一句话概括：我们不设账号、不接广告与分析 SDK，也不保存你的学习内容（只留一行不含内容的匿名请求指标，见第三节）。你的学习内容存在你自己的设备和 iCloud 里；只有生成练习和识别语音的那一刻，相关内容才会被转发给 AI 服务商处理，我们自己的服务器不留任何副本。服务商各自的留存政策不由我们控制，第三节逐个写明。
我们不收集任何可识别你个人身份的信息。本应用没有注册、登录或账号体系，不索取姓名、邮箱、手机号，也不读取通讯录、位置、相册或广告标识符（IDFA）。
为实现功能，以下内容会在你主动使用对应功能时被处理：
应用通过我们运行在 Cloudflare Workers 上的中转服务（butsu-japanese-ai.rokajun11.workers.dev）调用第三方 AI。该中转服务不保存你的任何文本、录音或学习内容——它只做转发，内容处理完即丢弃。两处例外，都不含你的内容：
① 匿名请求指标：为了排查故障和盯住用量，每次调用会记一行只有接口名、状态码、错误码、耗时、token 数、模型结束原因的元数据，定期清理（保留不超过 14 天），其中没有任何文本、录音或可识别到个人的信息。
② 朗读音频缓存：合成好的语音会在 Cloudflare 的边缘缓存里存最多 30 天，这样同一句重听时不必重新合成。缓存的是被朗读的那句学习语言文本——通常是应用生成的例句，也可能是由你输入的句子改写成的目标语言说法；按内容哈希索引，不与任何账号或设备关联。
whisper-large-v3）——接收你的录音，返回转写文字。仅在使用语音输入时发生。据其公开文档，推理数据默认不留存，仅在故障排查与滥用调查场景下临时保留，最长 30 天。deepseek-v4-flash）——接收你的文本内容（场景描述、对话语句），返回学习内容、翻译、批改建议。其付费 API 默认不将内容用于训练，但未公开通用的零留存承诺，且存在服务端上下文缓存。因此这部分文本可能在其服务器上留存一段时间。各服务商按其自身条款处理数据，其留存行为不在我们控制范围内。除上述必要的功能调用外，我们不向任何第三方出售、出租或共享你的任何内容。
因此请不要在对话里输入真实姓名、住址、证件号、银行卡号等敏感信息。练习场景用化名就够了。
高级版订阅完全通过 Apple 的 App 内购买完成。付款方式、账单地址等支付信息由 Apple 处理，开发者既看不到也不存储。应用只从系统获知「当前是否有有效订阅」这一个状态。
本应用面向普通语言学习者，不针对 13 岁以下儿童设计，也不会有意收集儿童的个人信息。
由于我们不持有你的数据，你随时可以自行完全清除：
本政策如有更新，将在此页面发布新版本并更新页首的生效日期。重大变更会在应用内提示。
对本政策或数据处理有任何疑问，请邮件联系：junesui00@gmail.com
In short: There are no accounts, no ads, and no analytics SDKs, and we do not store your study content (only one content-free anonymous metrics row per call — see section 3). Your study data lives on your own device and in your own iCloud. Content is forwarded to AI providers only at the moment you generate a lesson or use voice input, and our own server keeps no copy. Those providers’ retention policies are outside our control and are spelled out per-provider in section 3.
We collect no personally identifiable information. The app has no sign-up, sign-in, or account system. It does not ask for your name, email, or phone number, and does not access your contacts, location, photos, or advertising identifier (IDFA).
The following is processed only when you actively use the corresponding feature:
The app reaches third-party AI providers through our relay running on Cloudflare Workers (butsu-japanese-ai.rokajun11.workers.dev). That relay stores none of your text, audio, or study content — it only forwards requests and discards the content once done. Two exceptions, neither of which contains your content:
① Anonymous request metrics: each call appends one row containing only the endpoint name, status code, error code, duration, token counts, and the model's finish reason. These are purged regularly (retained no longer than 14 days) and contain no text, audio, or personally identifiable information.
② Read-aloud audio cache: synthesized speech is held in Cloudflare's edge cache for up to 30 days so replaying the same line doesn't re-synthesize it. What is cached is the line of target-language text that was spoken — usually an app-generated example, and possibly a target-language rendering of something you typed; indexed by a hash of its content and tied to no account or device.
whisper-large-v3) — receives your audio, returns a transcript. Only when you use voice input. Per their published documentation, inference data is not retained by default; it may be held temporarily for troubleshooting and abuse investigation, for up to 30 days.deepseek-v4-flash) — receives your text (scenario descriptions, conversation lines), returns study material, translations, and corrections. Their paid API does not use content for training by default, but they publish no general zero-retention commitment and operate server-side context caching. This text may therefore persist on their servers for some period.Each provider handles data under its own terms, and their retention behaviour is outside our control. Beyond these functional calls, we do not sell, rent, or share any of your content with anyone.
For this reason, please do not type real names, addresses, ID numbers, or card details into conversations. Made-up details are fine for practice.
Premium subscriptions are handled entirely through Apple In-App Purchase. Payment methods and billing details are processed by Apple; the developer never sees or stores them. The app learns only whether an active subscription exists.
This app is intended for general language learners. It is not directed at children under 13 and does not knowingly collect personal information from them.
Because we hold none of your data, you can erase it entirely at any time:
Updates to this policy will be published on this page with a new effective date. Material changes will also be announced in the app.
Questions about this policy or our data handling: junesui00@gmail.com