PocketPal AI: A Small Language Modeling Chat Tool for Offline Use on iOS and Android Devices

General Introduction

PocketPal AI is an open source mobile app designed to bring Small Language Models (SLMs) directly to your phone for both iOS and Android users. It provides a web-independent AI chat experience that ensures user privacy and supports automatic memory management and a rich selection of models. Developed by Asghar Ghorbani, this program integrates with the Hugging Face model library, allowing users to download and use a wide range of models directly through the app for different needs in chatting and problem solving.

Function List

• Offline AI Chat: Conversations using AI without an internet connection.
• Model diversity: Supports a variety of small language models, such as Danube, Phi, Gemma 2, etc., which can be selected and switched by users according to their needs.
• Automatic memory management: The app automatically uninstalls the model in the background, saving device resources.
• Hugging Face Model Integration: Search, download and use models directly from the Hugging Face model library.
• user: Provides a simple and intuitive interface for users to manage models and conduct dialogs.
• Privacy: All conversations are handled locally, ensuring that data is not transferred to external servers.

Using Help

Installation process

PocketPal AI is very easy to install:

iOS users.

1. Open the App Store.
2. Type "PocketPal AI" in the search bar.
3. Click to download and install the app.

Android users.

1. Open the Google Play Store.
2. Search for "PocketPal AI".
3. Click Install.

Guidelines for use

Launching the application.

• When you open the PocketPal AI app, you'll see a simple home page, usually with a menu button (which may show up as three lines or four dots).

Select and download models.

• Click on the menu to go to the "Models" page. This lists all the pre-installed models and those available for download from Hugging Face.
• Select the model you are interested in and click the "Download" button to start the download. After downloading, click the "Load" button to load the model into memory.

Starting a conversation.

• Select "Chat" from the menu to enter the chat screen.
• Type your question or dialog in the text input box and click Send. The model will give an answer based on the currently loaded model.

Model Advanced Settings.

• On the "Models" page, click on the arrow icon (v) next to the model, and you will see a series of advanced settings options, such as adjusting the temperature, BOS token, etc., which can optimize the output of the model.

Managing Memory.

• PocketPal AI automatically manages the memory usage of models. If you enable the "Auto Offload/Load" feature, the app intelligently unloads models in the background to save resources.

Reproducing the dialog.

• Long-pressing any part of the dialog copies a specific paragraph, while clicking the copy icon at the bottom of the dialog bubble copies the entire response.

Frequently Asked Questions and Solutions.

• Slow model loading or app crashes: First check to see if the Metal API is turned on (iOS only), and try turning it off if it's a problem. If the problem persists, report the issue on GitHub.
• Can't find the required model: You can search for more models by using the search function integrated in Hugging Face. If you still can't find it, you can try to download the model in GGUF format manually and import it with the "Add Local Model" option.

The experience of using PocketPal AI is highly personalized, you can adjust and use different models according to your needs and enjoy instant and private AI conversation services. In the process of using it, you may encounter some minor problems, but they can be effectively solved through community support and continuous updates from the developers. Whether you need a private conversation assistant or want to use AI without internet, PocketPal AI is an open source tool worth trying.

PocketPal AI Android Installer and Model Download

Quark: https://pan.quark.cn/s/06f5cf4ece85 Extract code: D8m2