Documentation

使用文档

Gleanbox Guides

Gleanbox 使用教程

Follow the complete loop, connect your own AI, and keep processed knowledge portable with Git.

从第一次收集到完成复习,再配置自己的 AI 与 Git 同步。三个教程都在这里。

Guide contents 教程目录

Guide 1

Use Gleanbox

Gleanbox turns something you encountered into a focused knowledge entry, then brings it back when you need to recall it.

CaptureProcessOrganizePractice
The key idea

Your raw collection is temporary working material. The processed knowledge entry is the durable source of truth used by search, decks, practice, export, and Git sync.

01

Capture the moment before it disappears

Open Collect and tap +. Save text, take a photo, choose an image or file, record audio, or capture a short video. Start quickly; you can remove the irrelevant parts later.

  1. Choose the lightest useful format. Text is fastest, a photo preserves visual context, and audio or video is useful when timing or pronunciation matters.
  2. Do not organize yet. The collection is an inbox, not your permanent library.
  3. Process it when you have time. Items remain under To process until you finish them.
A Japanese welcome sign in a railway station
A real encounter is enough. Here, the Japanese sign becomes the starting point.
The captured image shown in Gleanbox Collect
The item waits in Collect until you are ready to process it.
02

Process only the useful part

Open an item and choose Process. This stage removes noise and turns the encounter into a compact knowledge entry.

  1. Focus the media. Crop, rotate, adjust brightness and contrast, or highlight the important area. Trim audio and video; processed video is limited and compressed before entering the durable library.
  2. Name the target. Enter the word, phrase, sentence, or concept you want to remember.
  3. Add an explanation. Write it yourself, or select Meaning, Usage examples, and Pronunciation before asking your configured AI. Use the additional request field when you need a specific style or focus.
  4. Review and finish. The result becomes a knowledge entry. The original capture stays local so you can reprocess it, and can be cleared later when no longer needed.
Highlighting the useful part of a captured image
Keep the part that carries the memory cue.
AI explanation generated for the Japanese word
Check the explanation before saving the entry.
03

Organize without building silos

Processed entries appear in Kernels. Search the entry text and explanation, add your own notes, or place entries into decks.

  • To plan: the entry is not in a deck yet.
  • In progress: the entry belongs to at least one deck.
  • Mastered: you marked the entry as mastered.
  • Decks are scopes, not containers: one entry can belong to several decks, and removing it from a deck never deletes the entry itself.
Processed entries in the Gleanbox knowledge list
Use search and small decks to keep a growing library manageable.
04

Choose planned review or free practice

Open a deck, then tap Start. The practice page offers two modes:

Planned review
Follows FSRS. Recall first, reveal the answer, then rate your memory. The next due time adapts to your result.
Free practice
Shuffles every entry in the deck for one round. It records a simple count and does not change FSRS scheduling.

Enable local notifications if you want due-review reminders. A planned practice run completes when every included entry has been marked mastered; closing a run manually is permanent.

A Gleanbox recall card before revealing the answer
Try to recall before flipping the card.
FSRS review rating buttons after revealing the answer
Your rating schedules the next planned review.
A practical rhythm

Keep the active batch small. Process a few new items, practise what is due, and add more only when the current set feels manageable.

Guide 2

Get a free Gemini API key

AI is optional. Quick mode connects Gleanbox directly to Google Gemini with your own key; Nixlight does not proxy the request.

Free tier availability varies

Google controls supported regions, models, quotas, and billing requirements. Check the current status in AI Studio. Never send your API key in a screenshot or support message.

01

Create the key in Google AI Studio

  1. Open Google AI Studio · API keys and sign in with your Google account.
  2. Accept the Gemini API terms when prompted.
  3. Select Create API key. New users may receive a default Google Cloud project automatically; otherwise choose or create a project you control.
  4. Copy the new key and keep it private. You only need to paste it into Gleanbox once.

Reference: Google's API key documentation.

02

Connect the key to Gleanbox

  1. Open Settings → AI Settings.
  2. Choose Quick. The Gemini provider, endpoint, and model preset are already filled in.
  3. Paste the key into API key, tap Test, then tap Save.
  4. Back in Settings, choose AI output language. It can follow the interface or use a different language.
SettingsAI SettingsQuickTestSave

Advanced mode is for an OpenAI-compatible endpoint. It can also use one model for image recognition and another for the final explanation. Start with Quick mode unless you already operate another API.

03

Fix common connection errors

MessageWhat to check
Invalid API key / 401Copy the full key again, remove spaces, and confirm it has not been deleted or restricted incorrectly.
Forbidden / 403Check region availability, project access, accepted terms, and whether the Gemini API is available to the account.
Too many requests / 429The current quota is exhausted. Wait for it to reset or review usage and billing in AI Studio.
Model unavailableUpdate Gleanbox first. In Advanced mode, verify the model name supported by your endpoint.

Guide 3

Set up private Git sync

Premium syncs processed knowledge, decks, review history, and processed media to a private repository you own.

  • The raw Collect inbox stays only on the device and does not enter Git.
  • Your access token stays in secure local storage and is not committed.
  • Use a dedicated private repository for each Gleanbox dataset.
  • Git history is preserved so advanced users still have a recovery path.
01A

Option A: prepare GitHub

  1. Create a new private repository, for example gleanbox-data. An empty repository is simplest.
  2. Open GitHub fine-grained personal access tokens.
  3. Choose your account as the resource owner, select Only select repositories, and choose only the Gleanbox repository.
  4. Under Repository permissions, grant Contents: Read and write. No Issues, Actions, Packages, or administration permission is needed.
  5. Generate the token and copy it now; GitHub will not show the complete token again.

Reference: GitHub personal access token documentation.

01B

Option B: prepare GitLab

  1. Create a new private blank project, for example gleanbox-data.
  2. Open your GitLab avatar menu, then Edit profile → Access tokens.
  3. Create a token with the write_repository scope. This scope allows Git pull and push without granting the full API scope.
  4. Copy the token and store it securely.

Reference: GitLab access token scopes.

02

Connect the repository and make the first sync

  1. In Gleanbox, open Settings → Git sync.
  2. Paste the full HTTPS repository URL, such as https://github.com/you/gleanbox-data.git.
  3. Enter your GitHub or GitLab username and paste the personal access token.
  4. Tap Test connection. Resolve any permission error before continuing.
  5. Tap Save settings, then Sync now. After the first successful sync, edits are queued automatically and the selected interval performs a safety check.
If the repository already contains files

A README-only repository can be adopted after confirmation. A repository containing unrelated data or another Gleanbox dataset is rejected; use a clean repository instead of forcing the histories together.

03

Restore on a new or reinstalled device

  1. Install Gleanbox and unlock Premium with Restore purchases if needed.
  2. Enter the same repository URL, username, and token in Git sync, then save the settings.
  3. Open Settings → Restore from remote Git.
  4. Tap Check Git settings. Gleanbox verifies access and validates the remote dataset before touching local data.
  5. When the check succeeds, tap Start restore and confirm. The device's learning data is replaced by the remote copy, then the local search index is rebuilt.
Restore replaces local learning data

Use it on a new device or only when you intentionally want the remote repository to become the complete local dataset. The remote repository and its Git history are not erased.

Git troubleshooting

ProblemAction
403 / authentication rejectedUse the complete HTTPS URL and verify repository access plus Contents read/write on GitHub or write_repository on GitLab.
Different datasetDo not merge it. Use another empty repository for local data, or use Restore from remote Git if the remote data is the one you want.
No common ancestorThe histories cannot be merged safely. Restore the intended remote dataset or connect local data to a new clean repository.
Sync appears idleOpen Git sync, check Recent activity, then use Sync now. Automatic sync normally runs silently unless an error needs attention.

Need help?

Describe the step and error message, but never include an API key or Git token.

Contact support

教程一

使用 Gleanbox

Gleanbox 把你真实遇到的问题加工成聚焦的知识点,再在需要回忆的时候把它带回来。

收集加工整理复习
先理解一个核心关系

收集箱里的原始素材只是临时工作材料;加工后的知识点才是长期基准数据。搜索、卡组、复习、导出和 Git 同步都围绕知识点进行。

01

先把稍纵即逝的内容收进来

打开收集,点击 +。你可以输入文字、拍照、从相册或文件中选择、录音,或者拍摄一段短视频。先快速保存,之后再去掉无关部分。

  1. 选择最轻的有效形式。文字最快;照片保留视觉语境;发音、动作或时间过程重要时再使用音频和视频。
  2. 此时不用整理。收集页是待办箱,不是永久知识库。
  3. 有空时统一加工。完成前,素材会留在“待处理”中。
车站里的日语欢迎标语
一次真实遇见就够了。这里从车站里的日语「ようこそ」开始。
照片进入 Gleanbox 收集页
素材会留在收集页,等待你有时间再加工。
02

加工时只留下真正有用的部分

打开待处理项目并选择加工。这一阶段负责去除噪音,把偶然遇见的素材变成紧凑、可复习的知识点。

  1. 聚焦素材。裁剪、旋转、调整亮度和对比度,或用荧光笔标出重点。音频和视频可以截取片段;加工后的视频会限制时长并压缩,再进入长期资料库。
  2. 填写目标词句。输入真正想记住的词、短语、句子或概念。
  3. 补充解释。可以自己写,也可以选择含义解释、用法举例和发音,再让已配置的 AI 生成。需要特殊角度时填写“补充要求”。
  4. 检查后完成加工。结果会成为知识点。原始素材继续留在本机,便于重新加工;确认无用后可清理已处理项目。
高亮图片中真正有用的区域
只保留能够唤起记忆的部分。
为日语词语生成 AI 解释
保存知识点前,先检查生成结果是否准确。
03

整理知识点,但不要把数据关进卡组

加工结果会出现在知识点页。你可以搜索本体和解释、补充自己的 Notes,或者把知识点加入卡组。

  • 待安排:还没有加入任何卡组。
  • 进行中:已经属于至少一个卡组。
  • 已掌握:你主动把它标记为已掌握。
  • 卡组只是练习范围:一个知识点可以属于多个卡组;从卡组移除只删除关联,不会删除知识点。
Gleanbox 知识点列表
知识点变多后,用搜索和小批量卡组保持清晰。
04

选择计划复习或自主复习

打开一个卡组,点击开始练习。练习页提供两种模式:

计划复习
按照 FSRS 排期。先回忆、再翻面,然后根据真实记忆情况评分;系统据此计算下次复习时间。
自主复习
随机打乱卡组内全部知识点,完整练习一轮。只记录简单次数,不改变 FSRS 排期。

需要提醒时开启本地通知。计划练习中的知识点全部被你标记为已掌握后,本轮练习完成;手动关闭练习后不能再次开启。

翻面前主动回忆知识点
先尝试回忆,再翻面看答案。
翻面后选择 FSRS 记忆评分
评分会决定下一次计划复习时间。
推荐节奏

让当前批次保持精简:先加工少量新知识点,完成到期复习,觉得负担合适时再加入下一批。

教程二

获取免费的 Gemini API Key

AI 完全可选。快捷模式会使用你自己的 Key,从 Gleanbox 直接连接 Google Gemini;Nixlight 不代理请求。

免费额度以 Google 当前规则为准

支持地区、可用模型、额度和是否要求结算账号都由 Google 决定,请以 AI Studio 显示为准。不要在截图或客服邮件中发送 API Key。

01

在 Google AI Studio 创建 Key

  1. 打开 Google AI Studio · API Keys,登录 Google 账号。
  2. 首次使用时,按页面提示接受 Gemini API 条款。
  3. 点击 Create API key。新用户通常会自动获得默认 Google Cloud 项目;页面要求选择时,选择或创建一个自己控制的项目。
  4. 复制新生成的 Key 并妥善保管。它只需要在 Gleanbox 中粘贴一次。

官方参考:Gemini API Key 文档

02

把 Key 连接到 Gleanbox

  1. 打开设置 → AI 设置
  2. 选择快捷模式。Gemini 服务商、接口地址和模型预设已经填好,不需要自己修改。
  3. 把 Key 粘贴到 API Key,点击测试,成功后点击保存
  4. 回到设置页,打开 AI 输出语言。它可以跟随界面语言,也可以单独选择另一种语言。
设置AI 设置快捷模式测试保存

高级模式用于兼容 OpenAI 的接口,还可以让一个模型识图、另一个模型生成最终解释。没有明确需求时,先使用快捷模式即可。

03

处理常见连接错误

提示检查方法
API Key 无效 / 401重新复制完整 Key,去掉前后空格,并确认它没有被删除或错误限制。
禁止访问 / 403检查地区是否支持、项目权限、条款是否接受,以及该账号能否使用 Gemini API。
请求过多 / 429当前额度已用完。等待额度重置,或在 AI Studio 查看用量与结算设置。
模型不可用先更新 Gleanbox;高级模式还需要检查接口实际支持的模型名称。

教程三

设置 Git 同步

Premium 会把加工后的知识点、卡组、复习记录和加工后媒体同步到你拥有的私有仓库。

  • 收集页的原始素材只保存在设备上,不进入 Git。
  • 访问令牌保存在本机安全存储中,不会写进仓库。
  • 每套 Gleanbox 数据使用一个独立的私有仓库。
  • 远端 Git 历史会完整保留,为高级用户留下手动恢复路径。
01A

方案 A:准备 GitHub

  1. 新建一个私有仓库,例如 gleanbox-data。直接创建空仓库最简单。
  2. 打开 GitHub Fine-grained personal access tokens
  3. Resource owner 选择自己的账号;Repository access 选择 Only select repositories,并且只选刚才的 Gleanbox 仓库。
  4. 在 Repository permissions 中把 Contents 设置为 Read and write。不需要 Issues、Actions、Packages 或管理权限。
  5. 生成并立即复制 Token;离开页面后 GitHub 不会再次显示完整内容。

官方参考:GitHub Personal Access Token 文档

01B

方案 B:准备 GitLab

  1. 新建一个私有空白项目,例如 gleanbox-data
  2. 打开 GitLab 头像菜单,进入 Edit profile → Access tokens
  3. 创建 Token,并勾选 write_repository。这个范围足够进行 Git pull 和 push,不需要授予完整 API 权限。
  4. 复制 Token 并妥善保存。

官方参考:GitLab Access Token Scopes

02

连接仓库并完成首次同步

  1. 在 Gleanbox 中打开设置 → Git 同步
  2. 粘贴完整的 HTTPS 仓库地址,例如 https://github.com/you/gleanbox-data.git
  3. 填写 GitHub 或 GitLab 用户名,并粘贴个人访问令牌。
  4. 点击测试连接。如果权限有误,先修正后再继续。
  5. 点击保存设置,再点击立即同步。首次同步成功后,数据修改会自动排队同步,设定的同步频率还会做保底检查。
如果仓库里已经有文件

只有 README 等说明文件时,可以在确认后使用。仓库包含无关数据或另一套 Gleanbox 数据时会被拒绝;请换一个干净仓库,不要强行拼接两套历史。

03

在新设备或重装后恢复

  1. 安装 Gleanbox;如有需要,先用恢复购买重新解锁 Premium。
  2. Git 同步中填写同一个仓库地址、用户名和 Token,然后保存设置。
  3. 打开设置 → 从远端 Git 恢复
  4. 点击检测 Git 设置。Gleanbox 会先检查访问权限和远端数据标识,此时不会修改本机数据。
  5. 检测成功后点击开始恢复并确认。本机学习数据会被远端副本替换,随后自动重建本地搜索索引。
恢复操作会替换本机学习数据

它适合新设备,或你明确要让远端仓库成为本机完整数据源的情况。远端仓库及其 Git 历史不会被删除。

Git 常见问题

问题处理方法
403 / 凭据被拒绝使用完整 HTTPS 地址;GitHub 检查 Contents 读写权限,GitLab 检查 write_repository。
发现不同数据集不要合并。为本机数据使用新的空仓库;如果想保留远端数据,则使用“从远端 Git 恢复”。
没有共同祖先两套历史无法安全合并。恢复你想要的远端数据,或把本机数据连接到新的干净仓库。
同步看起来没有动作打开 Git 同步查看“最近活动”,再点击“立即同步”。自动同步通常静默运行,只有错误需要处理时才提示。