生成式 AI 服務框架 (GAISF) 是企業的中央人工智慧大腦，透過 API 在企業內提供生成式 AI 服務，同時遵守用戶資料隱私規範，不斷改進和擴展以採用更多生成型 AI 服務。

DaVinci Ai platform (達哥) 聊天機器人是一款基於生成式 AI 服務框架 (GAISF)，類似於 ChatGPT 的應用服務。DaVinci Ai platform 被設計為你的個人助理，以提你的日常工作的效率。除此之外，DaVinci Ai platform 同時也是一個開放平台，開發人員可在此建立插件與現有的系統及服務進行整合。

DaVinci Ai platform 是一個開方式平台，我們相容所有市面上的語言模型，不論從閉源的 Azure OpenAI GPT-4, Claude3, 到開源 (llama2, mistral) 或是自行微調的模型 (breeze-7b, breexe 8x7B 由 MediaTek Research 微調的繁體中文大模型)。因此在服務體驗 (模型成功達成任務的穩定性)、資料機敏性 (可提供地端部署模型，讓所有資料都鎖在企業內部)、成本控管角度 (任務有難易之分，難的可請 GPT-4，簡易的可請Claude3 或甚至開源模型，以達到更有性價比的使用生成式 AI 服務) 等都能夠打造符合企業要求的版本。

DaVinci Ai platform 提供兩種開發 plugin 的方式：

❶ 以 OpenAPI 方式

❷ 以 python 寫程式方式開發

因此，以支援 OpenAPI 的方式來說確實相容 OpenAI 的 plugin；然而，若有更多客製化的開發需求，DaVinci Ai platform 提供 python 寫程式的方式開發，可達到更多樣化的方式開發。

◆ 重要元件的介紹、使用方式、上架方式請看這裡。

◆ 開發方式請看這裡與這裡。

答案是 “不會”。生成式 AI 服務提供商 (MSFT) 不能也不會重複使用你的 Prompt、結果或任何客戶數據來重新訓練其自己的模型。

答案是”不會”。我們已申請 “選擇不進行日誌記錄和人工審核程序“，並已獲得微軟的批准。這意味著你的 Prompt 和結果不會記錄在 Azure 雲端服務中。簡而言之，你在使用 Azure OpenAI 服務時不會在雲端留下任何痕跡。

答案是”否”。我們使用的是微軟 Azure OpenAI 服務，而不是 OpenAI ChatGPT 服務。

◆ Chrome 版本需要 Chrome 102 或者更新

◆ Edge 需要使用 Chromium core

❶ Tokens 是 Azure OpenAI GPT 模型（包括 GPT3.5/4）用來計算文本長度的基本單位。它們是字符的組合，有時候與單詞數相同，但並不總是如此。具體來說，它取決於字符數量並包括標點符號或表情符號。這就是為什麼 Tokens 數量通常與單詞數量不同。

❷ 為了準確知道你的文本中花費了多少個 Tokens，你需要將你的文本進行 Tokens 化，意味著運行一個 Tokens 化算法來計數所有的 Tokens。聽起來困難？實際上只需複製粘貼而已，十分簡單。

❸ 你可以使用 https://www.gptcalculator.xyz/ 來計算。

❹ 想知更多關於 Token 的資訊

我們強烈建議從下列資源學習：

❶ 以下是一些極佳的 Prompt 提示技巧範本，供你參考：https://github.com/f/awesome-chatgpt-prompts

❷ 外部資源：大量的影片可以協助你提升 Prompt 提示技巧

The GAI Service Framework (GAISF) is the central AI brain that delivers generative AI services via API services across the enterprise, in compliance with user data privacy, and is continuously improved and extensible to adopt new GAI services.

DaVinci Ai platform chatbot is a web-based, ChatGPT-like application based on the GAI Service Framework (GAISF). DaVinci Ai platform is designed to be a personal assistant to increase your productivity in daily work.  More than this, DaVinci Ai platform is also an open platform for developers to create plugins to integrate with existing system & services.

DaVinci Ai platform is an open platform that is compatible with all language models available on the market, ranging from proprietary models like Azure OpenAI GPT-4 and Claude3, to open-source models (llama2, mistral), and even custom-tuned models (breeze-7b, breeze 8x7B by MediaTek Research tailored for Traditional Chinese). This compatibility allows for a service experience (model success stability), data sensitivity (offering on-premise deployment to keep all data within the company), and cost control (tasks vary in difficulty, with complex ones possibly requiring GPT-4, simpler ones could use Claude3 or even open-source models, achieving a more cost-effective generative AI service) that can meet enterprise requirements.

DaVinci Ai platform offers two ways to develop plugins:

❶ Using the OpenAPI specification

❷ Programming in Python.

Thus, in terms of supporting OpenAPI, it is indeed compatible with OpenAI plugins; however, for more customized development needs, DaVinci Ai platform allows for programming in Python, enabling a more diverse development approach.

◆ For an introduction to the key components, their uses, and listing methods, please refer to link.

◆For development methods, please see here and here.

The Answer is “No“.  GAI service provider (MSFT) cannot and will not reuse your prompt, result or any customer data to retrain its own model. 

The Answer is “No”.  We has applied “opted out of the logging and human review process ” and has been approved by Microsoft. This means that your prompts and results are not logged in the Azure cloud services. In short, no footprint in the cloud while you use Azure OpenAI Service. 

The Answer is “No”.  We use Microsoft Azure OpenAI Service, instead of OpenAI ChatGPT service. 

◆ The Chrome version required must be Chrome 102 or above.

◆ Edge needs to use the Chromium core.

❶ Tokens are the basic unit that Azure OpenAI GPT models (including GPT3.5/4) use to compute the length of a text. They are groups of characters, which sometimes align with words, but not always. In particular, it depends on the number of characters and includes punctuation signs or emojis. This is why the tokens count is usually different from the word count.

❷ In order to know exactly how many tokens spent in your text, you need to tokenize your text, which means running a tokenizer algorithm that will count all the tokens. Sounds difficult? It’s actually as easy as copy/pasting.

❸ You can use https://www.gptcalculator.xyz/  

❹ For more information about Token. 

We recommend you to learn better prompt from the following resources:

❶ External resources : Tons of videos to help you improve your prompt skill

❷ Here are some best prompt template for your reference: https://github.com/f/awesome-chatgpt-prompts