@promptbook/website-crawler 0.71.0-17 → 0.71.0-19

package/README.md

@@ -2,7 +2,7 @@
 
 # ![Promptbook logo - cube with letters P and B](./other/design/logo-h1.png) Promptbook
 
-Supercharge your use of large language models
+Build responsible, controlled and transparent applications on top of LLM models!
 
 
 
@@ -21,11 +21,9 @@ Supercharge your use of large language models
 -   ✨ **Support of [OpenAI o1 model](https://openai.com/o1/)**
 
 <blockquote style="color: #ff8811">
-<b>⚠ Warning:</b> This is a pre-release version of the library. It is not yet ready for production use. Please look at <a href="https://www.npmjs.com/package/@promptbook/core?activeTab=versions">latest stable release</a>.
+    <b>⚠ Warning:</b> This is a pre-release version of the library. It is not yet ready for production use. Please look at <a href="https://www.npmjs.com/package/@promptbook/core?activeTab=versions">latest stable release</a>.
 </blockquote>
 
-
-
 ## 📦 Package `@promptbook/website-crawler`
 
 - Promptbooks are [divided into several](#-packages) packages, all are published from [single monorepo](https://github.com/webgptorg/promptbook).

package/esm/index.es.js

@@ -2,12 +2,12 @@ import spaceTrim$1, { spaceTrim } from 'spacetrim';
 import { Readability } from '@mozilla/readability';
 import { mkdir, rm, writeFile } from 'fs/promises';
 import { JSDOM } from 'jsdom';
-import { forTime } from 'waitasecond';
 import { SHA256 } from 'crypto-js';
 import hexEncoder from 'crypto-js/enc-hex';
 import { basename, join, dirname } from 'path';
 import { format } from 'prettier';
 import parserHtml from 'prettier/parser-html';
+import { forTime } from 'waitasecond';
 import { lookup } from 'mime-types';
 import { unparse, parse } from 'papaparse';
 import { Converter } from 'showdown';
@@ -16,7 +16,7 @@ import { Converter } from 'showdown';
 /**
  * The version of the Promptbook library
  */
-var PROMPTBOOK_VERSION = '0.71.0-16';
+var PROMPTBOOK_VERSION = '0.71.0-18';
 // TODO: [main] !!!! List here all the versions and annotate + put into script
 
 /*! *****************************************************************************
@@ -364,18 +364,25 @@ var LOOP_LIMIT = 1000;
  * @private within the repository - too low-level in comparison with other `MAX_...`
  */
 var IMMEDIATE_TIME = 10;
+/**
+ * Strategy for caching the intermediate results for knowledge sources
+ *
+ * @public exported from `@promptbook/core`
+ */
+var DEFAULT_INTERMEDIATE_FILES_STRATEGY = 'HIDE_AND_KEEP';
+//                                                     <- TODO: [😡] Change to 'VISIBLE'
 /**
  * The maximum number of (LLM) tasks running in parallel
  *
  * @public exported from `@promptbook/core`
  */
-var MAX_PARALLEL_COUNT = 5; // <- TODO: [🤹‍♂️]
+var DEFAULT_MAX_PARALLEL_COUNT = 5; // <- TODO: [🤹‍♂️]
 /**
  * The maximum number of attempts to execute LLM task before giving up
  *
  * @public exported from `@promptbook/core`
  */
-var MAX_EXECUTION_ATTEMPTS = 3; // <- TODO: [🤹‍♂️]
+var DEFAULT_MAX_EXECUTION_ATTEMPTS = 3; // <- TODO: [🤹‍♂️]
 /**
  * Where to store the scrape cache
  *
@@ -383,7 +390,7 @@ var MAX_EXECUTION_ATTEMPTS = 3; // <- TODO: [🤹‍♂️]
  *
  * @public exported from `@promptbook/core`
  */
-var SCRAPE_CACHE_DIRNAME = '/.promptbook/scrape-cache';
+var DEFAULT_SCRAPE_CACHE_DIRNAME = '/.promptbook/scrape-cache';
 /**
  * Nonce which is used for replacing things in strings
  *
@@ -435,7 +442,7 @@ var DEFAULT_CSV_SETTINGS = Object.freeze({
  *
  * @public exported from `@promptbook/core`
  */
-var IS_VERBOSE = false;
+var DEFAULT_IS_VERBOSE = false;
 /**
  * @@@
  *
@@ -576,7 +583,7 @@ var $Register = /** @class */ (function () {
         this.storage = globalScope[storageName];
     }
     $Register.prototype.list = function () {
-        // <- TODO: ReadonlyDeep<Array<TRegistered>>
+        // <- TODO: ReadonlyDeep<ReadonlyArray<TRegistered>>
         return this.storage;
     };
     $Register.prototype.register = function (registered) {
@@ -628,7 +635,7 @@ var websiteScraperMetadata = $deepFreeze({
     mimeTypes: ['text/html'],
     documentationUrl: 'https://github.com/webgptorg/promptbook/discussions/@@',
     isAvilableInBrowser: false,
-    requiredExecutables: ['!!!!!!'],
+    requiredExecutables: [],
 }); /* <- TODO: [🤛] */
 /**
  * Registration of known scraper metadata
@@ -1106,12 +1113,12 @@ function TODO_USE() {
  */
 function getScraperIntermediateSource(source, options) {
     return __awaiter(this, void 0, void 0, function () {
-        var sourceFilename, url, rootDirname, cacheDirname, isCacheCleaned, extension, isVerbose, hash, semanticName, pieces, name, cacheFilename, isDestroyed, fileHandler;
+        var sourceFilename, url, rootDirname, cacheDirname, intermediateFilesStrategy, extension, isVerbose, hash, semanticName, pieces, name, cacheFilename, isDestroyed, fileHandler;
         return __generator(this, function (_a) {
             switch (_a.label) {
                 case 0:
                     sourceFilename = source.filename, url = source.url;
-                    rootDirname = options.rootDirname, cacheDirname = options.cacheDirname, isCacheCleaned = options.isCacheCleaned, extension = options.extension, isVerbose = options.isVerbose;
+                    rootDirname = options.rootDirname, cacheDirname = options.cacheDirname, intermediateFilesStrategy = options.intermediateFilesStrategy, extension = options.extension, isVerbose = options.isVerbose;
                     hash = SHA256(
                     //    <- TODO: [🥬] Encapsulate sha256 to some private utility function
                     hexEncoder.parse(sourceFilename || url || 'untitled'))
@@ -1121,7 +1128,7 @@ function getScraperIntermediateSource(source, options) {
                     pieces = ['intermediate', semanticName, hash].filter(function (piece) { return piece !== ''; });
                     name = pieces.join('-').split('--').join('-');
                     // <- TODO: Use MAX_FILENAME_LENGTH
-                    TODO_USE(rootDirname); // <- TODO: !!!!!!
+                    TODO_USE(rootDirname); // <- TODO: [😡]
                     cacheFilename = join.apply(void 0, __spreadArray(__spreadArray([process.cwd(),
                         cacheDirname], __read(nameToSubfolderPath(hash /* <- TODO: [🎎] Maybe add some SHA256 prefix */)), false), [name], false)).split('\\')
                         .join('/') +
@@ -1141,7 +1148,7 @@ function getScraperIntermediateSource(source, options) {
                                 return __generator(this, function (_a) {
                                     switch (_a.label) {
                                         case 0:
-                                            if (!isCacheCleaned) return [3 /*break*/, 2];
+                                            if (!(intermediateFilesStrategy === 'HIDE_AND_CLEAN')) return [3 /*break*/, 2];
                                             if (isVerbose) {
                                                 console.info('legacyDocumentScraper: Clening cache');
                                             }
@@ -1165,7 +1172,7 @@ function getScraperIntermediateSource(source, options) {
 /**
  * Note: Not using `FileCacheStorage` for two reasons:
  * 1) Need to store more than serialized JSONs
- * 2) Need to switch between a `rootDirname` and `cacheDirname` <- TODO: !!!!
+ * 2) Need to switch between a `rootDirname` and `cacheDirname` <- TODO: [😡]
  * TODO: [🐱‍🐉][🧠] Make some smart crop
  * Note: [🟢] Code in this file should never be never released in packages that could be imported into browser environment
  */
@@ -1230,7 +1237,6 @@ function pipelineJsonToString(pipelineJson) {
         pipelineString += '\n\n';
         pipelineString += description;
     }
-    // TODO:> const commands: Array<Command>
     var commands = [];
     if (pipelineUrl) {
         commands.push("PIPELINE URL ".concat(pipelineUrl));
@@ -1286,7 +1292,6 @@ function pipelineJsonToString(pipelineJson) {
                 pipelineString += '\n\n';
                 pipelineString += description_1;
             }
-            // TODO:> const commands: Array<Command>
             var commands_1 = [];
             var contentLanguage = 'text';
             if (templateType === 'PROMPT_TEMPLATE') {
@@ -2187,6 +2192,7 @@ function assertsExecutionSuccessful(executionResult) {
     }
 }
 /**
+ * TODO: [🐚] This function should be removed OR changed OR be completely rewritten
  * TODO: [🧠] Can this return type be better typed than void
  */
 
@@ -2527,8 +2533,7 @@ function countTotalUsage(llmTools) {
  * Multiple LLM Execution Tools is a proxy server that uses multiple execution tools internally and exposes the executor interface externally.
  *
  * Note: Internal utility of `joinLlmExecutionTools` but exposed type
- * @public exported from `@promptbook/types`
- * TODO: !!!!!! Export as runtime class not just type
+ * @public exported from `@promptbook/core`
  */
 var MultipleLlmExecutionTools = /** @class */ (function () {
     /**
@@ -2848,7 +2853,7 @@ function preparePersona(personaDescription, tools, options) {
         return __generator(this, function (_d) {
             switch (_d.label) {
                 case 0:
-                    _a = options.isVerbose, isVerbose = _a === void 0 ? IS_VERBOSE : _a;
+                    _a = options.isVerbose, isVerbose = _a === void 0 ? DEFAULT_IS_VERBOSE : _a;
                     if (tools === undefined || tools.llm === undefined) {
                         throw new MissingToolsError('LLM tools are required for preparing persona');
                     }
@@ -2915,7 +2920,7 @@ var $scrapersRegister = new $Register('scraper_constructors');
  * TODO: [®] DRY Register logic
  */
 
-// TODO: !!!!!! Maybe delete this function
+// TODO: !!!!!!last - Maybe delete this function
 /**
  * Creates a message with all registered scrapers
  *
@@ -3023,7 +3028,6 @@ function $registeredScrapersMessage() {
  * @private within the repository
  */
 function sourceContentToName(sourceContent) {
-    // TODO: !!!!!! Better name for source than gibberish hash
     var hash = SHA256(hexEncoder.parse(JSON.stringify(sourceContent)))
         //    <- TODO: [🥬] Encapsulate sha256 to some private utility function
         .toString( /* hex */)
@@ -3097,7 +3101,7 @@ function isFileExisting(filename, fs) {
 /**
  * @@@
  *
- * @private for scraper utilities
+ * @public exported from `@promptbook/core`
  */
 function makeKnowledgeSourceHandler(knowledgeSource, tools, options) {
     var _a;
@@ -3110,7 +3114,7 @@ function makeKnowledgeSourceHandler(knowledgeSource, tools, options) {
                 case 0:
                     sourceContent = knowledgeSource.sourceContent;
                     name = knowledgeSource.name;
-                    _b = options || {}, _c = _b.rootDirname, rootDirname = _c === void 0 ? null : _c, _d = _b.isVerbose, isVerbose = _d === void 0 ? IS_VERBOSE : _d;
+                    _b = options || {}, _c = _b.rootDirname, rootDirname = _c === void 0 ? null : _c, _d = _b.isVerbose, isVerbose = _d === void 0 ? DEFAULT_IS_VERBOSE : _d;
                     TODO_USE(isVerbose);
                     if (!name) {
                         name = sourceContentToName(sourceContent);
@@ -3200,7 +3204,7 @@ function makeKnowledgeSourceHandler(knowledgeSource, tools, options) {
                                                 content = _a.sent();
                                                 return [2 /*return*/, new Blob([
                                                         content,
-                                                        // <- TODO: !!!!!! Maybe not working
+                                                        // <- TODO: !!!!!! Test that this is working
                                                     ], { type: mimeType_1 })];
                                         }
                                     });
@@ -3263,7 +3267,7 @@ function prepareKnowledgePieces(knowledgeSources, tools, options) {
         return __generator(this, function (_c) {
             switch (_c.label) {
                 case 0:
-                    _a = options.maxParallelCount, maxParallelCount = _a === void 0 ? MAX_PARALLEL_COUNT : _a, rootDirname = options.rootDirname, _b = options.isVerbose, isVerbose = _b === void 0 ? IS_VERBOSE : _b;
+                    _a = options.maxParallelCount, maxParallelCount = _a === void 0 ? DEFAULT_MAX_PARALLEL_COUNT : _a, rootDirname = options.rootDirname, _b = options.isVerbose, isVerbose = _b === void 0 ? DEFAULT_IS_VERBOSE : _b;
                     knowledgePreparedUnflatten = new Array(knowledgeSources.length);
                     return [4 /*yield*/, forEachAsync(knowledgeSources, { maxParallelCount: maxParallelCount }, function (knowledgeSource, index) { return __awaiter(_this, void 0, void 0, function () {
                             var partialPieces, sourceHandler, _a, _b, scraper, partialPiecesUnchecked, e_1_1, pieces;
@@ -3292,7 +3296,8 @@ function prepareKnowledgePieces(knowledgeSources, tools, options) {
                                     case 4:
                                         partialPiecesUnchecked = _d.sent();
                                         if (partialPiecesUnchecked !== null) {
-                                            partialPieces = partialPiecesUnchecked;
+                                            partialPieces = __spreadArray([], __read(partialPiecesUnchecked), false);
+                                            // <- TODO: [🪓] Here should be no need for spreading new array, just `partialPieces = partialPiecesUnchecked`
                                             return [3 /*break*/, 6];
                                         }
                                         _d.label = 5;
@@ -3340,7 +3345,7 @@ TODO: [🧊] This is how it can look in future
 >   /**
 >    * Unprepared knowledge
 >    * /
->   readonly knowledgeSources: Array<KnowledgeSourceJson>;
+>   readonly knowledgeSources: ReadonlyArray<KnowledgeSourceJson>;
 > };
 >
 > export async function prepareKnowledgePieces(
@@ -3398,7 +3403,7 @@ function prepareTemplates(pipeline, tools, options) {
         return __generator(this, function (_b) {
             switch (_b.label) {
                 case 0:
-                    _a = options.maxParallelCount, maxParallelCount = _a === void 0 ? MAX_PARALLEL_COUNT : _a;
+                    _a = options.maxParallelCount, maxParallelCount = _a === void 0 ? DEFAULT_MAX_PARALLEL_COUNT : _a;
                     templates = pipeline.templates, parameters = pipeline.parameters, knowledgePiecesCount = pipeline.knowledgePiecesCount;
                     // TODO: [main] !!!!! Apply samples to each template (if missing and is for the template defined)
                     TODO_USE(parameters);
@@ -3460,7 +3465,7 @@ function preparePipeline(pipeline, tools, options) {
                     if (isPipelinePrepared(pipeline)) {
                         return [2 /*return*/, pipeline];
                     }
-                    rootDirname = options.rootDirname, _a = options.maxParallelCount, maxParallelCount = _a === void 0 ? MAX_PARALLEL_COUNT : _a, _b = options.isVerbose, isVerbose = _b === void 0 ? IS_VERBOSE : _b;
+                    rootDirname = options.rootDirname, _a = options.maxParallelCount, maxParallelCount = _a === void 0 ? DEFAULT_MAX_PARALLEL_COUNT : _a, _b = options.isVerbose, isVerbose = _b === void 0 ? DEFAULT_IS_VERBOSE : _b;
                     parameters = pipeline.parameters, templates = pipeline.templates, knowledgeSources = pipeline.knowledgeSources, personas = pipeline.personas;
                     if (tools === undefined || tools.llm === undefined) {
                         throw new MissingToolsError('LLM tools are required for preparing the pipeline');
@@ -3518,7 +3523,9 @@ function preparePipeline(pipeline, tools, options) {
                     // ----- /Templates preparation -----
                     // Note: Count total usage
                     currentPreparation.usage = llmToolsWithUsage.getTotalUsage();
-                    return [2 /*return*/, $asDeeplyFrozenSerializableJson('Prepared PipelineJson', __assign(__assign({}, clonePipeline(pipeline)), { templates: templatesPrepared, knowledgeSources: knowledgeSourcesPrepared, knowledgePieces: knowledgePiecesPrepared, personas: preparedPersonas, preparations: preparations }))];
+                    return [2 /*return*/, $asDeeplyFrozenSerializableJson('Prepared PipelineJson', __assign(__assign({}, clonePipeline(pipeline)), { templates: __spreadArray([], __read(templatesPrepared), false), 
+                            // <- TODO: [🪓] Here should be no need for spreading new array, just ` templates: templatesPrepared`
+                            knowledgeSources: knowledgeSourcesPrepared, knowledgePieces: knowledgePiecesPrepared, personas: preparedPersonas, preparations: __spreadArray([], __read(preparations), false) }))];
             }
         });
     });
@@ -4489,12 +4496,11 @@ function checkExpectations(expectations, value) {
  */
 function executeAttempts(options) {
     return __awaiter(this, void 0, void 0, function () {
-        var jokerParameterNames, priority, maxAttempts, preparedContent, parameters, template, preparedPipeline, tools, settings, $executionReport, pipelineIdentification, maxExecutionAttempts, $ongoingTemplateResult, _llms, llmTools, _loop_1, attempt, state_1;
+        var jokerParameterNames, priority, maxAttempts, preparedContent, parameters, template, preparedPipeline, tools, $executionReport, pipelineIdentification, maxExecutionAttempts, $ongoingTemplateResult, _llms, llmTools, _loop_1, attempt, state_1;
         return __generator(this, function (_a) {
             switch (_a.label) {
                 case 0:
-                    jokerParameterNames = options.jokerParameterNames, priority = options.priority, maxAttempts = options.maxAttempts, preparedContent = options.preparedContent, parameters = options.parameters, template = options.template, preparedPipeline = options.preparedPipeline, tools = options.tools, settings = options.settings, $executionReport = options.$executionReport, pipelineIdentification = options.pipelineIdentification;
-                    maxExecutionAttempts = settings.maxExecutionAttempts;
+                    jokerParameterNames = options.jokerParameterNames, priority = options.priority, maxAttempts = options.maxAttempts, preparedContent = options.preparedContent, parameters = options.parameters, template = options.template, preparedPipeline = options.preparedPipeline, tools = options.tools, $executionReport = options.$executionReport, pipelineIdentification = options.pipelineIdentification, maxExecutionAttempts = options.maxExecutionAttempts;
                     $ongoingTemplateResult = {
                         $result: null,
                         $resultString: null,
@@ -4860,12 +4866,12 @@ function executeAttempts(options) {
  */
 function executeFormatSubvalues(options) {
     return __awaiter(this, void 0, void 0, function () {
-        var template, jokerParameterNames, parameters, priority, pipelineIdentification, settings, parameterValue, formatDefinition, subvalueDefinition, formatSettings, resultString;
+        var template, jokerParameterNames, parameters, priority, csvSettings, pipelineIdentification, parameterValue, formatDefinition, subvalueDefinition, formatSettings, resultString;
         var _this = this;
         return __generator(this, function (_a) {
             switch (_a.label) {
                 case 0:
-                    template = options.template, jokerParameterNames = options.jokerParameterNames, parameters = options.parameters, priority = options.priority, pipelineIdentification = options.pipelineIdentification, settings = options.settings;
+                    template = options.template, jokerParameterNames = options.jokerParameterNames, parameters = options.parameters, priority = options.priority, csvSettings = options.csvSettings, pipelineIdentification = options.pipelineIdentification;
                     if (template.foreach === undefined) {
                         return [2 /*return*/, /* not await */ executeAttempts(options)];
                     }
@@ -4895,7 +4901,7 @@ function executeFormatSubvalues(options) {
                             .join('\n')), "\n\n                    [\u26F7] This should never happen because subformat name should be validated during parsing\n\n                    ").concat(block(pipelineIdentification), "\n                "); }));
                     }
                     if (formatDefinition.formatName === 'CSV') {
-                        formatSettings = settings.csvSettings;
+                        formatSettings = csvSettings;
                         // <- TODO: [🤹‍♂️] More universal, make simmilar pattern for other formats for example \n vs \r\n in text
                     }
                     return [4 /*yield*/, subvalueDefinition.mapValues(parameterValue, template.foreach.outputSubparameterName, formatSettings, function (subparameters, index) { return __awaiter(_this, void 0, void 0, function () {
@@ -5048,13 +5054,12 @@ function getReservedParametersForTemplate(options) {
  */
 function executeTemplate(options) {
     return __awaiter(this, void 0, void 0, function () {
-        var currentTemplate, preparedPipeline, parametersToPass, tools, onProgress, settings, $executionReport, pipelineIdentification, maxExecutionAttempts, name, title, priority, usedParameterNames, dependentParameterNames, definedParameters, _a, _b, _c, definedParameterNames, parameters, _loop_1, _d, _e, parameterName, maxAttempts, jokerParameterNames, preparedContent, resultString;
-        var e_1, _f, _g;
-        return __generator(this, function (_h) {
-            switch (_h.label) {
+        var currentTemplate, preparedPipeline, parametersToPass, tools, onProgress, $executionReport, pipelineIdentification, _a, maxExecutionAttempts, name, title, priority, usedParameterNames, dependentParameterNames, definedParameters, _b, _c, _d, definedParameterNames, parameters, _loop_1, _e, _f, parameterName, maxAttempts, jokerParameterNames, preparedContent, resultString;
+        var e_1, _g, _h;
+        return __generator(this, function (_j) {
+            switch (_j.label) {
                 case 0:
-                    currentTemplate = options.currentTemplate, preparedPipeline = options.preparedPipeline, parametersToPass = options.parametersToPass, tools = options.tools, onProgress = options.onProgress, settings = options.settings, $executionReport = options.$executionReport, pipelineIdentification = options.pipelineIdentification;
-                    maxExecutionAttempts = settings.maxExecutionAttempts;
+                    currentTemplate = options.currentTemplate, preparedPipeline = options.preparedPipeline, parametersToPass = options.parametersToPass, tools = options.tools, onProgress = options.onProgress, $executionReport = options.$executionReport, pipelineIdentification = options.pipelineIdentification, _a = options.maxExecutionAttempts, maxExecutionAttempts = _a === void 0 ? DEFAULT_MAX_EXECUTION_ATTEMPTS : _a;
                     name = "pipeline-executor-frame-".concat(currentTemplate.name);
                     title = currentTemplate.title;
                     priority = preparedPipeline.templates.length - preparedPipeline.templates.indexOf(currentTemplate);
@@ -5069,7 +5074,7 @@ function executeTemplate(options) {
                             // <- [🍸]
                         })];
                 case 1:
-                    _h.sent();
+                    _j.sent();
                     usedParameterNames = extractParameterNamesFromTemplate(currentTemplate);
                     dependentParameterNames = new Set(currentTemplate.dependentParameterNames);
                     // TODO: [👩🏾‍🤝‍👩🏻] Use here `mapAvailableToExpectedParameters`
@@ -5080,15 +5085,15 @@ function executeTemplate(options) {
                             .map(function (name) { return "{".concat(name, "}"); })
                             .join(', '), "\n\n                    ").concat(block(pipelineIdentification), "\n\n                "); }));
                     }
-                    _b = (_a = Object).freeze;
-                    _c = [{}];
+                    _c = (_b = Object).freeze;
+                    _d = [{}];
                     return [4 /*yield*/, getReservedParametersForTemplate({
                             preparedPipeline: preparedPipeline,
                             template: currentTemplate,
                             pipelineIdentification: pipelineIdentification,
                         })];
                 case 2:
-                    definedParameters = _b.apply(_a, [__assign.apply(void 0, [__assign.apply(void 0, _c.concat([(_h.sent())])), parametersToPass])]);
+                    definedParameters = _c.apply(_b, [__assign.apply(void 0, [__assign.apply(void 0, _d.concat([(_j.sent())])), parametersToPass])]);
                     definedParameterNames = new Set(Object.keys(definedParameters));
                     parameters = {};
                     _loop_1 = function (parameterName) {
@@ -5108,15 +5113,15 @@ function executeTemplate(options) {
                     try {
                         // Note: [2] Check that all used parameters are defined and removing unused parameters for this template
                         // TODO: [👩🏾‍🤝‍👩🏻] Use here `mapAvailableToExpectedParameters`
-                        for (_d = __values(Array.from(union(definedParameterNames, usedParameterNames, dependentParameterNames))), _e = _d.next(); !_e.done; _e = _d.next()) {
-                            parameterName = _e.value;
+                        for (_e = __values(Array.from(union(definedParameterNames, usedParameterNames, dependentParameterNames))), _f = _e.next(); !_f.done; _f = _e.next()) {
+                            parameterName = _f.value;
                             _loop_1(parameterName);
                         }
                     }
                     catch (e_1_1) { e_1 = { error: e_1_1 }; }
                     finally {
                         try {
-                            if (_e && !_e.done && (_f = _d.return)) _f.call(_d);
+                            if (_f && !_f.done && (_g = _e.return)) _g.call(_e);
                         }
                         finally { if (e_1) throw e_1.error; }
                     }
@@ -5136,12 +5141,11 @@ function executeTemplate(options) {
                             template: currentTemplate,
                             preparedPipeline: preparedPipeline,
                             tools: tools,
-                            settings: settings,
                             $executionReport: $executionReport,
                             pipelineIdentification: pipelineIdentification,
                         })];
                 case 3:
-                    resultString = _h.sent();
+                    resultString = _j.sent();
                     return [4 /*yield*/, onProgress({
                             name: name,
                             title: title,
@@ -5153,12 +5157,12 @@ function executeTemplate(options) {
                             // <- [🍸]
                         })];
                 case 4:
-                    _h.sent();
-                    return [2 /*return*/, Object.freeze((_g = {},
-                            _g[currentTemplate.resultingParameterName] = 
+                    _j.sent();
+                    return [2 /*return*/, Object.freeze((_h = {},
+                            _h[currentTemplate.resultingParameterName] = 
                             // <- Note: [👩‍👩‍👧] No need to detect parameter collision here because pipeline checks logic consistency during construction
                             resultString,
-                            _g))];
+                            _h))];
             }
         });
     });
@@ -5217,13 +5221,12 @@ function filterJustOutputParameters(options) {
  */
 function executePipeline(options) {
     return __awaiter(this, void 0, void 0, function () {
-        var inputParameters, tools, onProgress, pipeline, setPreparedPipeline, pipelineIdentification, settings, maxParallelCount, rootDirname, _a, isVerbose, preparedPipeline, errors, warnings, executionReport, isReturned, _b, _c, parameter, e_1_1, _loop_1, _d, _e, parameterName, state_1, e_2_1, parametersToPass, resovedParameterNames_1, unresovedTemplates_1, resolving_1, loopLimit, _loop_2, error_1, usage_1, outputParameters_1, usage, outputParameters;
+        var inputParameters, tools, onProgress, pipeline, setPreparedPipeline, pipelineIdentification, maxParallelCount, rootDirname, _a, isVerbose, preparedPipeline, errors, warnings, executionReport, isReturned, _b, _c, parameter, e_1_1, _loop_1, _d, _e, parameterName, state_1, e_2_1, parametersToPass, resovedParameterNames_1, unresovedTemplates_1, resolving_1, loopLimit, _loop_2, error_1, usage_1, outputParameters_1, usage, outputParameters;
         var e_1, _f, e_2, _g;
         return __generator(this, function (_h) {
             switch (_h.label) {
                 case 0:
-                    inputParameters = options.inputParameters, tools = options.tools, onProgress = options.onProgress, pipeline = options.pipeline, setPreparedPipeline = options.setPreparedPipeline, pipelineIdentification = options.pipelineIdentification, settings = options.settings;
-                    maxParallelCount = settings.maxParallelCount, rootDirname = settings.rootDirname, _a = settings.isVerbose, isVerbose = _a === void 0 ? IS_VERBOSE : _a;
+                    inputParameters = options.inputParameters, tools = options.tools, onProgress = options.onProgress, pipeline = options.pipeline, setPreparedPipeline = options.setPreparedPipeline, pipelineIdentification = options.pipelineIdentification, maxParallelCount = options.maxParallelCount, rootDirname = options.rootDirname, _a = options.isVerbose, isVerbose = _a === void 0 ? DEFAULT_IS_VERBOSE : _a;
                     preparedPipeline = options.preparedPipeline;
                     if (!(preparedPipeline === undefined)) return [3 /*break*/, 2];
                     return [4 /*yield*/, preparePipeline(pipeline, tools, {
@@ -5408,12 +5411,7 @@ function executePipeline(options) {
                                     return [3 /*break*/, 4];
                                 case 3:
                                     unresovedTemplates_1 = unresovedTemplates_1.filter(function (template) { return template !== currentTemplate; });
-                                    work_1 = executeTemplate({
-                                        currentTemplate: currentTemplate,
-                                        preparedPipeline: preparedPipeline,
-                                        parametersToPass: parametersToPass,
-                                        tools: tools,
-                                        onProgress: function (progress) {
+                                    work_1 = executeTemplate(__assign(__assign({}, options), { currentTemplate: currentTemplate, preparedPipeline: preparedPipeline, parametersToPass: parametersToPass, tools: tools, onProgress: function (progress) {
                                             if (isReturned) {
                                                 throw new UnexpectedError(spaceTrim(function (block) { return "\n                                        Can not call `onProgress` after pipeline execution is finished\n\n                                        ".concat(block(pipelineIdentification), "\n\n                                        ").concat(block(JSON.stringify(progress, null, 4)
                                                     .split('\n')
@@ -5423,11 +5421,7 @@ function executePipeline(options) {
                                             if (onProgress) {
                                                 onProgress(progress);
                                             }
-                                        },
-                                        settings: settings,
-                                        $executionReport: executionReport,
-                                        pipelineIdentification: spaceTrim(function (block) { return "\n                            ".concat(block(pipelineIdentification), "\n                            Template name: ").concat(currentTemplate.name, "\n                            Template title: ").concat(currentTemplate.title, "\n                        "); }),
-                                    })
+                                        }, $executionReport: executionReport, pipelineIdentification: spaceTrim(function (block) { return "\n                            ".concat(block(pipelineIdentification), "\n                            Template name: ").concat(currentTemplate.name, "\n                            Template title: ").concat(currentTemplate.title, "\n                        "); }) }))
                                         .then(function (newParametersToPass) {
                                         parametersToPass = __assign(__assign({}, newParametersToPass), parametersToPass);
                                         resovedParameterNames_1 = __spreadArray(__spreadArray([], __read(resovedParameterNames_1), false), [currentTemplate.resultingParameterName], false);
@@ -5531,8 +5525,7 @@ function executePipeline(options) {
  */
 function createPipelineExecutor(options) {
     var _this = this;
-    var pipeline = options.pipeline, tools = options.tools, _a = options.settings, settings = _a === void 0 ? {} : _a;
-    var _b = settings.maxExecutionAttempts, maxExecutionAttempts = _b === void 0 ? MAX_EXECUTION_ATTEMPTS : _b, _c = settings.maxParallelCount, maxParallelCount = _c === void 0 ? MAX_PARALLEL_COUNT : _c, _d = settings.csvSettings, csvSettings = _d === void 0 ? DEFAULT_CSV_SETTINGS : _d, _e = settings.isVerbose, isVerbose = _e === void 0 ? IS_VERBOSE : _e, _f = settings.isNotPreparedWarningSupressed, isNotPreparedWarningSupressed = _f === void 0 ? false : _f, _g = settings.rootDirname, rootDirname = _g === void 0 ? null : _g;
+    var pipeline = options.pipeline, tools = options.tools, _a = options.maxExecutionAttempts, maxExecutionAttempts = _a === void 0 ? DEFAULT_MAX_EXECUTION_ATTEMPTS : _a, _b = options.maxParallelCount, maxParallelCount = _b === void 0 ? DEFAULT_MAX_PARALLEL_COUNT : _b, _c = options.csvSettings, csvSettings = _c === void 0 ? DEFAULT_CSV_SETTINGS : _c, _d = options.isVerbose, isVerbose = _d === void 0 ? DEFAULT_IS_VERBOSE : _d, _e = options.isNotPreparedWarningSupressed, isNotPreparedWarningSupressed = _e === void 0 ? false : _e, _f = options.rootDirname, rootDirname = _f === void 0 ? null : _f;
     validatePipeline(pipeline);
     var pipelineIdentification = (function () {
         // Note: This is a 😐 implementation of [🚞]
@@ -5566,14 +5559,12 @@ function createPipelineExecutor(options) {
                     tools: tools,
                     onProgress: onProgress,
                     pipelineIdentification: spaceTrim(function (block) { return "\n                    ".concat(block(pipelineIdentification), "\n                    ").concat(runCount === 1 ? '' : "Run #".concat(runCount), "\n                "); }),
-                    settings: {
-                        maxExecutionAttempts: maxExecutionAttempts,
-                        maxParallelCount: maxParallelCount,
-                        csvSettings: csvSettings,
-                        isVerbose: isVerbose,
-                        isNotPreparedWarningSupressed: isNotPreparedWarningSupressed,
-                        rootDirname: rootDirname,
-                    },
+                    maxExecutionAttempts: maxExecutionAttempts,
+                    maxParallelCount: maxParallelCount,
+                    csvSettings: csvSettings,
+                    isVerbose: isVerbose,
+                    isNotPreparedWarningSupressed: isNotPreparedWarningSupressed,
+                    rootDirname: rootDirname,
                 })];
         });
     }); };
@@ -5595,7 +5586,7 @@ var markdownScraperMetadata = $deepFreeze({
     mimeTypes: ['text/markdown', 'text/plain'],
     documentationUrl: 'https://github.com/webgptorg/promptbook/discussions/@@',
     isAvilableInBrowser: true,
-    requiredExecutables: ['!!!!!!'],
+    requiredExecutables: [],
 }); /* <- TODO: [🤛] */
 /**
  * Registration of known scraper metadata
@@ -5639,7 +5630,7 @@ var MarkdownScraper = /** @class */ (function () {
             return __generator(this, function (_k) {
                 switch (_k.label) {
                     case 0:
-                        _a = this.options, _b = _a.maxParallelCount, maxParallelCount = _b === void 0 ? MAX_PARALLEL_COUNT : _b, _c = _a.isVerbose, isVerbose = _c === void 0 ? IS_VERBOSE : _c;
+                        _a = this.options, _b = _a.maxParallelCount, maxParallelCount = _b === void 0 ? DEFAULT_MAX_PARALLEL_COUNT : _b, _c = _a.isVerbose, isVerbose = _c === void 0 ? DEFAULT_IS_VERBOSE : _c;
                         llm = this.tools.llm;
                         if (llm === undefined) {
                             throw new MissingToolsError('LLM tools are required for scraping external files');
@@ -5738,7 +5729,8 @@ var MarkdownScraper = /** @class */ (function () {
                                             embeddingResult = _c.sent();
                                             index.push({
                                                 modelName: embeddingResult.modelName,
-                                                position: embeddingResult.content,
+                                                position: __spreadArray([], __read(embeddingResult.content), false),
+                                                // <- TODO: [🪓] Here should be no need for spreading new array, just `position: embeddingResult.content`
                                             });
                                             _c.label = 6;
                                         case 6: return [3 /*break*/, 8];
@@ -5778,32 +5770,29 @@ var MarkdownScraper = /** @class */ (function () {
  */
 
 /**
- * A converter instance that uses showdown and highlight extensions
+ * Create a new showdown converter instance
  *
- * @type {Converter}
- * @private for markdown and html knowledge scrapers
- */
-var markdownConverter = new Converter({
-    flavor: 'github', // <- TODO: !!!!!! Explicitly specify the flavor of promptbook markdown
-    /*
-    > import showdownHighlight from 'showdown-highlight';
-    > extensions: [
-    >     showdownHighlight({
-    >         // Whether to add the classes to the <pre> tag, default is false
-    >         pre: true,
-    >         // Whether to use hljs' auto language detection, default is true
-    >         auto_detection: true,
-    >     }),
-    > ],
-    */
-});
-/**
- * TODO: !!!!!! Figure out better name not to confuse with `Converter`
- * TODO: !!!!!! Lazy-make converter
+ * @private utility of `WebsiteScraper`
  */
+function createShowdownConverter() {
+    return new Converter({
+        flavor: 'github',
+        /*
+        > import showdownHighlight from 'showdown-highlight';
+        > extensions: [
+        >     showdownHighlight({
+        >         // Whether to add the classes to the <pre> tag, default is false
+        >         pre: true,
+        >         // Whether to use hljs' auto language detection, default is true
+        >         auto_detection: true,
+        >     }),
+        > ],
+        */
+    });
+}
 
 /**
- * Scraper for .docx files
+ * Scraper for websites
  *
  * @see `documentationUrl` for more details
  * @public exported from `@promptbook/website-crawler`
@@ -5813,6 +5802,7 @@ var WebsiteScraper = /** @class */ (function () {
         this.tools = tools;
         this.options = options;
         this.markdownScraper = new MarkdownScraper(tools, options);
+        this.showdownConverter = createShowdownConverter();
     }
     Object.defineProperty(WebsiteScraper.prototype, "metadata", {
         /**
@@ -5833,12 +5823,11 @@ var WebsiteScraper = /** @class */ (function () {
         return __awaiter(this, void 0, void 0, function () {
             var _a, _b, 
             // TODO: [🧠] Maybe in node use headless browser not just JSDOM
-            // externalProgramsPaths = {},
-            rootDirname, _c, cacheDirname, _d, isCacheCleaned, _e, isVerbose, jsdom, _f, reader, article, html, i, cacheFilehandler, markdown;
+            rootDirname, _c, cacheDirname, _d, intermediateFilesStrategy, _e, isVerbose, jsdom, _f, reader, article, html, i, cacheFilehandler, markdown;
             return __generator(this, function (_g) {
                 switch (_g.label) {
                     case 0:
-                        _a = this.options, _b = _a.rootDirname, rootDirname = _b === void 0 ? process.cwd() : _b, _c = _a.cacheDirname, cacheDirname = _c === void 0 ? SCRAPE_CACHE_DIRNAME : _c, _d = _a.isCacheCleaned, isCacheCleaned = _d === void 0 ? false : _d, _e = _a.isVerbose, isVerbose = _e === void 0 ? IS_VERBOSE : _e;
+                        _a = this.options, _b = _a.rootDirname, rootDirname = _b === void 0 ? process.cwd() : _b, _c = _a.cacheDirname, cacheDirname = _c === void 0 ? DEFAULT_SCRAPE_CACHE_DIRNAME : _c, _d = _a.intermediateFilesStrategy, intermediateFilesStrategy = _d === void 0 ? DEFAULT_INTERMEDIATE_FILES_STRATEGY : _d, _e = _a.isVerbose, isVerbose = _e === void 0 ? DEFAULT_IS_VERBOSE : _e;
                         // TODO: !!!!!! Does this work in browser? Make it work.
                         if (source.url === null) {
                             throw new KnowledgeScrapeError('Website scraper requires URL');
@@ -5851,10 +5840,6 @@ var WebsiteScraper = /** @class */ (function () {
                             }]))();
                         reader = new Readability(jsdom.window.document);
                         article = reader.parse();
-                        console.log(article);
-                        return [4 /*yield*/, forTime(10000)];
-                    case 2:
-                        _g.sent();
                         html = (article === null || article === void 0 ? void 0 : article.content) || (article === null || article === void 0 ? void 0 : article.textContent) || jsdom.window.document.body.innerHTML;
                         // Note: Unwrap html such as it is convertable by `markdownConverter`
                         for (i = 0; i < 2; i++) {
@@ -5866,16 +5851,16 @@ var WebsiteScraper = /** @class */ (function () {
                         return [4 /*yield*/, getScraperIntermediateSource(source, {
                                 rootDirname: rootDirname,
                                 cacheDirname: cacheDirname,
-                                isCacheCleaned: isCacheCleaned,
+                                intermediateFilesStrategy: intermediateFilesStrategy,
                                 extension: 'html',
                                 isVerbose: isVerbose,
                             })];
-                    case 3:
+                    case 2:
                         cacheFilehandler = _g.sent();
                         return [4 /*yield*/, writeFile(cacheFilehandler.filename, html, 'utf-8')];
-                    case 4:
+                    case 3:
                         _g.sent();
-                        markdown = markdownConverter.makeMarkdown(html, jsdom.window.document);
+                        markdown = this.showdownConverter.makeMarkdown(html, jsdom.window.document);
                         return [2 /*return*/, __assign(__assign({}, cacheFilehandler), { markdown: markdown })];
                 }
             });
@@ -5919,7 +5904,6 @@ var WebsiteScraper = /** @class */ (function () {
     return WebsiteScraper;
 }());
 /**
- * TODO: !!!!!! Put into separate package
  * TODO: [👣] Scraped website in .md can act as cache item - there is no need to run conversion each time
  * TODO: [🪂] Do it in parallel 11:11
  * Note: No need to aggregate usage here, it is done by intercepting the llmTools