UNPKG

langextract

Version:

A TypeScript library for extracting structured and grounded information from text using LLMs

249 lines 10.3 kB
"use strict"; /** * Copyright 2025 kmbro. * * This is a TypeScript translation of the original Python LangExtract library * by Google LLC (https://github.com/google/langextract). * * Licensed under the Apache License, Version 2.0 (the "License"); * you may not use this file except in compliance with the License. * You may obtain a copy of the License at * * http://www.apache.org/licenses/LICENSE-2.0 * * Unless required by applicable law or agreed to in writing, software * distributed under the License is distributed on an "AS IS" BASIS, * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. * See the License for the specific language governing permissions and * limitations under the License. */ var __createBinding = (this && this.__createBinding) || (Object.create ? (function(o, m, k, k2) { if (k2 === undefined) k2 = k; var desc = Object.getOwnPropertyDescriptor(m, k); if (!desc || ("get" in desc ? !m.__esModule : desc.writable || desc.configurable)) { desc = { enumerable: true, get: function() { return m[k]; } }; } Object.defineProperty(o, k2, desc); }) : (function(o, m, k, k2) { if (k2 === undefined) k2 = k; o[k2] = m[k]; })); var __setModuleDefault = (this && this.__setModuleDefault) || (Object.create ? (function(o, v) { Object.defineProperty(o, "default", { enumerable: true, value: v }); }) : function(o, v) { o["default"] = v; }); var __importStar = (this && this.__importStar) || (function () { var ownKeys = function(o) { ownKeys = Object.getOwnPropertyNames || function (o) { var ar = []; for (var k in o) if (Object.prototype.hasOwnProperty.call(o, k)) ar[ar.length] = k; return ar; }; return ownKeys(o); }; return function (mod) { if (mod && mod.__esModule) return mod; var result = {}; if (mod != null) for (var k = ownKeys(mod), i = 0; i < k.length; i++) if (k[i] !== "default") __createBinding(result, mod, k[i]); __setModuleDefault(result, mod); return result; }; })(); Object.defineProperty(exports, "__esModule", { value: true }); exports.Resolver = exports.ResolverParsingError = void 0; /** * Library for resolving LLM output. */ const yaml = __importStar(require("js-yaml")); const types_1 = require("./types"); const tokenizer_1 = require("./tokenizer"); const FUZZY_ALIGNMENT_MIN_THRESHOLD = 0.75; class ResolverParsingError extends Error { constructor(message) { super(message); this.name = "ResolverParsingError"; } } exports.ResolverParsingError = ResolverParsingError; class Resolver { constructor(options = {}) { this._fenceOutput = options.fenceOutput ?? true; this._constraint = options.constraint ?? { constraintType: "none" }; this._formatType = options.formatType ?? types_1.FormatType.JSON; this.extractionIndexSuffix = options.extractionIndexSuffix; this.extractionAttributesSuffix = options.extractionAttributesSuffix ?? "_attributes"; } get fenceOutput() { return this._fenceOutput; } set fenceOutput(value) { this._fenceOutput = value; } get formatType() { return this._formatType; } set formatType(value) { this._formatType = value; } resolve(inputText, options = {}) { try { const extractionData = this.stringToExtractionData(inputText); return this.extractOrderedExtractions(extractionData); } catch (error) { if (options.suppressParseErrors) { console.warn("Parse error suppressed:", error); return []; } throw new ResolverParsingError(`Failed to resolve input text: ${error}`); } } align(extractions, sourceText, tokenOffset, charOffset = 0, enableFuzzyAlignment = true, fuzzyAlignmentThreshold = FUZZY_ALIGNMENT_MIN_THRESHOLD) { const alignedExtractions = []; const sourceTokens = (0, tokenizer_1.tokenize)(sourceText).tokens; for (const extraction of extractions) { const alignedExtraction = this.alignSingleExtraction(extraction, sourceTokens, sourceText, tokenOffset, charOffset, enableFuzzyAlignment, fuzzyAlignmentThreshold); if (alignedExtraction) { alignedExtractions.push(alignedExtraction); } } return alignedExtractions; } extractAndParseContent(inputString) { let content = inputString.trim(); // Remove fence markers if present if (this._fenceOutput) { const fenceRegex = /```(?:json|yaml|yml)\n?([\s\S]*?)\n?```/; const match = content.match(fenceRegex); if (match) { content = match[1].trim(); } } try { if (this._formatType === types_1.FormatType.JSON) { return JSON.parse(content); } else if (this._formatType === types_1.FormatType.YAML) { return yaml.load(content); } else { throw new Error(`Unsupported format type: ${this._formatType}`); } } catch (error) { throw new Error(`Failed to parse content as ${this._formatType}: ${error}`); } } stringToExtractionData(inputString) { const parsed = this.extractAndParseContent(inputString); if (Array.isArray(parsed)) { return parsed; } else if (parsed && typeof parsed === "object" && "extractions" in parsed) { return parsed.extractions; } else { throw new Error("Invalid extraction data format"); } } extractOrderedExtractions(extractionData) { const extractions = []; for (let i = 0; i < extractionData.length; i++) { const data = extractionData[i]; for (const [key, value] of Object.entries(data)) { if (key === this.extractionAttributesSuffix || key.endsWith(this.extractionAttributesSuffix)) { continue; // Skip attribute fields } if (typeof value === "string") { const extraction = { extractionClass: key, extractionText: value, extractionIndex: i, }; // Add attributes if available const attributesKey = `${key}${this.extractionAttributesSuffix}`; if (data[attributesKey] && typeof data[attributesKey] === "object") { extraction.attributes = data[attributesKey]; } extractions.push(extraction); } } } return extractions; } alignSingleExtraction(extraction, sourceTokens, sourceText, tokenOffset, charOffset, enableFuzzyAlignment, fuzzyAlignmentThreshold) { const extractionTokens = (0, tokenizer_1.tokenize)(extraction.extractionText).tokens; // Try exact match first const exactMatch = this.findExactMatch(extractionTokens, sourceTokens, tokenOffset); if (exactMatch) { return { ...extraction, charInterval: this.tokenToCharInterval(exactMatch.start, exactMatch.end, sourceText, charOffset), alignmentStatus: types_1.AlignmentStatus.MATCH_EXACT, }; } // Try fuzzy alignment if enabled if (enableFuzzyAlignment) { const fuzzyMatch = this.findFuzzyMatch(extractionTokens, sourceTokens, tokenOffset, fuzzyAlignmentThreshold); if (fuzzyMatch) { return { ...extraction, charInterval: this.tokenToCharInterval(fuzzyMatch.start, fuzzyMatch.end, sourceText, charOffset), alignmentStatus: types_1.AlignmentStatus.MATCH_FUZZY, }; } } return null; } findExactMatch(extractionTokens, sourceTokens, tokenOffset) { const normalizedExtraction = extractionTokens.map(tokenizer_1.normalizeToken); for (let i = tokenOffset; i <= sourceTokens.length - normalizedExtraction.length; i++) { let match = true; for (let j = 0; j < normalizedExtraction.length; j++) { if ((0, tokenizer_1.normalizeToken)(sourceTokens[i + j]) !== normalizedExtraction[j]) { match = false; break; } } if (match) { return { start: i, end: i + normalizedExtraction.length }; } } return null; } findFuzzyMatch(extractionTokens, sourceTokens, tokenOffset, threshold) { // Simplified fuzzy matching - in a real implementation, you'd use more sophisticated algorithms const normalizedExtraction = extractionTokens.map(tokenizer_1.normalizeToken); for (let i = tokenOffset; i <= sourceTokens.length - normalizedExtraction.length; i++) { let matches = 0; for (let j = 0; j < normalizedExtraction.length; j++) { if ((0, tokenizer_1.normalizeToken)(sourceTokens[i + j]) === normalizedExtraction[j]) { matches++; } } const similarity = matches / normalizedExtraction.length; if (similarity >= threshold) { return { start: i, end: i + normalizedExtraction.length }; } } return null; } tokenToCharInterval(startToken, endToken, sourceText, charOffset) { const tokens = (0, tokenizer_1.tokenize)(sourceText); let startPos = charOffset; let endPos = charOffset; if (startToken < tokens.charIntervals.length) { startPos += tokens.charIntervals[startToken].startPos ?? 0; } if (endToken <= tokens.charIntervals.length) { endPos += tokens.charIntervals[endToken - 1].endPos ?? sourceText.length; } else { endPos += sourceText.length; } return { startPos, endPos }; } } exports.Resolver = Resolver; //# sourceMappingURL=resolver.js.map