS,YAAY,SAAS,GAAG;AACnD,UAAM,WAAW,MAAM,QAAQ,QAAQ,iBAAiB,OAAO,UAAU;AACzE,QAAI,aAAa,UAAa,SAAS,KAAK,MAAM,IAAI;AACpD,UAAI,SAAS,YAAY;AACvB,cAAM,IAAI,6BAA6B,QAAQ,GAAG,UAAU;AAAA,MAC9D;AACA;AAAA,IACF;AACA,QAAI;AAEJ,QAAI;AACF,YAAM,oBAAoB,gBAAgB,QAAQ;AAClD,YAAM,mBAAmB,eAAe,iBAAiB;AAEzD,mBAAa,KAAK,MAAM,gBAAgB;AAAA,IAC1C,SAAS,OAAO;AACd,YAAM,SAAS,MAAM,uBAAuB,SAAS,UAAU,KAAK;AAEpE,UAAI,WAAW,oBAAoB;AACjC,0CAAkC;AAClC,2CAAmC;AAEnC,YAAI,kCAAkC,KAAK,SAAS,YAAY;AAC9D,gBAAM,SACJ,SAAS,aACL,uDACA;AACN,gBAAM,IAAI,2BAA2B,QAAQ,GAAG,YAAY;AAAA,YAC1D;AAAA,YACA;AAAA,UACF,CAAC;AAAA,QACH;AAEA,0BAAkB;AAAA,UAChB;AAAA,UACA;AAAA,UACA,4BAA4B;AAAA,UAC5B;AAAA,QACF;AACA;AAAA,MACF;AAEA,uCAAiC;AACjC,0CAAoC;AACpC,wBAAkB;AAAA,QAChB;AAAA,QACA;AAAA,QACA,WAAW,mBACP,0BAA0B,cAAc,KAAK,CAAC,IAC9C,wBAAwB,cAAc,KAAK,CAAC;AAAA,QAChD,mCACE;AAAA,MACJ;AACA;AAAA,IACF;AACA,qCAAiC;AACjC,uCAAmC;AAEnC,UAAM,aAAa,MAAM,QAAQ,OAAO,eAAe,UAAU;AAEjE,QAAI,CAAC,WAAW,SAAS;AACvB,UACE,0CAA0C,WAAW,OAAO,UAAU,GACtE;AACA,cAAM,SAAS,MAAM,uBAAuB,SAAS,UAAU,KAAK;AAEpE,YAAI,WAAW,oBAAoB;AACjC,4CAAkC;AAClC,6CAAmC;AAEnC,cAAI,kCAAkC,KAAK,SAAS,YAAY;AAC9D,kBAAM,SACJ,SAAS,aACL,uDACA;AACN,kBAAM,IAAI,2BAA2B,QAAQ,GAAG,YAAY;AAAA,cAC1D;AAAA,cACA;AAAA,YACF,CAAC;AAAA,UACH;AAEA,4BAAkB;AAAA,YAChB;AAAA,YACA;AAAA,YACA,4BAA4B;AAAA,YAC5B;AAAA,UACF;AACA;AAAA,QACF;AAAA,MACF;AAEA,uCAAiC;AACjC,YAAM,WAAW,wBAAwB,WAAW,KAAK;AAEzD,UAAI,SAAS,YAAY;AACvB,cAAM,IAAI;AAAA,UACR,QAAQ;AAAA,UACR;AAAA,UACA;AAAA,YACE,QAAQ,iBAAiB,WAAW,KAAK;AAAA,YACzC;AAAA,UACF;AAAA,UACA,WAAW;AAAA,QACb;AAAA,MACF;AACA,wBAAkB;AAAA,QAChB;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,MACF;AACA;AAAA,IACF;AAEA,QAAI;AACF,aAAO,MAAM,QAAQ,MAAM,WAAW,MAAM,OAAO,UAAU;AAAA,IAC/D,SAAS,OAAO;AACd,UAAI,EAAE,iBAAiB,kBAAkB;AACvC,cAAM;AAAA,MACR;AACA,YAAM,WAAW,0BAA0B,MAAM,MAAM;AAEvD,UAAI,SAAS,YAAY;AACvB,cAAM,IAAI;AAAA,UACR,QAAQ;AAAA,UACR;AAAA,UACA;AAAA,YACE,QAAQ,MAAM;AAAA,YACd;AAAA,UACF;AAAA,UACA;AAAA,QACF;AAAA,MACF;AACA,wBAAkB;AAAA,QAChB;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,MACF;AAAA,IACF;AAAA,EACF;AACA,QAAM,IAAI,MAAM,2CAA2C;AAC7D;AAEA,SAAS,mBACP,iBACA,UACA,UACA,0BACa;AACb,SAAO;AAAA,IACL,GAAG;AAAA,IACH,GAAI,2BACA;AAAA,MACE;AAAA,QACE,MAAM;AAAA,QACN,SAAS;AAAA,MACX;AAAA,IACF,IACA,CAAC;AAAA,IACL;AAAA,MACE,MAAM;AAAA,MACN,SAAS;AAAA,IACX;AAAA,EACF;AACF;AAEA,SAAS,cAAc,OAA6B;AAClD,MAAI,iBAAiB,aAAa;AAChC,WAAO;AAAA,EACT;AAEA,SAAO,IAAI,YAAY,OAAO,KAAK,CAAC;AACtC;AAEA,eAAe,uBACb,SACA,UACA,OACmC;AACnC,QAAM,cAAc,8BAA8B,QAAQ;AAE1D,MAAI,gBAAgB,aAAa;AAC/B,WAAO;AAAA,EACT;AAEA,MAAI;AACF,UAAM,qBAAqB,MAAM,QAAQ;AAAA,MACvC;AAAA,QACE,QAAQ;AAAA,QACR;AAAA,MACF;AAAA,MACA;AAAA,MACA,QAAQ,cAAc;AAAA,IACxB;AAEA,QAAI,uBAAuB,UAAa,mBAAmB,KAAK,MAAM,IAAI;AACxE,aAAO;AAAA,IACT;AAEA,UAAM,mBACJ,kCAAkC,kBAAkB;AAEtD,WAAO,qBAAqB,cACxB,qBACA;AAAA,EACN,QAAQ;AACN,WAAO;AAAA,EACT;AACF;AAEA,SAAS,0CACP,OAOA,YACS;AACT,MAAI,CAAC,qBAAqB,UAAU,GAAG;AACrC,WAAO;AAAA,EACT;AAEA,SAAO,MAAM,OAAO;AAAA,IAClB,CAAC,UACC,MAAM,SAAS,kBACf,MAAM,KAAK,WAAW,MACrB,MAAM,aAAa,YAAY,MAAM,aAAa;AAAA,EACvD;AACF;AAEA,SAAS,qBAAqB,OAAyB;AACrD,SACE,UAAU,QAAS,OAAO,UAAU,YAAY,OAAO,UAAU;AAErE;","names":[]}

package/dist/external/guaranteed/response.js.map

@@ -0,0 +1 @@
+{"version":3,"sources":["../../../src/external/guaranteed/response.ts"],"sourcesContent":["import { jsonrepair } from \"jsonrepair\";\nimport type { ZodError, core } from \"zod\";\n\nexport function extractJsonText(response: string): string {\n  const withoutCodeFence = response.replace(\n    /```(?:json)?\\s*\\n(.*?)\\n```/gs,\n    \"$1\",\n  );\n  const matchedArray = withoutCodeFence.match(/\\[[\\s\\S]*\\]/);\n  const matchedObject = withoutCodeFence.match(/\\{[\\s\\S]*\\}/);\n\n  if (\n    matchedArray?.index !== undefined &&\n    matchedArray[0] !== undefined &&\n    (matchedObject?.index === undefined ||\n      matchedArray.index < matchedObject.index)\n  ) {\n    return matchedArray[0];\n  }\n\n  if (matchedObject?.[0] !== undefined) {\n    return matchedObject[0];\n  }\n\n  return withoutCodeFence.trim();\n}\n\nexport function repairJsonText(jsonText: string): string {\n  try {\n    return jsonrepair(jsonText);\n  } catch {\n    return jsonText;\n  }\n}\n\nexport function buildSyntaxErrorMessage(error: SyntaxError): string {\n  const lines = [\n    \"Your JSON has a syntax error.\",\n    `Error: ${error.message}`,\n    \"\",\n    \"Common issues:\",\n    \"- Trailing commas in arrays or objects\",\n    '- Unescaped quotes in strings, for example: \"He said \\\\\"hi\\\\\"\"',\n    '- Single quotes instead of double quotes, for example: {\"key\": \"value\"}',\n    \"- Missing commas between items\",\n    \"\",\n    \"Regenerate the complete and valid JSON directly. Do not add explanations or markdown fences.\",\n  ];\n\n  return lines.join(\"\\n\");\n}\n\nexport function buildMalformedJsonMessage(error: SyntaxError): string {\n  const lines = [\n    \"Your previous reply looked like malformed JSON.\",\n    `Error: ${error.message}`,\n    \"\",\n    \"Return complete and valid JSON directly.\",\n    \"Do not answer conversationally.\",\n    \"Do not add explanations or markdown fences.\",\n  ];\n\n  return lines.join(\"\\n\");\n}\n\nexport function buildNaturalLanguageMessage(): string {\n  const lines = [\n    \"Your previous reply was plain natural language, not a JSON object or array.\",\n    \"\",\n    \"Do not apologize or explain.\",\n    \"Do not answer conversationally.\",\n    \"Return complete and valid JSON directly.\",\n  ];\n\n  return lines.join(\"\\n\");\n}\n\nexport function buildSchemaErrorMessage(error: ZodError): string {\n  const issues = error.issues.map(formatZodIssue);\n  const lines = [\n    \"Your JSON has structural issues.\",\n    \"\",\n    ...issues,\n    \"\",\n    \"Regenerate the complete and valid JSON directly. Do not add explanations.\",\n  ];\n\n  return lines.join(\"\\n\");\n}\n\nexport function listSchemaIssues(error: ZodError): readonly string[] {\n  return error.issues.map(formatZodIssue);\n}\n\nexport function buildBusinessErrorMessage(issues: readonly string[]): string {\n  const lines = [\n    \"Your JSON has the following issues.\",\n    \"\",\n    ...issues.map((issue) => `- ${issue}`),\n    \"\",\n    \"Regenerate the complete and valid JSON directly. Do not add explanations.\",\n  ];\n\n  return lines.join(\"\\n\");\n}\n\nfunction formatZodIssue(issue: core.$ZodIssue): string {\n  const location = issue.path.length === 0 ? \"<root>\" : issue.path.join(\".\");\n\n  return `- **${location}**: ${issue.message}`;\n}\n"],"mappings":"AAAA,SAAS,kBAAkB;AAGpB,SAAS,gBAAgB,UAA0B;AACxD,QAAM,mBAAmB,SAAS;AAAA,IAChC;AAAA,IACA;AAAA,EACF;AACA,QAAM,eAAe,iBAAiB,MAAM,aAAa;AACzD,QAAM,gBAAgB,iBAAiB,MAAM,aAAa;AAE1D,MACE,cAAc,UAAU,UACxB,aAAa,CAAC,MAAM,WACnB,eAAe,UAAU,UACxB,aAAa,QAAQ,cAAc,QACrC;AACA,WAAO,aAAa,CAAC;AAAA,EACvB;AAEA,MAAI,gBAAgB,CAAC,MAAM,QAAW;AACpC,WAAO,cAAc,CAAC;AAAA,EACxB;AAEA,SAAO,iBAAiB,KAAK;AAC/B;AAEO,SAAS,eAAe,UAA0B;AACvD,MAAI;AACF,WAAO,WAAW,QAAQ;AAAA,EAC5B,QAAQ;AACN,WAAO;AAAA,EACT;AACF;AAEO,SAAS,wBAAwB,OAA4B;AAClE,QAAM,QAAQ;AAAA,IACZ;AAAA,IACA,UAAU,MAAM,OAAO;AAAA,IACvB;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,EACF;AAEA,SAAO,MAAM,KAAK,IAAI;AACxB;AAEO,SAAS,0BAA0B,OAA4B;AACpE,QAAM,QAAQ;AAAA,IACZ;AAAA,IACA,UAAU,MAAM,OAAO;AAAA,IACvB;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,EACF;AAEA,SAAO,MAAM,KAAK,IAAI;AACxB;AAEO,SAAS,8BAAsC;AACpD,QAAM,QAAQ;AAAA,IACZ;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,EACF;AAEA,SAAO,MAAM,KAAK,IAAI;AACxB;AAEO,SAAS,wBAAwB,OAAyB;AAC/D,QAAM,SAAS,MAAM,OAAO,IAAI,cAAc;AAC9C,QAAM,QAAQ;AAAA,IACZ;AAAA,IACA;AAAA,IACA,GAAG;AAAA,IACH;AAAA,IACA;AAAA,EACF;AAEA,SAAO,MAAM,KAAK,IAAI;AACxB;AAEO,SAAS,iBAAiB,OAAoC;AACnE,SAAO,MAAM,OAAO,IAAI,cAAc;AACxC;AAEO,SAAS,0BAA0B,QAAmC;AAC3E,QAAM,QAAQ;AAAA,IACZ;AAAA,IACA;AAAA,IACA,GAAG,OAAO,IAAI,CAAC,UAAU,KAAK,KAAK,EAAE;AAAA,IACrC;AAAA,IACA;AAAA,EACF;AAEA,SAAO,MAAM,KAAK,IAAI;AACxB;AAEA,SAAS,eAAe,OAA+B;AACrD,QAAM,WAAW,MAAM,KAAK,WAAW,IAAI,WAAW,MAAM,KAAK,KAAK,GAAG;AAEzE,SAAO,OAAO,QAAQ,OAAO,MAAM,OAAO;AAC5C;","names":[]}