[inference] NL-to-ESQL: improve doc generation (#192378)

## Summary

Follow-up of #190433

Fix [#192762](#192762)

- Cleanup and refactor the documentation generation script
- Make some tweak to the documentation to improve efficiency and make a
better user of tokens
- Perform human review of the generated content to make sure everything
is accurate

---------

Co-authored-by: kibanamachine <[email protected]>

x-pack/plugins/inference/scripts/evaluation/scenarios/esql/index.spec.ts

@@ -192,7 +192,7 @@ const buildTestDefinitions = (): Section[] => {
         {
           title: 'Generates a query to show employees filtered by name and grouped by hire_date',
           question: `From the employees index, I want to see how many employees with a "B" in their first name
-      where hired each month over the past 2 years.
+      were hired each month over the past 2 years.
       Assume the following fields:
       - hire_date
       - first_name
@@ -208,10 +208,10 @@ const buildTestDefinitions = (): Section[] => {
          (which can be read the same backward and forward), and then return their last name and first name
       - last_name
       - first_name`,
-          expected: `FROM employees
-      | EVAL reversed_last_name = REVERSE(last_name)
-      | WHERE TO_LOWER(last_name) == TO_LOWER(reversed_last_name)
-      | KEEP last_name, first_name`,
+          criteria: [
+            `The assistant should not provide an ES|QL query, and explicitly mention that there is no
+            way to check for palindromes using ES|QL.`,
+          ],
         },
         {
           title: 'Generates a query to show the top 10 domains by doc count',

x-pack/plugins/inference/scripts/load_esql_docs/extract_doc_entries.ts

@@ -0,0 +1,288 @@
+/*
+ * Copyright Elasticsearch B.V. and/or licensed to Elasticsearch B.V. under one
+ * or more contributor license agreements. Licensed under the Elastic License
+ * 2.0; you may not use this file except in compliance with the Elastic License
+ * 2.0.
+ */
+
+import Fs from 'fs/promises';
+import Path from 'path';
+import fastGlob from 'fast-glob';
+import $, { load, Cheerio, AnyNode } from 'cheerio';
+import { partition } from 'lodash';
+import { ToolingLog } from '@kbn/tooling-log';
+import pLimit from 'p-limit';
+import { ScriptInferenceClient } from '../util/kibana_client';
+import { convertToMarkdownPrompt } from './prompts/convert_to_markdown';
+import { bindOutput, PromptCaller } from './utils/output_executor';
+
+/**
+ * The pages that will be extracted but only used as context
+ * for the LLM for the enhancement tasks of the documentation entries.
+ */
+const contextArticles = [
+  'esql.html',
+  'esql-syntax.html',
+  'esql-kibana.html',
+  'esql-query-api.html',
+  'esql-limitations.html',
+  'esql-cross-clusters.html',
+  'esql-examples.html',
+  'esql-metadata-fields.html',
+  'esql-multi-index.html',
+];
+
+interface ExtractedPage {
+  sourceFile: string;
+  name: string;
+  content: string;
+}
+
+export interface ExtractedCommandOrFunc {
+  name: string;
+  markdownContent: string;
+  command: boolean;
+}
+
+export interface ExtractionOutput {
+  commands: ExtractedCommandOrFunc[];
+  functions: ExtractedCommandOrFunc[];
+  pages: ExtractedPage[];
+  skippedFile: string[];
+}
+
+export async function extractDocEntries({
+  builtDocsDir,
+  log,
+  inferenceClient,
+}: {
+  builtDocsDir: string;
+  log: ToolingLog;
+  inferenceClient: ScriptInferenceClient;
+}): Promise<ExtractionOutput> {
+  const files = await fastGlob(`${builtDocsDir}/html/en/elasticsearch/reference/master/esql*.html`);
+  if (!files.length) {
+    throw new Error('No files found');
+  }
+
+  const output: ExtractionOutput = {
+    commands: [],
+    functions: [],
+    pages: [],
+    skippedFile: [],
+  };
+
+  const executePrompt = bindOutput({
+    output: inferenceClient.output,
+    connectorId: inferenceClient.getConnectorId(),
+  });
+
+  const limiter = pLimit(10);
+
+  await Promise.all(
+    files.map(async (file) => {
+      return await processFile({
+        file,
+        log,
+        executePrompt,
+        output,
+        limiter,
+      });
+    })
+  );
+
+  return output;
+}
+
+async function processFile({
+  file: fileFullPath,
+  output,
+  executePrompt,
+  log,
+  limiter,
+}: {
+  file: string;
+  output: ExtractionOutput;
+  executePrompt: PromptCaller;
+  log: ToolingLog;
+  limiter: pLimit.Limit;
+}) {
+  const basename = Path.basename(fileFullPath);
+  const fileContent = (await Fs.readFile(fileFullPath)).toString('utf-8');
+
+  if (basename === 'esql-commands.html') {
+    // process commands
+    await processCommands({
+      fileContent,
+      log,
+      output,
+      limiter,
+      executePrompt,
+    });
+  } else if (basename === 'esql-functions-operators.html') {
+    // process functions / operators
+    await processFunctionsAndOperators({
+      fileContent,
+      log,
+      output,
+      limiter,
+      executePrompt,
+    });
+  } else if (contextArticles.includes(basename)) {
+    const $element = load(fileContent)('*');
+    output.pages.push({
+      sourceFile: basename,
+      name: basename === 'esql.html' ? 'overview' : basename.substring(5, basename.length - 5),
+      content: getSimpleText($element),
+    });
+  } else {
+    output.skippedFile.push(basename);
+  }
+}
+
+async function processFunctionsAndOperators({
+  fileContent,
+  output,
+  executePrompt,
+  log,
+  limiter,
+}: {
+  fileContent: string;
+  output: ExtractionOutput;
+  executePrompt: PromptCaller;
+  log: ToolingLog;
+  limiter: pLimit.Limit;
+}) {
+  const $element = load(fileContent.toString())('*');
+
+  const sections = extractSections($element);
+
+  const searches = [
+    'Binary operators',
+    'Equality',
+    'Inequality',
+    'Less than',
+    'Less than or equal to',
+    'Greater than',
+    'Greater than or equal to',
+    'Add +',
+    'Subtract -',
+    'Multiply *',
+    'Divide /',
+    'Modulus %',
+    'Unary operators',
+    'Logical operators',
+    'IS NULL and IS NOT NULL',
+    'Cast (::)',
+  ];
+
+  const matches = ['IN', 'LIKE', 'RLIKE'];
+
+  const [operatorSections, allOtherSections] = partition(sections, (section) => {
+    return (
+      matches.includes(section.title) ||
+      searches.some((search) => section.title.toLowerCase().startsWith(search.toLowerCase()))
+    );
+  });
+
+  const functionSections = allOtherSections.filter(({ title }) => !!title.match(/^[A-Z_]+$/));
+
+  const markdownFiles = await Promise.all(
+    functionSections.map(async (section) => {
+      return limiter(async () => {
+        return {
+          name: section.title,
+          markdownContent: await executePrompt(
+            convertToMarkdownPrompt({ htmlContent: section.content })
+          ),
+          command: false,
+        };
+      });
+    })
+  );
+
+  output.functions.push(...markdownFiles);
+
+  output.pages.push({
+    sourceFile: 'esql-functions-operators.html',
+    name: 'operators',
+    content: operatorSections.map(({ title, content }) => `${title}\n${content}`).join('\n'),
+  });
+}
+
+async function processCommands({
+  fileContent,
+  output,
+  executePrompt,
+  log,
+  limiter,
+}: {
+  fileContent: string;
+  output: ExtractionOutput;
+  executePrompt: PromptCaller;
+  log: ToolingLog;
+  limiter: pLimit.Limit;
+}) {
+  const $element = load(fileContent.toString())('*');
+
+  const sections = extractSections($element).filter(({ title }) => !!title.match(/^[A-Z_]+$/));
+
+  const markdownFiles = await Promise.all(
+    sections.map(async (section) => {
+      return limiter(async () => {
+        return {
+          name: section.title,
+          markdownContent: await executePrompt(
+            convertToMarkdownPrompt({ htmlContent: section.content })
+          ),
+          command: true,
+        };
+      });
+    })
+  );
+
+  output.commands.push(...markdownFiles);
+}
+
+function getSimpleText($element: Cheerio<AnyNode>) {
+  $element.remove('.navfooter');
+  $element.remove('#sticky_content');
+  $element.remove('.edit_me');
+  $element.find('code').each(function () {
+    $(this).replaceWith('`' + $(this).text() + '`');
+  });
+  return $element
+    .find('.section,section,.part')
+    .last()
+    .text()
+    .replaceAll(/([\n]\s*){2,}/g, '\n');
+}
+
+export function extractSections(cheerio: Cheerio<AnyNode>) {
+  const sections: Array<{
+    title: string;
+    content: string;
+  }> = [];
+  cheerio.find('.section .position-relative').each((index, element) => {
+    const untilNextHeader = $(element).nextUntil('.position-relative');
+
+    const title = $(element).text().trim().replace('edit', '');
+
+    untilNextHeader.find('svg defs').remove();
+    untilNextHeader.find('.console_code_copy').remove();
+    untilNextHeader.find('.imageblock').remove();
+    untilNextHeader.find('table').remove();
+
+    const htmlContent = untilNextHeader
+      .map((i, node) => $(node).prop('outerHTML'))
+      .toArray()
+      .join('');
+
+    sections.push({
+      title: title === 'STATS ... BY' ? 'STATS' : title,
+      content: `<div><h1>${title}</h1> ${htmlContent}</div>`,
+    });
+  });
+
+  return sections;
+}