const { NativeEmbedder } = require("../../EmbeddingEngines/native"); const { handleDefaultStreamResponseV2, } = require("../../helpers/chat/responses"); class LocalAiLLM { constructor(embedder = null, modelPreference = null) { if (!process.env.LOCAL_AI_BASE_PATH) throw new Error("No LocalAI Base Path was set."); const { OpenAI: OpenAIApi } = require("openai"); this.openai = new OpenAIApi({ baseURL: process.env.LOCAL_AI_BASE_PATH, apiKey: process.env.LOCAL_AI_API_KEY ?? null, }); this.model = modelPreference || process.env.LOCAL_AI_MODEL_PREF; this.limits = { history: this.promptWindowLimit() * 0.15, system: this.promptWindowLimit() * 0.15, user: this.promptWindowLimit() * 0.7, }; this.embedder = embedder ?? new NativeEmbedder(); this.defaultTemp = 0.7; } #appendContext(contextTexts = []) { if (!contextTexts || !contextTexts.length) return ""; return ( "\nContext:\n" + contextTexts .map((text, i) => { return `[CONTEXT ${i}]:\n${text}\n[END CONTEXT ${i}]\n\n`; }) .join("") ); } streamingEnabled() { return "streamGetChatCompletion" in this; } // Ensure the user set a value for the token limit // and if undefined - assume 4096 window. promptWindowLimit() { const limit = process.env.LOCAL_AI_MODEL_TOKEN_LIMIT || 4096; if (!limit || isNaN(Number(limit))) throw new Error("No LocalAi token context limit was set."); return Number(limit); } async isValidChatCompletionModel(_ = "") { return true; } constructPrompt({ systemPrompt = "", contextTexts = [], chatHistory = [], userPrompt = "", }) { const prompt = { role: "system", content: `${systemPrompt}${this.#appendContext(contextTexts)}`, }; return [prompt, ...chatHistory, { role: "user", content: userPrompt }]; } async isSafe(_input = "") { // Not implemented so must be stubbed return { safe: true, reasons: [] }; } async getChatCompletion(messages = null, { temperature = 0.7 }) { if (!(await this.isValidChatCompletionModel(this.model))) throw new Error( `LocalAI chat: ${this.model} is not valid for chat completion!` ); const result = await this.openai.chat.completions.create({ model: this.model, messages, temperature, }); if (!result.hasOwnProperty("choices") || result.choices.length === 0) return null; return result.choices[0].message.content; } async streamGetChatCompletion(messages = null, { temperature = 0.7 }) { if (!(await this.isValidChatCompletionModel(this.model))) throw new Error( `LocalAi chat: ${this.model} is not valid for chat completion!` ); const streamRequest = await this.openai.chat.completions.create({ model: this.model, stream: true, messages, temperature, }); return streamRequest; } handleStream(response, stream, responseProps) { return handleDefaultStreamResponseV2(response, stream, responseProps); } // Simple wrapper for dynamic embedder & normalize interface for all LLM implementations async embedTextInput(textInput) { return await this.embedder.embedTextInput(textInput); } async embedChunks(textChunks = []) { return await this.embedder.embedChunks(textChunks); } async compressMessages(promptArgs = {}, rawHistory = []) { const { messageArrayCompressor } = require("../../helpers/chat"); const messageArray = this.constructPrompt(promptArgs); return await messageArrayCompressor(this, messageArray, rawHistory); } } module.exports = { LocalAiLLM, };