Intlayer

import { readAsset } from 'utils:asset'; import { generateObject } from 'ai'; import { z } from 'zod'; import type { AIConfig, AIOptions } from '../aiSdk'; type Tag = { key: string; description?: string; }; export type AuditDictionaryMetadataOptions = { fileContent: string; tags?: Tag[]; aiConfig: AIConfig; applicationContext?: string; }; export type AuditFileResultData = { fileContent: { title: string; description: string; tags: string[]; }; tokenUsed: number; }; export const aiDefaultOptions: AIOptions = { // Keep default options }; /** * Audits a content declaration file by constructing a prompt for AI models. * The prompt includes details about the project's locales, file paths of content declarations, * and requests for identifying issues or inconsistencies. */ export const auditDictionaryMetadata = async ({ fileContent, tags, aiConfig, applicationContext, }: AuditDictionaryMetadataOptions): Promise< AuditFileResultData | undefined > => { const CHAT_GPT_PROMPT = readAsset('./PROMPT.md'); const EXAMPLE_REQUEST = readAsset('./EXAMPLE_REQUEST.md'); const EXAMPLE_RESPONSE = readAsset('./EXAMPLE_RESPONSE.md'); // Prepare the prompt for AI by replacing placeholders with actual values. const prompt = CHAT_GPT_PROMPT.replace( '{{applicationContext}}', applicationContext ?? '' ).replace( '{{tags}}', tags ? JSON.stringify( tags .map(({ key, description }) => `- ${key}: ${description}`) .join('\n\n'), null, 2 ) : '' ); // Use the AI SDK to generate the completion const { object, usage } = await generateObject({ ...aiConfig, schema: z.object({ title: z.string(), description: z.string(), tags: z.array(z.string()), }), messages: [ { role: 'system', content: prompt }, { role: 'user', content: EXAMPLE_REQUEST }, { role: 'assistant', content: EXAMPLE_RESPONSE }, { role: 'user', content: fileContent, }, ], }); return { fileContent: object, tokenUsed: usage?.totalTokens ?? 0, }; };