理解了 Generative AI 幾個基本概念，以及利用語言模型的能力判斷文字/圖片的能力，並帶入程式開發的範例。接著要談開發 Generative AI 應用時的困難。

這一片主要是以開發者角度之一，技術工具的選擇與使用困難。

大家都聽過 LangChain, Llamaindex, Azure Power Platform, Google Agent Builder, 這麼多個工具，到底該選擇哪一個作為開發工具呢？

我們先以之前提到的 RAG 範例，除了語言模型的選擇，還有資料的來源，清理，建立 Embedding 等等，加上前端查詢畫面，部署與維護，以及後續的資料收集，模型重新訓練。

另外，針對數入特定資料的整理與建立索引，以及後續的資料更新，也是一個困難的問題。

我們在接下來的篇幅會繼續深入討論。

English

After understanding the basic concepts of Generative AI and the ability to judge text/images using language models, and bringing examples of software development. Next, let's talk about the difficulties of developing Generative AI applications.

This article is mainly from the perspective of developers, the difficulties of choosing and using technical tools.

Everyone has heard of LangChain, Llamaindex, Azure Power Platform, Google Agent Builder, so many tools, which one should be chosen as a development tool?

Let's take the RAG example mentioned earlier. In addition to the choice of language model, there are data sources, cleaning, building Embedding, etc., plus the front-end query screen, deployment and maintenance, and subsequent data collection, model retraining.

In addition, the organization and indexing of specific data inputs, as well as the subsequent data updates, are also difficult problems.

We will continue to discuss in the following articles.

本篇同步發表於 iThome 與 個人電子報