FlowDown: AI Client for a Smooth Experience on iOS/macOS

General Introduction

FlowDown-App is a lightweight and efficient AI conversation client, developed by a team of developers using Swift and UIKit, aiming to provide users with a fast and smooth intelligent conversation experience. The app is divided into a standard version (FlowDown) and a community version (FlowDown Community). The standard version is available for download through the App Store, with a built-in free cloud model and support for more advanced features; the community version will soon be open-sourced, with the source code published on GitHub, and free for users to access and customize; FlowDown supports all OpenAI-compatible service interfaces, allowing users to access their own Large Language Models (LLMs), and has been designed with privacy protection as a core concept to ensure data security. ensure data security. The highlight of FlowDown is the extremely fast text rendering for a seamless dialog experience.

FlowDown macOS interface

FlowDown IOS Interface

Function List

• Extremely fast text rendering: Conversation text is displayed with virtually no delay, providing a smooth experience.
• Markdown Support: Output in rich text format for easy reading and organization.
• Universal compatibility: All OpenAI-compliant service providers are supported.
• Privacy by Design: User data is not collected by default, conversations stay local.
• Auto Chat Header: Intelligently generate dialog titles to improve management efficiency.
• Comes with LLM support: The Standard Edition supports both offline LLM (based on MLX) and visual LLM.
• Attachment upload: The standard version supports uploading files to interact with AI.
• Web Search: The standard version has a built-in search function to enhance access to information.

Using Help

Installation process

FlowDown is available in two ways: download the standard version from the App Store, or get the Community Edition source code from GitHub. Here are the detailed installation steps:

Standard (FlowDown)

1. system requirements::
   • iOS 16.0 or later (iPhone and iPad supported).
   • macOS 13.0 or later.
2. Download Apps::
   • Open the App Store and search for "FlowDown".
   • Click the "Get" button to complete the download and installation.
3. launch an application::
   • Once the installation is complete, click on the app icon to open it and the free cloud model will be automatically configured for the first time.

Community Edition (FlowDown Community)

1. system requirements: Same as Standard, see GitHub for dependencies.
2. Download source code::
   • Visit https://github.com/Lakr233/FlowDown-App.
   • Click the "Code" button and select "Download ZIP" or use the Git command:

     git clone https://github.com/Lakr233/FlowDown-App.git
     

3. environmental preparation::
   • Make sure Xcode (for Swift development) is installed.
   • Open a terminal, go to the project directory and run the dependency installation (if any):

     pod install # if using CocoaPods
     

4. Build and Run::
   • Open it with Xcode .xcodeproj Documentation.
   • Select the target device (emulator or real machine), click the "Run" button to compile and run.

Main function operation flow

Extremely Fast Text Rendering with Markdown Support

1. launch an application: Open FlowDown and go to the main screen.
2. Input Issues: Enter text into the dialog box, e.g. "Write me a note in Markdown format".
3. Get a reply: Click Send, and the Markdown-formatted text returned by AI is rendered at blazing speed, with headlines, lists, and more clearly rendered.
4. Adjusting the display: Long press on the text area to adjust the font size or copy the content.

Customizing LLM Service Access

1. Go to Settings::
   • Click on the gear icon in the upper right corner to enter the settings page.
2. Add Service::
   • Select Connect AI Services and enter an OpenAI-compatible API key and endpoint, for example:

     API Key: sk-xxxxxxxxxxxxxxxxxxxxxxxxxxx
     Endpoint: https://api.openai.com/v1
     

3. test connection::
   • Click "Test" to confirm that the service is available and save.
4. Switching Models::
   • Select the newly added service in the dialog screen to get started.

Auto Chat Header

1. Start a new conversation::
   • Click on the "New dialog" button and enter a question.
2. Generate Title::
   • The system automatically generates a title based on the input content, such as "Markdown note writing".
3. Management Dialogue::
   • View all conversations in the list on the left and click on the title to quickly switch between them.

Featured Functions

Extremely fast text rendering

• Operation Experience: After entering a question, the response text is displayed almost instantaneously, without waiting for it to load.
• Usage Scenarios: Ideal for users who need real-time interaction, such as taking quick notes on inspiration or handling long conversations.
• Optimization adjustments: The rendering priority can be adjusted in the settings to ensure smooth operation even on low-equipped devices.

Privacy by Design

• local storage: Conversation data is saved on the device by default and is not uploaded to the cloud.
• Service Options: Users can manually specify the LLM service to avoid data leakage.
• View Privacy Policy: The settings page provides links to detailed descriptions of how the data is handled.

Standard Edition Exclusive Features: Offline LLM and Attachment Uploading

• Offline LLM (based on MLX)::
  1. Select "Local Model" in the settings.
  2. Download the supported MLX model files (official documentation is required).
  3. Import the model and use it offline, no network required.
• Attachment upload::
  1. Click the "+" button next to the dialog box.
  2. Select a file (supports images, PDFs, etc.), upload it, and then enter a command, such as "Analyze this image".
  3. The AI returns the results of the analysis, which are displayed in the dialog area.

Web search (standard version)

1. Activate Search::
   • After entering your question in the dialog box, click on the "Search" icon.
2. Getting results::
   • The system returns web page information that is integrated into the AI response.
3. Optimize experience::
   • You can adjust the search priority or turn off this feature in the settings.

Recommendations for use

• network environment: Maintain network stability when using cloud-based models.
• hardware requirement: Offline LLM requires a high performance device, at least 8GB of RAM is recommended.
• Access to documents: Visit https://apps.qaq.wiki/docs/flowdown/ for an official guide if you encounter problems.