应用不自带 Ollama 服务，必须先自行搭好 Ollama 主机并在应用内填写 API 端点；桌面端仍属实验性功能，部分功能可能异常；Windows 安装包未签名，需在 Defender 提示中选“更多信息”再“仍要运行”；Linux 启动报 libgtk-3.so.0 缺失时需安装 packagekit-gtk3-module
从 releases 页面下载对应平台安装包；Linux 便携版解压后执行 ./ollama 硬件要求按 README 说明
README 未说明具体显存或 GPU 要求；应用本身不托管 Ollama 服务，需另有一台运行 Ollama 的主机，硬件要求取决于该主机。
Ollama 的现代化客户端，通过 API 连接本地或局域网内的 Ollama 服务，聊天数据不出内网
Ollama 的现代化跨平台客户端，支持 Android、Windows、Linux，让本地大模型对话和管理更顺手。
核心能力
- 通过 Ollama API 端点连接远端服务，不在本机托管模型
- 提供模型选择器与侧边栏的聊天界面
- 支持 Android 及实验性的 Windows、Linux 桌面端
- 实验性语音模式，细节见 Wiki 文档
容易踩的坑
内容由模型根据 README 生成，可能不完整或不准确，以官方文档为准。看原文 README
- 1.2.0v1.2.0最新正式版2024-09-14 · 2024-09-14 13:30 UTC
Hi there everybody 👋
It's been some time since the last update, three months to be precise (god, time passes quickly). I've been collecting various features and improvements to the app. And today's the day I'll be releasing them all into the wild.
As always, if you stumble across any errors or inconsistencies, or you have an idea for a feature, feel free to open a new issue report. Have fun.
~ Jakob
Localization
Thanks to many users on Crowdin, I'm happy to announce that Ollama App will be available in three new languages:
- Chinese (Simplified), thanks to Ruifeng and Pinkman Zhang
- Italian, thanks to Intranox
- Turkish, thanks to ShLerP
If you'd like to see your language added as well, help to localize the project. Read more in the Contribution Guide.
Added features
- Ability to add models
- You remember seeing the message that adding models in the app is not supported? That's a thing of the past, in the latest version of Ollama App, you'll be able to download your favorite model directly from within the app
- Timeout multiplier
- Were you ever in the situation that a request for a message fails because of your slow internet connection or because your host is too slow? Timeout multiplier is the solution. It'll multiply every timeout value in the app, so those situations can be prevented
- Toggle to disable system message
- This is implemented, so the system message specified in the Modelfile of the model can be used. It is used global and not model specific, maybe that'll be added in the future
- Ollama App for Desktop
- As you might have already noticed, down below you're able to download a Windows and a Linux version of the app. Yes, that's correct; Flutter, the underlying framework, enables me to easily port Ollama App to multiple platforms without a major rewrite.
- You should be able to use the app in production use, but (potentially severe) issues might occur. Please make an issue report if you come across any
- Learn how to install here
Improvements
- Reworked titles to be more stream lined, and actually working (they were broken before)
- No restart is required when changing theme or brightness of the app
- The new theme worker makes the app more fluent that ever before
- Android 15's predictive back is now supported
- Tooltips are now available in most spots
- Voice Mode reports if TTS is available
- Improved update worker
- Automatically enable show model tags if there are two identical models installed (only works if not toggled before)
- All licenses are now shown in the About settings section
- Technical: Properly implemented the toggle, it's no longer hacky
- Various other additions ...
- 1.1.1v1.1.12024-06-13 · 2024-06-13 19:10 UTC
查看更新说明
Hi everyone 👋
this update fixes a "critical" issue I discovered while adding the new Voice Mode. Basically, all messages would be mixed up when sent to the model, and I was already wondering about the bad outputs by Ollama. Hopefully that fixes it.
Another big thing in this update is Voice Mode. While it's still experimental, it is already a fully working phone call with an AI model. The feature hasn't undergone massive testing and has to manually be enabled in settings, so expect running into issues. I'm of course having an eye on the issue tab, so feel free to report bugs there. (Note: there are a few new permissions required. They are only asked for in case you want to enable Voice Mode in settings and will only be used for that)
Your Material You theme is now supported. Enabling it in settings will allow the app to use the device theme.
~ Jakob
- 1.1.0v1.1.02024-06-08 · 2024-06-08 20:57 UTC
查看更新说明
Hi everyone 👋
In release, I reworked the settings UI; that was the main focus. I wasn't happy how clunky everything looked.
Here are the more detailed changelogs:
- Grouped settings UI
- Added option to keep models loaded on the server
- Model tags are now greyed out to give the focus to the model name
- Fixed icons in dark mode
- Version now visible in settings
I'm open for any feedback. ~ Jakob
- 1.0.3v1.0.32024-06-05 · 2024-06-05 19:03 UTC
查看更新说明
Hi everyone 👋
this release includes the following changes:
- Fixed framerate issues on some devices
- Added update dialog
- Extended sidebar drag space
- Added updater
~ Jakob
- 1.0.2v1.0.22024-06-03 · 2024-06-03 15:08 UTC
查看更新说明
Hello everyone 👋
I added a few fixes and a few small changes.
- Added toggle to enable tags in model selector
- Fixed tips being buggy
- Edited messages now have to have a content
- The send button isn't shown constantly anymore
- The send button has a new design (let me know what you think, if not enough like it, I'll revert it)
~ Jakob
用户点评
登录用户的真实使用记录，可按版本和硬件筛选共 0 条
还没有人点评。第一个说出真实体验的人最有价值。
Ollama App
A modern and easy-to-use client for Ollama. Have the greatest experience while keeping everything private and in your local network.
 |  |  |  |
|---|
[!IMPORTANT] This app does not host a Ollama server on device, but rather connects to one using its api endpoint. You don't know what Ollama is? Learn more at ollama.com.
Getting Started
Ollama App has a pretty simple and intuitive interface to be as open as possible. Everything just works out of the box, you just have to follow the next steps.
Installation
You'll find the latest recommended version of the Ollama App under the releases tab. Download the correct executable onto your device and install it. If you want to install on a desktop platform, you might also have to follow the steps listed below, under Ollama App for Desktop.
Alternatively, you can also download the app from any of the following stores:
That's it, you've successfully installed Ollama App! Now just proceed to Initial Setup or alternatively Setup below.
Ollama App for Desktop
[!WARNING] This is still an experimental feature! Some functions may not work as intended. If you come across any errors, please create a new issue report.
There are a few things you might have to keep in mind if you're planning to use the experimental desktop support.
Windows
The Windows version is provided in the form of an installer, you can find it attached on the latest release. It's not signed, you might have to dismiss the Windows Defender screen by pressing "View More" > "Run Anyway".
Windows app data is kept at: C:\Users\[user]\AppData\Roaming\JHubi1\Ollama App
Linux
The Linux version is provided in the form of a portable executable, you can find it attached on the latest release. To start it, execute ./ollama in the extracted folder.
If a message like error while loading shared libraries: libgtk-3.so.0: cannot open shared object file appears when you start the app, run the following commands:
sudo apt-get updatesudo apt-get upgradesudo apt-get install packagekit-gtk3-module
Linux app data is kept at: /home/[user]/.local/share/ollama
Setup
The most difficult part is setting up the host. To learn more visit the wiki guide on how to do so. After setting up, you normally don't have to enter it again.
And you're done! Just start chatting with your local AI and have fun!
[!TIP] The new Voice Mode is now avaliable as an experimental feature. Learn more about it in the documentation.
Documentation
The documentation for components, settings, functions, etc. has moved to the Wiki Page of this repository. The steps there will be updated with future versions. Still having questions? Feel free to open an issue.
Translations and Contribution
You want to help me make this project even better? Great, help is always appresheated.
Ollama App is created using Flutter, a modern and robust frontend framework designed to make a single codebase run on multiple target platforms. The framework itself is based on the Dart programming language.
Read more in the Contribution Guide.