Privatemode: an AI chat app that offers end-to-end encryption to protect enterprise data privacy

General Introduction

Privatemode AI is a generative AI service platform focused on data privacy and security, designed to allow users to process sensitive data without fear of privacy breaches. It ensures that data is always encrypted during transmission, storage, and processing through confidential computing technology so that even service providers cannot access user data. The website offers features similar to ChatGPT but focuses on privacy protection and supports the latest open source AI models for both individual and business users. The service is hosted in top EU data centers, with plans to expand to more regions in the future. Whether it's generating content or analyzing documents, Privatemode provides efficient support while protecting data privacy, and is particularly suited to industries that need to handle sensitive information, such as the healthcare, financial and public sectors.

Function List

• End-to-end data encryption: Data is encrypted from the moment it is sent from the device to the moment it is processed, ensuring privacy and security.
• Confidential computing support: Utilize hardware-level encryption to prevent AI models or infrastructure from accessing user data.
• Open Source AI Model Integration: Support the latest open source big language models, such as Meta's Llama series, for users to choose to use.
• Privatemode Chat App: Provides desktop chat tools, supporting Windows and macOS, for everyday interactions.
• API Services: Supports developers in building privacy-preserving AI applications through API integration into user projects.
• Document Analysis and Content Generation:: Enables secure processing of documents and generation of text for business scenarios.
• Service Validation: Verify the integrity of the back-end service with hardware-issued cryptographic certificates.

Using Help

Installation and use process

Privatemode AI offers two main ways to use it: either through the Privatemode Chat app on the desktop, or through API integration into your own projects. Below is a detailed how-to guide:

1. Installation and use of the Privatemode Chat application

Installation Steps:

• Download Apps: Visit the official website and find the "Privatemode Chat" download link on the homepage, which supports Windows and macOS systems. Click on the download button of the corresponding system to get the installation package.
• installation process:
  • Windows users: Double-click the downloaded.exefile, follow the prompts to click "Next", select the installation path to complete the installation.
  • macOS users: open the downloaded.dmgDrag the app into the "Apps" folder, you may need to allow apps "from unknown developers" at startup (check in "System Settings - Security & Privacy"). Drag and drop the app into the "Apps" folder.
• initialization: After the installation is complete and you open the app, you will be prompted to enter your API Key. You need to register your account on the official website and get the key (see "Get API Key" section below). Enter the key and click "Verify" to enter the main interface.

Function Operation:

• Chat Interaction: The main interface is similar to a common chat tool, with a dialog list on the left and an input box on the right. Enter a question or command in the input box, such as "Write an email for me" or "Analyze this report", click send and the AI will return the result within a few seconds.
• Upload a document: Click the "Upload" button at the top of the interface, support for PDF, TXT and other formats, after uploading, you can enter commands such as "summarize the content of the document", AI will be based on the encrypted document processing and return results.
• Language Support: Multi-language input is supported by default, you can ask a question in Chinese directly, such as "reply me in German", AI will switch the language output.

2. Installation and use of the API

Installation Steps:

• Getting the API key: After registering an account on the official website, log in and go to the "Developer" page, click "Generate API key", copy the key and save it.
• Installation of Proxy: The Privatemode API needs to run through a proxy (privatemode-proxy) to enable encrypted communication. A Docker installation is recommended:
  1. Make sure you have Docker installed on your system (downloadable from docker.com).
  2. Open a terminal and run the following command to start the agent:

     docker run -p 8080:8080 ghcr.io/edgelesssys/privatemode/privatemode-proxy:latest --apiKey <你的API密钥>
     

  3. When the agent starts, it listens on local port 8080 by default, which can be changed via thelocalhost:8080Access.

API function operation:

• Send Request: Call the API using a programming language (e.g. Python). take chat as an example, the sample code is as follows:

  import requests
  url = "http://localhost:8080/v1/chat/completions"
  headers = {"Content-Type": "application/json"}
  data = {
  "model": "ibnzterrell/Meta-Llama-3.3-70B-Instruct-AWQ-INT4",
  "messages": [{"role": "user", "content": "你好，请介绍一下自己"}]
  }
  response = requests.post(url, headers=headers, json=data)
  print(response.json())
  

  The return result will contain the AI's response, and the data is encrypted and transmitted throughout.

• Model Selection: Indatamodifications"model"Fields to switch between different open source models, the list of supported models can be viewed in the official website documentation.
• Advanced Features: Support upload file analysis, you need to convert the file to Base64 encoding and then attached to the request, the specific method refer to the official website "API documentation".

3. Operation of special functions

• Privacy Verification: Each time the service is used, the agent will automatically verify the integrity of the backend (via hardware encryption certificate). If the verification fails, the interface or API will indicate that the service is not trustworthy to ensure data security.
• file processing: After uploading a document in the Chat app, you can enter detailed instructions, such as "Extract key data from the document and generate a table," and the AI will complete the task in a privacy-protected manner./v1/documentEndpoints implement similar functionality.
• Multi-language support: Enter "Please reply in Chinese" or "Please generate content in German", and the AI will adjust the output language according to the command, making it suitable for internationalization.

Tips and Precautions for Use

• network requirement: Ensure that the device is connected to a stable network, and do not close the terminal window while the agent is running, otherwise the API service will be interrupted.
• Key Management: The API key is the only credential to access the service, do not leak it to others. If you suspect leakage, you can regenerate it on the official website.
• performance optimization: If processing large files or complex tasks is slow, try using a quantization model (e.g. AWQ-INT4), just specify it in the API request.

With these steps, users can quickly get started with Privatemode AI and utilize its powerful privacy protection and AI features for daily tasks or development projects.