const { toChunks } = require("../../helpers"); class OpenAiEmbedder { constructor() { const { Configuration, OpenAIApi } = require("openai"); if (!process.env.OPEN_AI_KEY) throw new Error("No OpenAI API key was set."); const config = new Configuration({ apiKey: process.env.OPEN_AI_KEY, }); const openai = new OpenAIApi(config); this.openai = openai; this.model = process.env.EMBEDDING_MODEL_PREF || "text-embedding-ada-002"; // Limit of how many strings we can process in a single pass to stay with resource or network limits this.maxConcurrentChunks = 500; // https://platform.openai.com/docs/guides/embeddings/embedding-models this.embeddingMaxChunkLength = 8_191; } async embedTextInput(textInput) { const result = await this.embedChunks(textInput); return result?.[0] || []; } async embedChunks(textChunks = []) { // Because there is a hard POST limit on how many chunks can be sent at once to OpenAI (~8mb) // we concurrently execute each max batch of text chunks possible. // Refer to constructor maxConcurrentChunks for more info. const embeddingRequests = []; for (const chunk of toChunks(textChunks, this.maxConcurrentChunks)) { embeddingRequests.push( new Promise((resolve) => { this.openai .createEmbedding({ model: this.model, input: chunk, }) .then((res) => { resolve({ data: res.data?.data, error: null }); }) .catch((e) => { e.type = e?.response?.data?.error?.code || e?.response?.status || "failed_to_embed"; e.message = e?.response?.data?.error?.message || e.message; resolve({ data: [], error: e }); }); }) ); } const { data = [], error = null } = await Promise.all( embeddingRequests ).then((results) => { // If any errors were returned from OpenAI abort the entire sequence because the embeddings // will be incomplete. const errors = results .filter((res) => !!res.error) .map((res) => res.error) .flat(); if (errors.length > 0) { let uniqueErrors = new Set(); errors.map((error) => uniqueErrors.add(`[${error.type}]: ${error.message}`) ); return { data: [], error: Array.from(uniqueErrors).join(", "), }; } return { data: results.map((res) => res?.data || []).flat(), error: null, }; }); if (!!error) throw new Error(`OpenAI Failed to embed: ${error}`); return data.length > 0 && data.every((embd) => embd.hasOwnProperty("embedding")) ? data.map((embd) => embd.embedding) : null; } } module.exports = { OpenAiEmbedder, };