wiki-graph-core 0.3.1 → 0.4.1

package/dist/text/reader/chunk-batch/language.js

@@ -0,0 +1,23 @@
+import {
+  detectLanguageCode,
+  getLanguageDetectionCode
+} from "../../../runtime/common/tinyld-language.js";
+import {} from "../../../runtime/common/language.js";
+function needsTranslation(input) {
+  const targetLanguageCode = getLanguageDetectionCode(input.targetLanguage);
+  return fieldNeedsTranslation(input.label, targetLanguageCode) || fieldNeedsTranslation(input.content, targetLanguageCode);
+}
+function fieldNeedsTranslation(text, targetLanguageCode) {
+  if (text.trim() === "") {
+    return false;
+  }
+  const detectedLanguageCode = detectLanguageCode(text);
+  if (detectedLanguageCode === void 0) {
+    return true;
+  }
+  return detectedLanguageCode !== targetLanguageCode;
+}
+export {
+  needsTranslation
+};
+//# sourceMappingURL=language.js.map

package/dist/text/reader/chunk-batch/language.js.map

@@ -0,0 +1 @@
+{"version":3,"sources":["../../../../src/text/reader/chunk-batch/language.ts"],"sourcesContent":["import {\n  detectLanguageCode,\n  getLanguageDetectionCode,\n} from \"../../../runtime/common/tinyld-language.js\";\nimport { type Language } from \"../../../runtime/common/language.js\";\n\nexport function needsTranslation(input: {\n  content: string;\n  label: string;\n  targetLanguage: Language;\n}): boolean {\n  const targetLanguageCode = getLanguageDetectionCode(input.targetLanguage);\n\n  return (\n    fieldNeedsTranslation(input.label, targetLanguageCode) ||\n    fieldNeedsTranslation(input.content, targetLanguageCode)\n  );\n}\n\nfunction fieldNeedsTranslation(\n  text: string,\n  targetLanguageCode: string,\n): boolean {\n  if (text.trim() === \"\") {\n    return false;\n  }\n\n  const detectedLanguageCode = detectLanguageCode(text);\n\n  if (detectedLanguageCode === undefined) {\n    return true;\n  }\n\n  return detectedLanguageCode !== targetLanguageCode;\n}\n"],"mappings":"AAAA;AAAA,EACE;AAAA,EACA;AAAA,OACK;AACP,eAA8B;AAEvB,SAAS,iBAAiB,OAIrB;AACV,QAAM,qBAAqB,yBAAyB,MAAM,cAAc;AAExE,SACE,sBAAsB,MAAM,OAAO,kBAAkB,KACrD,sBAAsB,MAAM,SAAS,kBAAkB;AAE3D;AAEA,SAAS,sBACP,MACA,oBACS;AACT,MAAI,KAAK,KAAK,MAAM,IAAI;AACtB,WAAO;AAAA,EACT;AAEA,QAAM,uBAAuB,mBAAmB,IAAI;AAEpD,MAAI,yBAAyB,QAAW;AACtC,WAAO;AAAA,EACT;AAEA,SAAO,yBAAyB;AAClC;","names":[]}

package/dist/text/reader/chunk-batch/parser/annotations.d.ts

@@ -0,0 +1,19 @@
+import { ChunkImportance } from '../../../../document/types.js';
+import { ChunkImportanceAnnotation } from '../types.js';
+import '../../../../runtime/common/language.js';
+import '../../../../external/llm/client/core.js';
+import '../../../../external/llm/context.js';
+import '../../../../external/llm/cache.js';
+import '../../../../external/llm/types.js';
+import 'ai';
+
+declare function validateImportanceAnnotations(input: {
+    readonly annotations: readonly {
+        readonly chunk_id: number;
+        readonly importance: ChunkImportance;
+    }[];
+    readonly issues: string[];
+    readonly validImportanceChunkIds: Readonly<Record<string, true>> | undefined;
+}): ChunkImportanceAnnotation[] | undefined;
+
+export { validateImportanceAnnotations };

package/dist/text/reader/chunk-batch/parser/annotations.js

@@ -0,0 +1,34 @@
+import {
+  expectChunkImportance
+} from "../../../../document/index.js";
+import { hasMembership } from "./helpers.js";
+function validateImportanceAnnotations(input) {
+  const { annotations, issues, validImportanceChunkIds } = input;
+  if (annotations.length === 0) {
+    return [];
+  }
+  if (validImportanceChunkIds === void 0) {
+    return annotations.map((annotation) => ({
+      chunkId: annotation.chunk_id,
+      importance: expectChunkImportance(annotation.importance)
+    }));
+  }
+  const result = [];
+  for (const annotation of annotations) {
+    if (!hasMembership(validImportanceChunkIds, annotation.chunk_id)) {
+      issues.push(
+        `importance_annotations references unknown chunk_id ${annotation.chunk_id}`
+      );
+      continue;
+    }
+    result.push({
+      chunkId: annotation.chunk_id,
+      importance: expectChunkImportance(annotation.importance)
+    });
+  }
+  return result;
+}
+export {
+  validateImportanceAnnotations
+};
+//# sourceMappingURL=annotations.js.map

package/dist/text/reader/chunk-batch/parser/annotations.js.map

@@ -0,0 +1 @@
+{"version":3,"sources":["../../../../../src/text/reader/chunk-batch/parser/annotations.ts"],"sourcesContent":["import {\n  expectChunkImportance,\n  type ChunkImportance,\n} from \"../../../../document/index.js\";\nimport type { ChunkImportanceAnnotation } from \"../types.js\";\nimport { hasMembership } from \"./helpers.js\";\n\nexport function validateImportanceAnnotations(input: {\n  readonly annotations: readonly {\n    readonly chunk_id: number;\n    readonly importance: ChunkImportance;\n  }[];\n  readonly issues: string[];\n  readonly validImportanceChunkIds: Readonly<Record<string, true>> | undefined;\n}): ChunkImportanceAnnotation[] | undefined {\n  const { annotations, issues, validImportanceChunkIds } = input;\n  if (annotations.length === 0) {\n    return [];\n  }\n\n  if (validImportanceChunkIds === undefined) {\n    return annotations.map((annotation) => ({\n      chunkId: annotation.chunk_id,\n      importance: expectChunkImportance(annotation.importance),\n    }));\n  }\n\n  const result: ChunkImportanceAnnotation[] = [];\n\n  for (const annotation of annotations) {\n    if (!hasMembership(validImportanceChunkIds, annotation.chunk_id)) {\n      issues.push(\n        `importance_annotations references unknown chunk_id ${annotation.chunk_id}`,\n      );\n      continue;\n    }\n\n    result.push({\n      chunkId: annotation.chunk_id,\n      importance: expectChunkImportance(annotation.importance),\n    });\n  }\n\n  return result;\n}\n"],"mappings":"AAAA;AAAA,EACE;AAAA,OAEK;AAEP,SAAS,qBAAqB;AAEvB,SAAS,8BAA8B,OAOF;AAC1C,QAAM,EAAE,aAAa,QAAQ,wBAAwB,IAAI;AACzD,MAAI,YAAY,WAAW,GAAG;AAC5B,WAAO,CAAC;AAAA,EACV;AAEA,MAAI,4BAA4B,QAAW;AACzC,WAAO,YAAY,IAAI,CAAC,gBAAgB;AAAA,MACtC,SAAS,WAAW;AAAA,MACpB,YAAY,sBAAsB,WAAW,UAAU;AAAA,IACzD,EAAE;AAAA,EACJ;AAEA,QAAM,SAAsC,CAAC;AAE7C,aAAW,cAAc,aAAa;AACpC,QAAI,CAAC,cAAc,yBAAyB,WAAW,QAAQ,GAAG;AAChE,aAAO;AAAA,QACL,sDAAsD,WAAW,QAAQ;AAAA,MAC3E;AACA;AAAA,IACF;AAEA,WAAO,KAAK;AAAA,MACV,SAAS,WAAW;AAAA,MACpB,YAAY,sBAAsB,WAAW,UAAU;AAAA,IACzD,CAAC;AAAA,EACH;AAEA,SAAO;AACT;","names":[]}

package/dist/text/reader/chunk-batch/parser/core.d.ts

@@ -0,0 +1,33 @@
+import { SentenceTextSource, ChunkExtractionSentence, CognitiveChunk } from '../types.js';
+import { FragmentProjection } from '../fragment-projection.js';
+import { UserFocusedResponseData, BookCoherenceResponseData, ChunkMetadataField, GuaranteedChoiceRequest, ExtractChunksResult } from './schema.js';
+import '../../../../runtime/common/language.js';
+import '../../../../external/llm/client/core.js';
+import '../../../../external/llm/context.js';
+import '../../../../external/llm/cache.js';
+import '../../../../external/llm/types.js';
+import 'ai';
+import '../../../../document/types.js';
+import 'zod';
+import '../../../../graph/evidence-selection/anchor/types.js';
+
+declare class ChunkBatchParser<TData extends UserFocusedResponseData | BookCoherenceResponseData> {
+    #private;
+    constructor(input: {
+        choiceSystemPrompt: string;
+        metadataField: ChunkMetadataField;
+        projection: FragmentProjection;
+        responseIntentClassifierPrompt: string;
+        requestChoice: GuaranteedChoiceRequest;
+        sentenceTextSource: SentenceTextSource;
+        sentences: readonly ChunkExtractionSentence[];
+        validImportanceChunkIds?: readonly number[];
+        visibleChunkIds: readonly number[];
+    });
+    parse(parsedData: TData, input: {
+        isLastGenerationAttempt: boolean;
+    }): Promise<ExtractChunksResult>;
+    getChunkSourceSentences(chunk: CognitiveChunk): Promise<string[]>;
+}
+
+export { ChunkBatchParser };

package/dist/text/reader/chunk-batch/parser/core.js

@@ -0,0 +1,171 @@
+import { ParsedJsonError } from "../../../../external/guaranteed/index.js";
+import {
+  expectChunkImportance,
+  expectChunkRetention
+} from "../../../../document/index.js";
+import { validateImportanceAnnotations } from "./annotations.js";
+import {
+  createMembershipRecord,
+  createSentenceKey,
+  createSentenceTextRecord,
+  createWordsCountRecord,
+  normalizeChunkLinks
+} from "./helpers.js";
+import { filterAndValidateLinks } from "./links.js";
+import { ChunkEvidenceResolver } from "./resolver.js";
+import {
+  ChunkMetadataField
+} from "./schema.js";
+class ChunkBatchParser {
+  #evidenceResolver;
+  #metadataField;
+  #sentenceTextByKey;
+  #sentenceTextSource;
+  #wordsCountByKey;
+  #validImportanceChunkIds;
+  #visibleChunkIds;
+  constructor(input) {
+    this.#evidenceResolver = new ChunkEvidenceResolver({
+      choiceSystemPrompt: input.choiceSystemPrompt,
+      projection: input.projection,
+      requestChoice: input.requestChoice,
+      responseIntentClassifierPrompt: input.responseIntentClassifierPrompt
+    });
+    this.#metadataField = input.metadataField;
+    this.#sentenceTextSource = input.sentenceTextSource;
+    this.#visibleChunkIds = input.visibleChunkIds;
+    this.#wordsCountByKey = createWordsCountRecord(
+      input.projection.sentences.map((sentence) => ({
+        sentenceId: sentence.sentenceId,
+        wordsCount: sentence.wordsCount
+      }))
+    );
+    this.#sentenceTextByKey = createSentenceTextRecord(input.projection);
+    this.#validImportanceChunkIds = input.validImportanceChunkIds === void 0 ? void 0 : createMembershipRecord(input.validImportanceChunkIds);
+  }
+  async parse(parsedData, input) {
+    const issues = [];
+    const chunks = [];
+    const tempIds = [];
+    const importanceAnnotations = "importance_annotations" in parsedData ? validateImportanceAnnotations({
+      annotations: parsedData.importance_annotations,
+      issues,
+      validImportanceChunkIds: this.#validImportanceChunkIds
+    }) : void 0;
+    const fragmentSummary = "fragment_summary" in parsedData ? parsedData.fragment_summary : void 0;
+    const links = normalizeChunkLinks(parsedData.links);
+    for (const [index, data] of parsedData.chunks.entries()) {
+      const chunkIndex = index + 1;
+      const chunkIssues = [];
+      const label = data.label.trim();
+      const content = data.content.trim();
+      if (label === "") {
+        chunkIssues.push(
+          `Chunk #${chunkIndex}: Missing or empty "label" field`
+        );
+      }
+      if (content === "") {
+        chunkIssues.push(
+          `Chunk #${chunkIndex}: Missing or empty "content" field`
+        );
+      }
+      const [matchedSentenceIds, evidenceFailure] = await this.#evidenceResolver.resolve({
+        chunkIndex,
+        chunkLabel: label,
+        data,
+        isLastGenerationAttempt: input.isLastGenerationAttempt
+      });
+      if (matchedSentenceIds.length === 0) {
+        if (evidenceFailure !== void 0) {
+          chunkIssues.push(
+            `Chunk #${chunkIndex} ("${label}"): ${evidenceFailure.message}`
+          );
+        } else {
+          chunkIssues.push(
+            `Chunk #${chunkIndex} ("${label}"): Missing evidence`
+          );
+        }
+      }
+      if (chunkIssues.length > 0) {
+        issues.push(...chunkIssues);
+        continue;
+      }
+      const primarySentenceId = matchedSentenceIds[0];
+      if (primarySentenceId === void 0) {
+        issues.push(
+          `Chunk #${chunkIndex} ("${label}"): Unable to resolve any sentence IDs`
+        );
+        continue;
+      }
+      const totalWordsCount = matchedSentenceIds.reduce((sum, sentenceId) => {
+        return sum + (this.#wordsCountByKey[createSentenceKey(sentenceId)] ?? 0);
+      }, 0);
+      if (this.#metadataField === ChunkMetadataField.Retention) {
+        const chunkData = data;
+        chunks.push({
+          content,
+          generation: 0,
+          id: 0,
+          label,
+          links: [],
+          retention: expectChunkRetention(chunkData.retention),
+          sentenceId: primarySentenceId,
+          sentenceIds: [...matchedSentenceIds],
+          wordsCount: totalWordsCount
+        });
+      } else {
+        const chunkData = data;
+        chunks.push({
+          content,
+          generation: 0,
+          id: 0,
+          importance: expectChunkImportance(chunkData.importance),
+          label,
+          links: [],
+          sentenceId: primarySentenceId,
+          sentenceIds: [...matchedSentenceIds],
+          wordsCount: totalWordsCount
+        });
+      }
+      tempIds.push(data.temp_id);
+    }
+    const validLinks = filterAndValidateLinks({
+      issues,
+      links,
+      tempIds,
+      visibleChunkIds: this.#visibleChunkIds
+    });
+    if (issues.length > 0) {
+      throw new ParsedJsonError(issues);
+    }
+    return {
+      chunkBatch: {
+        chunks,
+        links: validLinks,
+        orderCorrect: true,
+        tempIds,
+        ...importanceAnnotations === void 0 ? {} : { importanceAnnotations }
+      },
+      ...fragmentSummary === void 0 ? {} : { fragmentSummary }
+    };
+  }
+  async getChunkSourceSentences(chunk) {
+    const sourceSentences = [];
+    for (const sentenceId of chunk.sentenceIds) {
+      const sentenceKey = createSentenceKey(sentenceId);
+      const sentenceText = this.#sentenceTextByKey[sentenceKey];
+      if (sentenceText !== void 0) {
+        sourceSentences.push(sentenceText);
+        continue;
+      }
+      sourceSentences.push(
+        await this.#sentenceTextSource.getSentence(sentenceId)
+      );
+    }
+    return sourceSentences;
+  }
+}
+export {
+  ChunkBatchParser
+};
+//# sourceMappingURL=core.js.map

package/dist/text/reader/chunk-batch/parser/core.js.map

@@ -0,0 +1 @@
+{"version":3,"sources":["../../../../../src/text/reader/chunk-batch/parser/core.ts"],"sourcesContent":["import { ParsedJsonError } from \"../../../../external/guaranteed/index.js\";\nimport {\n  expectChunkImportance,\n  expectChunkRetention,\n} from \"../../../../document/index.js\";\nimport type {\n  ChunkBatch,\n  ChunkExtractionSentence,\n  CognitiveChunk,\n  SentenceTextSource,\n} from \"../types.js\";\nimport type { FragmentProjection } from \"../fragment-projection.js\";\nimport { validateImportanceAnnotations } from \"./annotations.js\";\nimport {\n  createMembershipRecord,\n  createSentenceKey,\n  createSentenceTextRecord,\n  createWordsCountRecord,\n  normalizeChunkLinks,\n} from \"./helpers.js\";\nimport { filterAndValidateLinks } from \"./links.js\";\nimport { ChunkEvidenceResolver } from \"./resolver.js\";\nimport {\n  ChunkMetadataField,\n  type BookCoherenceChunkData,\n  type BookCoherenceResponseData,\n  type ExtractChunksResult,\n  type GuaranteedChoiceRequest,\n  type UserFocusedChunkData,\n  type UserFocusedResponseData,\n} from \"./schema.js\";\n\nexport class ChunkBatchParser<\n  TData extends UserFocusedResponseData | BookCoherenceResponseData,\n> {\n  readonly #evidenceResolver: ChunkEvidenceResolver;\n  readonly #metadataField: ChunkMetadataField;\n  readonly #sentenceTextByKey: Readonly<Record<string, string>>;\n  readonly #sentenceTextSource: SentenceTextSource;\n  readonly #wordsCountByKey: Readonly<Record<string, number>>;\n  readonly #validImportanceChunkIds: Readonly<Record<string, true>> | undefined;\n  readonly #visibleChunkIds: readonly number[];\n\n  public constructor(input: {\n    choiceSystemPrompt: string;\n    metadataField: ChunkMetadataField;\n    projection: FragmentProjection;\n    responseIntentClassifierPrompt: string;\n    requestChoice: GuaranteedChoiceRequest;\n    sentenceTextSource: SentenceTextSource;\n    sentences: readonly ChunkExtractionSentence[];\n    validImportanceChunkIds?: readonly number[];\n    visibleChunkIds: readonly number[];\n  }) {\n    this.#evidenceResolver = new ChunkEvidenceResolver({\n      choiceSystemPrompt: input.choiceSystemPrompt,\n      projection: input.projection,\n      requestChoice: input.requestChoice,\n      responseIntentClassifierPrompt: input.responseIntentClassifierPrompt,\n    });\n    this.#metadataField = input.metadataField;\n    this.#sentenceTextSource = input.sentenceTextSource;\n    this.#visibleChunkIds = input.visibleChunkIds;\n\n    this.#wordsCountByKey = createWordsCountRecord(\n      input.projection.sentences.map((sentence) => ({\n        sentenceId: sentence.sentenceId,\n        wordsCount: sentence.wordsCount,\n      })),\n    );\n    this.#sentenceTextByKey = createSentenceTextRecord(input.projection);\n    this.#validImportanceChunkIds =\n      input.validImportanceChunkIds === undefined\n        ? undefined\n        : createMembershipRecord(input.validImportanceChunkIds);\n  }\n\n  public async parse(\n    parsedData: TData,\n    input: {\n      isLastGenerationAttempt: boolean;\n    },\n  ): Promise<ExtractChunksResult> {\n    const issues: string[] = [];\n    const chunks: ChunkBatch[\"chunks\"] = [];\n    const tempIds: string[] = [];\n    const importanceAnnotations =\n      \"importance_annotations\" in parsedData\n        ? validateImportanceAnnotations({\n            annotations: parsedData.importance_annotations,\n            issues,\n            validImportanceChunkIds: this.#validImportanceChunkIds,\n          })\n        : undefined;\n    const fragmentSummary =\n      \"fragment_summary\" in parsedData\n        ? parsedData.fragment_summary\n        : undefined;\n    const links = normalizeChunkLinks(parsedData.links);\n\n    for (const [index, data] of parsedData.chunks.entries()) {\n      const chunkIndex = index + 1;\n      const chunkIssues: string[] = [];\n      const label = data.label.trim();\n      const content = data.content.trim();\n\n      if (label === \"\") {\n        chunkIssues.push(\n          `Chunk #${chunkIndex}: Missing or empty \"label\" field`,\n        );\n      }\n\n      if (content === \"\") {\n        chunkIssues.push(\n          `Chunk #${chunkIndex}: Missing or empty \"content\" field`,\n        );\n      }\n\n      const [matchedSentenceIds, evidenceFailure] =\n        await this.#evidenceResolver.resolve({\n          chunkIndex,\n          chunkLabel: label,\n          data,\n          isLastGenerationAttempt: input.isLastGenerationAttempt,\n        });\n\n      if (matchedSentenceIds.length === 0) {\n        if (evidenceFailure !== undefined) {\n          chunkIssues.push(\n            `Chunk #${chunkIndex} (\"${label}\"): ${evidenceFailure.message}`,\n          );\n        } else {\n          chunkIssues.push(\n            `Chunk #${chunkIndex} (\"${label}\"): Missing evidence`,\n          );\n        }\n      }\n\n      if (chunkIssues.length > 0) {\n        issues.push(...chunkIssues);\n        continue;\n      }\n\n      const primarySentenceId = matchedSentenceIds[0];\n\n      if (primarySentenceId === undefined) {\n        issues.push(\n          `Chunk #${chunkIndex} (\"${label}\"): Unable to resolve any sentence IDs`,\n        );\n        continue;\n      }\n\n      const totalWordsCount = matchedSentenceIds.reduce((sum, sentenceId) => {\n        return (\n          sum + (this.#wordsCountByKey[createSentenceKey(sentenceId)] ?? 0)\n        );\n      }, 0);\n\n      if (this.#metadataField === ChunkMetadataField.Retention) {\n        const chunkData = data as UserFocusedChunkData;\n\n        chunks.push({\n          content,\n          generation: 0,\n          id: 0,\n          label,\n          links: [],\n          retention: expectChunkRetention(chunkData.retention),\n          sentenceId: primarySentenceId,\n          sentenceIds: [...matchedSentenceIds],\n          wordsCount: totalWordsCount,\n        });\n      } else {\n        const chunkData = data as BookCoherenceChunkData;\n\n        chunks.push({\n          content,\n          generation: 0,\n          id: 0,\n          importance: expectChunkImportance(chunkData.importance),\n          label,\n          links: [],\n          sentenceId: primarySentenceId,\n          sentenceIds: [...matchedSentenceIds],\n          wordsCount: totalWordsCount,\n        });\n      }\n\n      tempIds.push(data.temp_id);\n    }\n\n    const validLinks = filterAndValidateLinks({\n      issues,\n      links,\n      tempIds,\n      visibleChunkIds: this.#visibleChunkIds,\n    });\n\n    if (issues.length > 0) {\n      throw new ParsedJsonError(issues);\n    }\n\n    return {\n      chunkBatch: {\n        chunks,\n        links: validLinks,\n        orderCorrect: true,\n        tempIds,\n        ...(importanceAnnotations === undefined\n          ? {}\n          : { importanceAnnotations }),\n      },\n      ...(fragmentSummary === undefined ? {} : { fragmentSummary }),\n    };\n  }\n\n  public async getChunkSourceSentences(\n    chunk: CognitiveChunk,\n  ): Promise<string[]> {\n    const sourceSentences: string[] = [];\n\n    for (const sentenceId of chunk.sentenceIds) {\n      const sentenceKey = createSentenceKey(sentenceId);\n      const sentenceText = this.#sentenceTextByKey[sentenceKey];\n\n      if (sentenceText !== undefined) {\n        sourceSentences.push(sentenceText);\n        continue;\n      }\n\n      sourceSentences.push(\n        await this.#sentenceTextSource.getSentence(sentenceId),\n      );\n    }\n\n    return sourceSentences;\n  }\n}\n"],"mappings":"AAAA,SAAS,uBAAuB;AAChC;AAAA,EACE;AAAA,EACA;AAAA,OACK;AAQP,SAAS,qCAAqC;AAC9C;AAAA,EACE;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,OACK;AACP,SAAS,8BAA8B;AACvC,SAAS,6BAA6B;AACtC;AAAA,EACE;AAAA,OAOK;AAEA,MAAM,iBAEX;AAAA,EACS;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EAEF,YAAY,OAUhB;AACD,SAAK,oBAAoB,IAAI,sBAAsB;AAAA,MACjD,oBAAoB,MAAM;AAAA,MAC1B,YAAY,MAAM;AAAA,MAClB,eAAe,MAAM;AAAA,MACrB,gCAAgC,MAAM;AAAA,IACxC,CAAC;AACD,SAAK,iBAAiB,MAAM;AAC5B,SAAK,sBAAsB,MAAM;AACjC,SAAK,mBAAmB,MAAM;AAE9B,SAAK,mBAAmB;AAAA,MACtB,MAAM,WAAW,UAAU,IAAI,CAAC,cAAc;AAAA,QAC5C,YAAY,SAAS;AAAA,QACrB,YAAY,SAAS;AAAA,MACvB,EAAE;AAAA,IACJ;AACA,SAAK,qBAAqB,yBAAyB,MAAM,UAAU;AACnE,SAAK,2BACH,MAAM,4BAA4B,SAC9B,SACA,uBAAuB,MAAM,uBAAuB;AAAA,EAC5D;AAAA,EAEA,MAAa,MACX,YACA,OAG8B;AAC9B,UAAM,SAAmB,CAAC;AAC1B,UAAM,SAA+B,CAAC;AACtC,UAAM,UAAoB,CAAC;AAC3B,UAAM,wBACJ,4BAA4B,aACxB,8BAA8B;AAAA,MAC5B,aAAa,WAAW;AAAA,MACxB;AAAA,MACA,yBAAyB,KAAK;AAAA,IAChC,CAAC,IACD;AACN,UAAM,kBACJ,sBAAsB,aAClB,WAAW,mBACX;AACN,UAAM,QAAQ,oBAAoB,WAAW,KAAK;AAElD,eAAW,CAAC,OAAO,IAAI,KAAK,WAAW,OAAO,QAAQ,GAAG;AACvD,YAAM,aAAa,QAAQ;AAC3B,YAAM,cAAwB,CAAC;AAC/B,YAAM,QAAQ,KAAK,MAAM,KAAK;AAC9B,YAAM,UAAU,KAAK,QAAQ,KAAK;AAElC,UAAI,UAAU,IAAI;AAChB,oBAAY;AAAA,UACV,UAAU,UAAU;AAAA,QACtB;AAAA,MACF;AAEA,UAAI,YAAY,IAAI;AAClB,oBAAY;AAAA,UACV,UAAU,UAAU;AAAA,QACtB;AAAA,MACF;AAEA,YAAM,CAAC,oBAAoB,eAAe,IACxC,MAAM,KAAK,kBAAkB,QAAQ;AAAA,QACnC;AAAA,QACA,YAAY;AAAA,QACZ;AAAA,QACA,yBAAyB,MAAM;AAAA,MACjC,CAAC;AAEH,UAAI,mBAAmB,WAAW,GAAG;AACnC,YAAI,oBAAoB,QAAW;AACjC,sBAAY;AAAA,YACV,UAAU,UAAU,MAAM,KAAK,OAAO,gBAAgB,OAAO;AAAA,UAC/D;AAAA,QACF,OAAO;AACL,sBAAY;AAAA,YACV,UAAU,UAAU,MAAM,KAAK;AAAA,UACjC;AAAA,QACF;AAAA,MACF;AAEA,UAAI,YAAY,SAAS,GAAG;AAC1B,eAAO,KAAK,GAAG,WAAW;AAC1B;AAAA,MACF;AAEA,YAAM,oBAAoB,mBAAmB,CAAC;AAE9C,UAAI,sBAAsB,QAAW;AACnC,eAAO;AAAA,UACL,UAAU,UAAU,MAAM,KAAK;AAAA,QACjC;AACA;AAAA,MACF;AAEA,YAAM,kBAAkB,mBAAmB,OAAO,CAAC,KAAK,eAAe;AACrE,eACE,OAAO,KAAK,iBAAiB,kBAAkB,UAAU,CAAC,KAAK;AAAA,MAEnE,GAAG,CAAC;AAEJ,UAAI,KAAK,mBAAmB,mBAAmB,WAAW;AACxD,cAAM,YAAY;AAElB,eAAO,KAAK;AAAA,UACV;AAAA,UACA,YAAY;AAAA,UACZ,IAAI;AAAA,UACJ;AAAA,UACA,OAAO,CAAC;AAAA,UACR,WAAW,qBAAqB,UAAU,SAAS;AAAA,UACnD,YAAY;AAAA,UACZ,aAAa,CAAC,GAAG,kBAAkB;AAAA,UACnC,YAAY;AAAA,QACd,CAAC;AAAA,MACH,OAAO;AACL,cAAM,YAAY;AAElB,eAAO,KAAK;AAAA,UACV;AAAA,UACA,YAAY;AAAA,UACZ,IAAI;AAAA,UACJ,YAAY,sBAAsB,UAAU,UAAU;AAAA,UACtD;AAAA,UACA,OAAO,CAAC;AAAA,UACR,YAAY;AAAA,UACZ,aAAa,CAAC,GAAG,kBAAkB;AAAA,UACnC,YAAY;AAAA,QACd,CAAC;AAAA,MACH;AAEA,cAAQ,KAAK,KAAK,OAAO;AAAA,IAC3B;AAEA,UAAM,aAAa,uBAAuB;AAAA,MACxC;AAAA,MACA;AAAA,MACA;AAAA,MACA,iBAAiB,KAAK;AAAA,IACxB,CAAC;AAED,QAAI,OAAO,SAAS,GAAG;AACrB,YAAM,IAAI,gBAAgB,MAAM;AAAA,IAClC;AAEA,WAAO;AAAA,MACL,YAAY;AAAA,QACV;AAAA,QACA,OAAO;AAAA,QACP,cAAc;AAAA,QACd;AAAA,QACA,GAAI,0BAA0B,SAC1B,CAAC,IACD,EAAE,sBAAsB;AAAA,MAC9B;AAAA,MACA,GAAI,oBAAoB,SAAY,CAAC,IAAI,EAAE,gBAAgB;AAAA,IAC7D;AAAA,EACF;AAAA,EAEA,MAAa,wBACX,OACmB;AACnB,UAAM,kBAA4B,CAAC;AAEnC,eAAW,cAAc,MAAM,aAAa;AAC1C,YAAM,cAAc,kBAAkB,UAAU;AAChD,YAAM,eAAe,KAAK,mBAAmB,WAAW;AAExD,UAAI,iBAAiB,QAAW;AAC9B,wBAAgB,KAAK,YAAY;AACjC;AAAA,MACF;AAEA,sBAAgB;AAAA,QACd,MAAM,KAAK,oBAAoB,YAAY,UAAU;AAAA,MACvD;AAAA,IACF;AAEA,WAAO;AAAA,EACT;AACF;","names":[]}

package/dist/text/reader/chunk-batch/parser/evidence.d.ts

@@ -0,0 +1,21 @@
+import { EvidenceResolutionFailure } from '../../../../graph/evidence-selection/anchor/types.js';
+import { EvidenceSelectionList, EvidenceSelectionCandidate } from '../../../../graph/evidence-selection/types.js';
+import { RawChunkEvidence } from './schema.js';
+import '../../../../document/types.js';
+import 'zod';
+import '../../../../external/llm/types.js';
+import 'ai';
+import '../types.js';
+import '../../../../runtime/common/language.js';
+import '../../../../external/llm/client/core.js';
+import '../../../../external/llm/context.js';
+import '../../../../external/llm/cache.js';
+
+declare function createEvidenceSelectionList(evidence: RawChunkEvidence): EvidenceSelectionList | undefined;
+declare function toEvidenceResolutionFailure(failure: {
+    readonly candidates: readonly EvidenceSelectionCandidate[];
+    readonly code: string;
+    readonly message: string;
+}, fieldName: string): EvidenceResolutionFailure;
+
+export { createEvidenceSelectionList, toEvidenceResolutionFailure };

package/dist/text/reader/chunk-batch/parser/evidence.js

@@ -0,0 +1,40 @@
+function createEvidenceSelectionList(evidence) {
+  if (Array.isArray(evidence)) {
+    return evidence.map(createEvidenceSelection);
+  }
+  const hasSelectionEvidence = typeof evidence.quote === "string" || typeof evidence.sentence_id === "string";
+  return hasSelectionEvidence ? createEvidenceSelection(evidence) : void 0;
+}
+function createEvidenceSelection(evidence) {
+  return {
+    ...typeof evidence.quote === "string" ? { quote: evidence.quote } : {},
+    ...typeof evidence.sentence_id === "string" ? { sentence_id: evidence.sentence_id } : {}
+  };
+}
+function toEvidenceResolutionFailure(failure, fieldName) {
+  return {
+    candidates: failure.candidates.map(toRankedSentenceCandidate),
+    code: failure.code,
+    fieldName,
+    message: failure.message
+  };
+}
+function toRankedSentenceCandidate(candidate) {
+  return {
+    exactNormalized: candidate.exactNormalized,
+    exactRaw: candidate.exactRaw,
+    exactSubstring: candidate.exactSubstring,
+    index: candidate.index,
+    nextText: candidate.nextText,
+    occurrenceId: candidate.occurrenceId,
+    prevText: candidate.prevText,
+    score: candidate.score,
+    sentenceId: candidate.sentence.sentenceId,
+    text: candidate.sentence.text
+  };
+}
+export {
+  createEvidenceSelectionList,
+  toEvidenceResolutionFailure
+};
+//# sourceMappingURL=evidence.js.map

package/dist/text/reader/chunk-batch/parser/evidence.js.map

@@ -0,0 +1 @@
+{"version":3,"sources":["../../../../../src/text/reader/chunk-batch/parser/evidence.ts"],"sourcesContent":["import type {\n  EvidenceSelectionCandidate,\n  EvidenceSelectionList,\n  EvidenceResolutionFailure,\n  RankedSentenceCandidate,\n} from \"../../../../graph/evidence-selection/index.js\";\nimport type { RawChunkEvidence } from \"./schema.js\";\n\nexport function createEvidenceSelectionList(\n  evidence: RawChunkEvidence,\n): EvidenceSelectionList | undefined {\n  if (Array.isArray(evidence)) {\n    return evidence.map(createEvidenceSelection);\n  }\n\n  const hasSelectionEvidence =\n    typeof evidence.quote === \"string\" ||\n    typeof evidence.sentence_id === \"string\";\n\n  return hasSelectionEvidence ? createEvidenceSelection(evidence) : undefined;\n}\n\nfunction createEvidenceSelection(evidence: {\n  readonly quote?: unknown;\n  readonly sentence_id?: unknown;\n}): {\n  readonly quote?: string;\n  readonly sentence_id?: string;\n} {\n  return {\n    ...(typeof evidence.quote === \"string\" ? { quote: evidence.quote } : {}),\n    ...(typeof evidence.sentence_id === \"string\"\n      ? { sentence_id: evidence.sentence_id }\n      : {}),\n  };\n}\n\nexport function toEvidenceResolutionFailure(\n  failure: {\n    readonly candidates: readonly EvidenceSelectionCandidate[];\n    readonly code: string;\n    readonly message: string;\n  },\n  fieldName: string,\n): EvidenceResolutionFailure {\n  return {\n    candidates: failure.candidates.map(toRankedSentenceCandidate),\n    code: failure.code,\n    fieldName,\n    message: failure.message,\n  };\n}\n\nfunction toRankedSentenceCandidate(\n  candidate: EvidenceSelectionCandidate,\n): RankedSentenceCandidate {\n  return {\n    exactNormalized: candidate.exactNormalized,\n    exactRaw: candidate.exactRaw,\n    exactSubstring: candidate.exactSubstring,\n    index: candidate.index,\n    nextText: candidate.nextText,\n    occurrenceId: candidate.occurrenceId,\n    prevText: candidate.prevText,\n    score: candidate.score,\n    sentenceId: candidate.sentence.sentenceId,\n    text: candidate.sentence.text,\n  };\n}\n"],"mappings":"AAQO,SAAS,4BACd,UACmC;AACnC,MAAI,MAAM,QAAQ,QAAQ,GAAG;AAC3B,WAAO,SAAS,IAAI,uBAAuB;AAAA,EAC7C;AAEA,QAAM,uBACJ,OAAO,SAAS,UAAU,YAC1B,OAAO,SAAS,gBAAgB;AAElC,SAAO,uBAAuB,wBAAwB,QAAQ,IAAI;AACpE;AAEA,SAAS,wBAAwB,UAM/B;AACA,SAAO;AAAA,IACL,GAAI,OAAO,SAAS,UAAU,WAAW,EAAE,OAAO,SAAS,MAAM,IAAI,CAAC;AAAA,IACtE,GAAI,OAAO,SAAS,gBAAgB,WAChC,EAAE,aAAa,SAAS,YAAY,IACpC,CAAC;AAAA,EACP;AACF;AAEO,SAAS,4BACd,SAKA,WAC2B;AAC3B,SAAO;AAAA,IACL,YAAY,QAAQ,WAAW,IAAI,yBAAyB;AAAA,IAC5D,MAAM,QAAQ;AAAA,IACd;AAAA,IACA,SAAS,QAAQ;AAAA,EACnB;AACF;AAEA,SAAS,0BACP,WACyB;AACzB,SAAO;AAAA,IACL,iBAAiB,UAAU;AAAA,IAC3B,UAAU,UAAU;AAAA,IACpB,gBAAgB,UAAU;AAAA,IAC1B,OAAO,UAAU;AAAA,IACjB,UAAU,UAAU;AAAA,IACpB,cAAc,UAAU;AAAA,IACxB,UAAU,UAAU;AAAA,IACpB,OAAO,UAAU;AAAA,IACjB,YAAY,UAAU,SAAS;AAAA,IAC/B,MAAM,UAAU,SAAS;AAAA,EAC3B;AACF;","names":[]}

package/dist/text/reader/chunk-batch/parser/helpers.d.ts

@@ -0,0 +1,27 @@
+import { SentenceId } from '../../../../document/types.js';
+import { RankedSentenceCandidate } from '../../../../graph/evidence-selection/anchor/types.js';
+import { ChunkExtractionSentence, ChunkLink } from '../types.js';
+import { FragmentProjection, TextSpan } from '../fragment-projection.js';
+import { RawChunkLink, ChoiceFieldName } from './schema.js';
+import '../../../../runtime/common/language.js';
+import '../../../../external/llm/client/core.js';
+import '../../../../external/llm/context.js';
+import '../../../../external/llm/cache.js';
+import '../../../../external/llm/types.js';
+import 'ai';
+import 'zod';
+
+declare function createWordsCountRecord(sentences: readonly Pick<ChunkExtractionSentence, "sentenceId" | "wordsCount">[]): Readonly<Record<string, number>>;
+declare function createSentenceTextRecord(projection: FragmentProjection): Readonly<Record<string, string>>;
+declare function normalizeChunkLinks(links: readonly RawChunkLink[]): ChunkLink[];
+declare function formatChoiceCandidate(candidate: RankedSentenceCandidate): string;
+declare function toChoiceFieldName(value: string): ChoiceFieldName | undefined;
+declare function isRecord(value: unknown): value is Record<string, unknown>;
+declare function formatError(error: unknown): string;
+declare function isParsedJsonValidationFailure(error: unknown): boolean;
+declare function createSentenceKey(sentenceId: SentenceId): string;
+declare function createMembershipRecord(values: readonly (number | string)[]): Readonly<Record<string, true>>;
+declare function hasMembership(record: Readonly<Record<string, true>>, value: number | string): boolean;
+declare function expectSingleSpan(spans: readonly TextSpan[]): TextSpan | undefined;
+
+export { createMembershipRecord, createSentenceKey, createSentenceTextRecord, createWordsCountRecord, expectSingleSpan, formatChoiceCandidate, formatError, hasMembership, isParsedJsonValidationFailure, isRecord, normalizeChunkLinks, toChoiceFieldName };

package/dist/text/reader/chunk-batch/parser/helpers.js

@@ -0,0 +1,98 @@
+import {
+  GuaranteedParseValidationError,
+  ParsedJsonError
+} from "../../../../external/guaranteed/index.js";
+function createWordsCountRecord(sentences) {
+  const wordsCountByKey = createEmptyRecord();
+  for (const sentence of sentences) {
+    const sentenceKey = createSentenceKey(sentence.sentenceId);
+    wordsCountByKey[sentenceKey] = sentence.wordsCount;
+  }
+  return wordsCountByKey;
+}
+function createSentenceTextRecord(projection) {
+  const record = createEmptyRecord();
+  for (const sentence of projection.sentences) {
+    record[createSentenceKey(sentence.sentenceId)] = sentence.rawText;
+  }
+  return record;
+}
+function normalizeChunkLinks(links) {
+  return links.map((link) => {
+    if (link.strength === void 0) {
+      return {
+        from: link.from,
+        to: link.to
+      };
+    }
+    return {
+      from: link.from,
+      strength: link.strength,
+      to: link.to
+    };
+  });
+}
+function formatChoiceCandidate(candidate) {
+  return [
+    candidate.occurrenceId,
+    `prev: ${formatChoiceText(candidate.prevText)}`,
+    `text: ${formatChoiceText(candidate.text)}`,
+    `next: ${formatChoiceText(candidate.nextText)}`
+  ].join("\n");
+}
+function formatChoiceText(text) {
+  const collapsed = text.replace(/\s+/gu, " ").trim();
+  return collapsed === "" ? "(none)" : collapsed;
+}
+function toChoiceFieldName(value) {
+  return value === "evidence" || value === "start_anchor" || value === "end_anchor" ? value : void 0;
+}
+function isRecord(value) {
+  return typeof value === "object" && value !== null;
+}
+function formatError(error) {
+  if (error instanceof Error) {
+    return error.message;
+  }
+  return String(error);
+}
+function isParsedJsonValidationFailure(error) {
+  return error instanceof GuaranteedParseValidationError && error.cause instanceof ParsedJsonError;
+}
+function createSentenceKey(sentenceId) {
+  return sentenceId.join(":");
+}
+function createMembershipRecord(values) {
+  const record = createEmptyRecord();
+  for (const value of values) {
+    record[String(value)] = true;
+  }
+  return record;
+}
+function hasMembership(record, value) {
+  return hasIndexedValue(record, String(value));
+}
+function hasIndexedValue(record, key) {
+  return Object.hasOwn(record, key);
+}
+function createEmptyRecord() {
+  return /* @__PURE__ */ Object.create(null);
+}
+function expectSingleSpan(spans) {
+  return spans.length === 1 ? spans[0] : void 0;
+}
+export {
+  createMembershipRecord,
+  createSentenceKey,
+  createSentenceTextRecord,
+  createWordsCountRecord,
+  expectSingleSpan,
+  formatChoiceCandidate,
+  formatError,
+  hasMembership,
+  isParsedJsonValidationFailure,
+  isRecord,
+  normalizeChunkLinks,
+  toChoiceFieldName
+};
+//# sourceMappingURL=helpers.js.map

package/dist/text/reader/chunk-batch/parser/helpers.js.map

@@ -0,0 +1 @@
+{"version":3,"sources":["../../../../../src/text/reader/chunk-batch/parser/helpers.ts"],"sourcesContent":["import {\n  GuaranteedParseValidationError,\n  ParsedJsonError,\n} from \"../../../../external/guaranteed/index.js\";\nimport type { SentenceId } from \"../../../../document/index.js\";\nimport type { RankedSentenceCandidate } from \"../../../../graph/evidence-selection/index.js\";\nimport type { ChunkExtractionSentence, ChunkLink } from \"../types.js\";\nimport type { FragmentProjection, TextSpan } from \"../fragment-projection.js\";\nimport type { ChoiceFieldName, RawChunkLink } from \"./schema.js\";\n\nexport function createWordsCountRecord(\n  sentences: readonly Pick<\n    ChunkExtractionSentence,\n    \"sentenceId\" | \"wordsCount\"\n  >[],\n): Readonly<Record<string, number>> {\n  const wordsCountByKey = createEmptyRecord<number>();\n\n  for (const sentence of sentences) {\n    const sentenceKey = createSentenceKey(sentence.sentenceId);\n    wordsCountByKey[sentenceKey] = sentence.wordsCount;\n  }\n\n  return wordsCountByKey;\n}\n\nexport function createSentenceTextRecord(\n  projection: FragmentProjection,\n): Readonly<Record<string, string>> {\n  const record = createEmptyRecord<string>();\n\n  for (const sentence of projection.sentences) {\n    record[createSentenceKey(sentence.sentenceId)] = sentence.rawText;\n  }\n\n  return record;\n}\n\nexport function normalizeChunkLinks(\n  links: readonly RawChunkLink[],\n): ChunkLink[] {\n  return links.map((link) => {\n    if (link.strength === undefined) {\n      return {\n        from: link.from,\n        to: link.to,\n      };\n    }\n\n    return {\n      from: link.from,\n      strength: link.strength,\n      to: link.to,\n    };\n  });\n}\n\nexport function formatChoiceCandidate(\n  candidate: RankedSentenceCandidate,\n): string {\n  return [\n    candidate.occurrenceId,\n    `prev: ${formatChoiceText(candidate.prevText)}`,\n    `text: ${formatChoiceText(candidate.text)}`,\n    `next: ${formatChoiceText(candidate.nextText)}`,\n  ].join(\"\\n\");\n}\n\nfunction formatChoiceText(text: string): string {\n  const collapsed = text.replace(/\\s+/gu, \" \").trim();\n\n  return collapsed === \"\" ? \"(none)\" : collapsed;\n}\n\nexport function toChoiceFieldName(value: string): ChoiceFieldName | undefined {\n  return value === \"evidence\" ||\n    value === \"start_anchor\" ||\n    value === \"end_anchor\"\n    ? value\n    : undefined;\n}\n\nexport function isRecord(value: unknown): value is Record<string, unknown> {\n  return typeof value === \"object\" && value !== null;\n}\n\nexport function formatError(error: unknown): string {\n  if (error instanceof Error) {\n    return error.message;\n  }\n\n  return String(error);\n}\n\nexport function isParsedJsonValidationFailure(error: unknown): boolean {\n  return (\n    error instanceof GuaranteedParseValidationError &&\n    error.cause instanceof ParsedJsonError\n  );\n}\n\nexport function createSentenceKey(sentenceId: SentenceId): string {\n  return sentenceId.join(\":\");\n}\n\nexport function createMembershipRecord(\n  values: readonly (number | string)[],\n): Readonly<Record<string, true>> {\n  const record = createEmptyRecord<true>();\n\n  for (const value of values) {\n    record[String(value)] = true;\n  }\n\n  return record;\n}\n\nexport function hasMembership(\n  record: Readonly<Record<string, true>>,\n  value: number | string,\n): boolean {\n  return hasIndexedValue(record, String(value));\n}\n\nfunction hasIndexedValue<TValue>(\n  record: Readonly<Record<string, TValue>>,\n  key: string,\n): boolean {\n  return Object.hasOwn(record, key);\n}\n\nfunction createEmptyRecord<TValue>(): Record<string, TValue> {\n  return Object.create(null) as Record<string, TValue>;\n}\n\nexport function expectSingleSpan(\n  spans: readonly TextSpan[],\n): TextSpan | undefined {\n  return spans.length === 1 ? spans[0] : undefined;\n}\n"],"mappings":"AAAA;AAAA,EACE;AAAA,EACA;AAAA,OACK;AAOA,SAAS,uBACd,WAIkC;AAClC,QAAM,kBAAkB,kBAA0B;AAElD,aAAW,YAAY,WAAW;AAChC,UAAM,cAAc,kBAAkB,SAAS,UAAU;AACzD,oBAAgB,WAAW,IAAI,SAAS;AAAA,EAC1C;AAEA,SAAO;AACT;AAEO,SAAS,yBACd,YACkC;AAClC,QAAM,SAAS,kBAA0B;AAEzC,aAAW,YAAY,WAAW,WAAW;AAC3C,WAAO,kBAAkB,SAAS,UAAU,CAAC,IAAI,SAAS;AAAA,EAC5D;AAEA,SAAO;AACT;AAEO,SAAS,oBACd,OACa;AACb,SAAO,MAAM,IAAI,CAAC,SAAS;AACzB,QAAI,KAAK,aAAa,QAAW;AAC/B,aAAO;AAAA,QACL,MAAM,KAAK;AAAA,QACX,IAAI,KAAK;AAAA,MACX;AAAA,IACF;AAEA,WAAO;AAAA,MACL,MAAM,KAAK;AAAA,MACX,UAAU,KAAK;AAAA,MACf,IAAI,KAAK;AAAA,IACX;AAAA,EACF,CAAC;AACH;AAEO,SAAS,sBACd,WACQ;AACR,SAAO;AAAA,IACL,UAAU;AAAA,IACV,SAAS,iBAAiB,UAAU,QAAQ,CAAC;AAAA,IAC7C,SAAS,iBAAiB,UAAU,IAAI,CAAC;AAAA,IACzC,SAAS,iBAAiB,UAAU,QAAQ,CAAC;AAAA,EAC/C,EAAE,KAAK,IAAI;AACb;AAEA,SAAS,iBAAiB,MAAsB;AAC9C,QAAM,YAAY,KAAK,QAAQ,SAAS,GAAG,EAAE,KAAK;AAElD,SAAO,cAAc,KAAK,WAAW;AACvC;AAEO,SAAS,kBAAkB,OAA4C;AAC5E,SAAO,UAAU,cACf,UAAU,kBACV,UAAU,eACR,QACA;AACN;AAEO,SAAS,SAAS,OAAkD;AACzE,SAAO,OAAO,UAAU,YAAY,UAAU;AAChD;AAEO,SAAS,YAAY,OAAwB;AAClD,MAAI,iBAAiB,OAAO;AAC1B,WAAO,MAAM;AAAA,EACf;AAEA,SAAO,OAAO,KAAK;AACrB;AAEO,SAAS,8BAA8B,OAAyB;AACrE,SACE,iBAAiB,kCACjB,MAAM,iBAAiB;AAE3B;AAEO,SAAS,kBAAkB,YAAgC;AAChE,SAAO,WAAW,KAAK,GAAG;AAC5B;AAEO,SAAS,uBACd,QACgC;AAChC,QAAM,SAAS,kBAAwB;AAEvC,aAAW,SAAS,QAAQ;AAC1B,WAAO,OAAO,KAAK,CAAC,IAAI;AAAA,EAC1B;AAEA,SAAO;AACT;AAEO,SAAS,cACd,QACA,OACS;AACT,SAAO,gBAAgB,QAAQ,OAAO,KAAK,CAAC;AAC9C;AAEA,SAAS,gBACP,QACA,KACS;AACT,SAAO,OAAO,OAAO,QAAQ,GAAG;AAClC;AAEA,SAAS,oBAAoD;AAC3D,SAAO,uBAAO,OAAO,IAAI;AAC3B;AAEO,SAAS,iBACd,OACsB;AACtB,SAAO,MAAM,WAAW,IAAI,MAAM,CAAC,IAAI;AACzC;","names":[]}

package/dist/text/reader/chunk-batch/parser/index.d.ts

@@ -0,0 +1,13 @@
+export { ChunkBatchParser } from './core.js';
+export { BookCoherenceChunkData, BookCoherenceResponseData, ChoiceFieldName, ChunkMetadataField, ExtractChunksResult, ExtractedChunkData, GuaranteedChoiceRequest, RawChunkEvidence, RawChunkLink, ResolveChunkEvidenceInput, SelectAmbiguousCandidateInput, UserFocusedChunkData, UserFocusedResponseData, bookCoherenceResponseSchema, choiceResponseSchema, userFocusedResponseSchema } from './schema.js';
+import '../types.js';
+import '../../../../runtime/common/language.js';
+import '../../../../external/llm/client/core.js';
+import '../../../../external/llm/context.js';
+import '../../../../external/llm/cache.js';
+import '../../../../external/llm/types.js';
+import 'ai';
+import '../../../../document/types.js';
+import '../fragment-projection.js';
+import 'zod';
+import '../../../../graph/evidence-selection/anchor/types.js';

package/dist/text/reader/chunk-batch/parser/index.js

@@ -0,0 +1,3 @@
+export * from "./core.js";
+export * from "./schema.js";
+//# sourceMappingURL=index.js.map

package/dist/text/reader/chunk-batch/parser/index.js.map

@@ -0,0 +1 @@
+{"version":3,"sources":["../../../../../src/text/reader/chunk-batch/parser/index.ts"],"sourcesContent":["export * from \"./core.js\";\nexport * from \"./schema.js\";\n"],"mappings":"AAAA,cAAc;AACd,cAAc;","names":[]}

package/dist/text/reader/chunk-batch/parser/links.d.ts

@@ -0,0 +1,17 @@
+import { ChunkLink } from '../types.js';
+import '../../../../runtime/common/language.js';
+import '../../../../external/llm/client/core.js';
+import '../../../../external/llm/context.js';
+import '../../../../external/llm/cache.js';
+import '../../../../external/llm/types.js';
+import 'ai';
+import '../../../../document/types.js';
+
+declare function filterAndValidateLinks(input: {
+    readonly issues: string[];
+    readonly links: readonly ChunkLink[];
+    readonly tempIds: readonly string[];
+    readonly visibleChunkIds: readonly number[];
+}): ChunkLink[];
+
+export { filterAndValidateLinks };

package/dist/text/reader/chunk-batch/parser/links.js

@@ -0,0 +1,44 @@
+import { createMembershipRecord, hasMembership } from "./helpers.js";
+function filterAndValidateLinks(input) {
+  const validTempIds = createMembershipRecord(input.tempIds);
+  const validChunkIds = createMembershipRecord(input.visibleChunkIds);
+  const retainedLinks = [];
+  for (const [index, link] of input.links.entries()) {
+    const fromValid = validateLinkReference({
+      fieldName: "from",
+      index: index + 1,
+      issues: input.issues,
+      reference: link.from,
+      validChunkIds,
+      validTempIds
+    });
+    const toValid = validateLinkReference({
+      fieldName: "to",
+      index: index + 1,
+      issues: input.issues,
+      reference: link.to,
+      validChunkIds,
+      validTempIds
+    });
+    if (fromValid && toValid) {
+      retainedLinks.push(link);
+    }
+  }
+  return retainedLinks;
+}
+function validateLinkReference(input) {
+  if (typeof input.reference === "string") {
+    return hasMembership(input.validTempIds, input.reference);
+  }
+  if (!hasMembership(input.validChunkIds, input.reference)) {
+    input.issues.push(
+      `Link #${input.index}: "${input.fieldName}" chunk_id ${input.reference} does not exist in visible chunks`
+    );
+    return false;
+  }
+  return true;
+}
+export {
+  filterAndValidateLinks
+};
+//# sourceMappingURL=links.js.map

package/dist/text/reader/chunk-batch/parser/links.js.map

@@ -0,0 +1 @@
+{"version":3,"sources":["../../../../../src/text/reader/chunk-batch/parser/links.ts"],"sourcesContent":["import type { ChunkLink } from \"../types.js\";\nimport { createMembershipRecord, hasMembership } from \"./helpers.js\";\n\nexport function filterAndValidateLinks(input: {\n  readonly issues: string[];\n  readonly links: readonly ChunkLink[];\n  readonly tempIds: readonly string[];\n  readonly visibleChunkIds: readonly number[];\n}): ChunkLink[] {\n  const validTempIds = createMembershipRecord(input.tempIds);\n  const validChunkIds = createMembershipRecord(input.visibleChunkIds);\n  const retainedLinks: ChunkLink[] = [];\n\n  for (const [index, link] of input.links.entries()) {\n    const fromValid = validateLinkReference({\n      fieldName: \"from\",\n      index: index + 1,\n      issues: input.issues,\n      reference: link.from,\n      validChunkIds,\n      validTempIds,\n    });\n    const toValid = validateLinkReference({\n      fieldName: \"to\",\n      index: index + 1,\n      issues: input.issues,\n      reference: link.to,\n      validChunkIds,\n      validTempIds,\n    });\n\n    if (fromValid && toValid) {\n      retainedLinks.push(link);\n    }\n  }\n\n  return retainedLinks;\n}\n\nfunction validateLinkReference(input: {\n  fieldName: \"from\" | \"to\";\n  index: number;\n  issues: string[];\n  reference: number | string;\n  validChunkIds: Readonly<Record<string, true>>;\n  validTempIds: Readonly<Record<string, true>>;\n}): boolean {\n  if (typeof input.reference === \"string\") {\n    return hasMembership(input.validTempIds, input.reference);\n  }\n\n  if (!hasMembership(input.validChunkIds, input.reference)) {\n    input.issues.push(\n      `Link #${input.index}: \"${input.fieldName}\" chunk_id ${input.reference} does not exist in visible chunks`,\n    );\n\n    return false;\n  }\n\n  return true;\n}\n"],"mappings":"AACA,SAAS,wBAAwB,qBAAqB;AAE/C,SAAS,uBAAuB,OAKvB;AACd,QAAM,eAAe,uBAAuB,MAAM,OAAO;AACzD,QAAM,gBAAgB,uBAAuB,MAAM,eAAe;AAClE,QAAM,gBAA6B,CAAC;AAEpC,aAAW,CAAC,OAAO,IAAI,KAAK,MAAM,MAAM,QAAQ,GAAG;AACjD,UAAM,YAAY,sBAAsB;AAAA,MACtC,WAAW;AAAA,MACX,OAAO,QAAQ;AAAA,MACf,QAAQ,MAAM;AAAA,MACd,WAAW,KAAK;AAAA,MAChB;AAAA,MACA;AAAA,IACF,CAAC;AACD,UAAM,UAAU,sBAAsB;AAAA,MACpC,WAAW;AAAA,MACX,OAAO,QAAQ;AAAA,MACf,QAAQ,MAAM;AAAA,MACd,WAAW,KAAK;AAAA,MAChB;AAAA,MACA;AAAA,IACF,CAAC;AAED,QAAI,aAAa,SAAS;AACxB,oBAAc,KAAK,IAAI;AAAA,IACzB;AAAA,EACF;AAEA,SAAO;AACT;AAEA,SAAS,sBAAsB,OAOnB;AACV,MAAI,OAAO,MAAM,cAAc,UAAU;AACvC,WAAO,cAAc,MAAM,cAAc,MAAM,SAAS;AAAA,EAC1D;AAEA,MAAI,CAAC,cAAc,MAAM,eAAe,MAAM,SAAS,GAAG;AACxD,UAAM,OAAO;AAAA,MACX,SAAS,MAAM,KAAK,MAAM,MAAM,SAAS,cAAc,MAAM,SAAS;AAAA,IACxE;AAEA,WAAO;AAAA,EACT;AAEA,SAAO;AACT;","names":[]}