moeru-ai/airi

Clear and concise description of the problem As a developer using AIRI with Japanese TTS, I want assistant messages to retain natural Japanese orthography in chat while allowing the speech pipeline to receive an explicit pronunciation form, so that ambiguous kanji, names, and domain-specific terms are pronounced correctly. Today the same streamed literal is used for both the visible assistant message and TTS. Japanese TTS backends may infer the wrong reading for kanji or proper nouns. Writing the whole response in kana avoids the pronunciation error but makes the chat transcript less natural and less readable. Suggested solution / Ideas Introduce separate display and speech projections for assistant text. One possible input convention is silent ruby metadata: Raw model output: ｜約束《やくそく》は｜約束《やくそく》です。 Display text: 約束は約束です。 Speech text: やくそくはやくそくです。 The exact syntax is flexible; the important co

Clear and concise description of the problem It would be really helpful (for our use case) if we could have an adjustable temperature parameter. Currently, the model's responses are highly emotionally expressive and subjective, which hampers accurate error analysis. I have searched the settings but cannot find the option. If it has already been included (apologize if I missed it), I would be grateful for any pointers on how to use it. It would be desirable to expose a temperature parameter so that users can fine-tune the model’s behaviour. Suggested solution / Ideas Take DeepSeek as an example. https://api-docs.deepseek.com/zh-cn/api/create-chat-completion Above is the document. From it, we can see the key temperature can be changed in the request. Other models might be similar, I guess. Alternative 也许修改描述可以达到类似的效果，其相较于修改温度而言，已经实现，但操作不如修改温度直观。 Additional context N/A Validations Follow our Code o

Describe the bug Issue As expected tiling manager tiles the window, full screening the program results in screen being only AIRI. There are workarounds to this but both are prevented. Float window, stretch to fill screen and pin - Click-through does not work. Float window, resize to corner of screen and pin - Edge decorations clearly indicate it's intended to fill screen and still of course click-through will not work but I can manually move the window. Also this may be out of scope with this issue but could be related but is AIRI supposed to be movable within the window? I was certain that was a thing a while ago. I don't remember these issues being a thing in earlier versions so there has been a regression for me. Wish I remembered the versions I believed worked correctly. Looking forward to improvements and more features, not quite there yet but it's close! System Info Linux 6.19.11-1-cachyos Hyprland 0.54.3 (Wayland) Validations Follow our Code of Conduct Read the Contributin

Clear and concise description of the problem 摘要 希望能原生支持 Fish Audio 作为 TTS 提供商。他们生成的速度和质量都还不错，而且也已经有很多人生成过其他二次元角色语音，用起来会比较方便 Fish Audio API 文档：https://fish.audio/developers/ Suggested solution / Ideas 临时解决方案 在等待原生支持的同时，我让claude写了一个本地代理脚本，将 OpenAI 兼容格式的 TTS 请求转换为 Fish Audio 的 API 格式。 将软件的 Base URL 设置为 http://127.0.0.1:8880 即可直接使用。 配置步骤（新手友好） 步骤 1 和 2 只需要做一次。 之后每次使用，只需运行 python fish_proxy.py 启动服务即可。 步骤 1 — 一次性配置 1a. 保存并配置脚本 从此 Gist 下载 fish_proxy.py：https://gist.github.com/Find-A-Name-Is-Hard/c404da4580895d2f04f3bfb1e36c3fe4 保存到任意位置，例如 C:\fish_proxy\fish_proxy.py。 用记事本打开文件（右键 → 打开方

Clear and concise description of the problem Would be cool to one-click import, or at least analyze the chat sessions exported from other apps like chatgpt and save to memory. Suggested solution / Ideas No response Alternative No response Additional context No response Validations Follow our Code of Conduct Read the Contributing Guide. Check that there isn't already an issue that request the same feature to avoid creating a duplicate.