Add explicit license and dataset usage boundaries

[botaruibo]

Why this matters

MyVoiceTyping is positioned as an open-source, local-first Chinese voice typing project, and the promotion plan now relies heavily on three public assets:

• App: https://github.com/botaruibo/MyVoiceTyping
• Model: https://modelscope.cn/models/botaruibo/MyVoiceTyping-1.5b-q4
• Dataset: https://github.com/botaruibo/MyVoiceTyping-Dataset

However, the current GitHub API shows:

• App repo: license: null
• Dataset repo: license: null
• ModelScope model page: License: apache-2.0

For users, developers, newsletters, awesome lists, and teams evaluating the project, this is a trust and adoption blocker. People need to know whether they can fork, modify, redistribute, train with, or commercially evaluate the project and related assets.

Goals

• Add a clear project-level license to the app repository.
• Clarify that the app code, model artifacts, and dataset may have different license / usage boundaries.
• Add dataset source and usage notes so the training data can be evaluated responsibly.
• Make the README and landing page safer to share in open-source communities.

Suggested tasks

App repository

• Choose a project license: MIT, Apache-2.0, GPL-3.0, or AGPL-3.0.
• Add LICENSE to the repository root.
• Add a README license section / badge.
• Add a short license note to the landing page.

For early community growth and easier adoption, MIT or Apache-2.0 may be the most practical choices. Final choice should be made by the maintainer.

Dataset repository

• Add DATA_LICENSE.md or equivalent usage notes.
• Add docs/SOURCES.md listing each upstream dataset source, original link, license, citation requirements, commercial-use status, redistribution status, and sample count.
• Clarify that samples remain subject to upstream dataset licenses unless explicitly stated otherwise.
• Add README wording that users must verify upstream licenses before redistribution, publication, commercial training, or model release.

Model page / docs

• Keep the ModelScope Apache-2.0 signal visible.
• Mention that the model is based on Qwen/Qwen2.5-1.5B-Instruct and users must also comply with upstream model terms.
• Link to the dataset usage notes once added.

Suggested README wording

## License / Usage boundaries

The app code, model artifacts, and dataset may have different license or usage terms.

- App code: TODO, pending maintainer decision.
- Model: the ModelScope page currently marks the model as Apache-2.0; users should also comply with the upstream Qwen model terms.
- Dataset: built from multiple public upstream datasets. Before redistribution, publication, commercial training, or model release, please check each upstream dataset's license, citation requirements, and usage restrictions.

Related internal promotion notes

This issue tracks a P0 trust signal for the 0 -> 1000 stars growth plan. Clear licensing will help with:

• developer confidence;
• awesome-list / newsletter submissions;
• company/team evaluation;
• reuse and contribution decisions;
• accurate community promotion language.

[botaruibo]

Progress update:

The Dataset-side usage boundary docs have been added:

• Dataset commit: botaruibo/MyVoiceTyping-Dataset@a7b3aff
• DATA_LICENSE: https://github.com/botaruibo/MyVoiceTyping-Dataset/blob/main/DATA_LICENSE.md
• Source verification table: https://github.com/botaruibo/MyVoiceTyping-Dataset/blob/main/docs/SOURCES.md

What changed:

• The Dataset README now links to DATA_LICENSE.md and docs/SOURCES.md.
• DATA_LICENSE.md clarifies that the dataset is derived from multiple public upstream datasets and that upstream licenses / citation / redistribution terms still apply.
• docs/SOURCES.md lists the current sources and marks unknown license fields as needs_verification instead of treating them as allowed.

Remaining work:

• choose the App repository license;
• add App LICENSE and README badge / section;
• verify each upstream dataset source in docs/SOURCES.md;
• update README / landing wording once the App license and Dataset source table are more complete.

[botaruibo]

补充一个最近的隐私 / 安全边界更新。

MyVoiceTyping 一直强调“本地优先”，但我不想把它做成一句模糊的营销话术，所以这几天把几个具体边界补进了文档和 landing：

• 新增 Security Policy；
• landing 顶部和反馈区都增加了 Security 入口；
• Privacy / Data Safety 里补充了日志与调试信息边界；
• Contributing 里补了 AI Coding 30 秒测试、self-evaluation 样例边界和安全 / 隐私问题报告路径。

现在文档里明确写了几件事：

• 普通日志不应该记录原始音频、原始转写文本、润写后文本或用户最终确认文本；
• self-evaluation / 自进化样例比普通日志更敏感，应默认本地优先、用户可控、可删除、可预览；
• 如果要反馈问题，请不要在公开 issue 里贴公司机密、私人消息、token、密码、内部路径或未脱敏日志；
• 语音输入结果进入 agent / 聊天框前，应尽量保留 review-before-submit 的边界。

我更希望 MyVoiceTyping 的“本地优先”是可审查、可讨论、可反馈的，而不是一句“100% 隐私”的夸张承诺。

如果你在公司场景、AI Coding prompt、隐私敏感输入里试用，欢迎优先指出隐私边界、日志、调试信息、self-evaluation 数据流哪里还不清楚。

[botaruibo]

Progress update on the Dataset side:

The Dataset repo now has two additional conversion / trust updates:

• Dataset README now explains its relationship with the main MyVoiceTyping app and links back to:
  • App repo: https://github.com/botaruibo/MyVoiceTyping
  • ModelScope model: https://modelscope.cn/models/botaruibo/MyVoiceTyping-1.5b-q4
  • 30-second trial tasks: https://github.com/botaruibo/MyVoiceTyping/blob/main/docs/TRIAL_TASKS.zh-CN.md
  • Chinese alternatives comparison: https://github.com/botaruibo/MyVoiceTyping/blob/main/docs/ALTERNATIVES.zh-CN.md
• Dataset sample collection issue is now open:
  • 征集中文语音输入 / ASR 后处理样例：错词、标点、中英混合和技术词保护 MyVoiceTyping-Dataset#1

This means the Dataset repo now has three clearer public paths:

1. usage / license boundary: DATA_LICENSE.md and docs/SOURCES.md;
2. app referral path: Dataset README -> main App repo / trial tasks / model page;
3. feedback path: Dataset issue MyVoiceTyping：面向中文语音输入的本地优先开源方案 #1 for sanitized ASR post-processing samples.

Remaining blocker for broader newsletter / awesome-list / directory submissions:

• App repo license is still undecided (license: null via GitHub API).
• Once the maintainer chooses MIT / Apache-2.0 / another license, the next step is to add root LICENSE, README license section/badge, landing note, then re-submit or follow up on the existing community submissions.

[botaruibo]

Progress update: added an App license decision guide.

New guide:

• https://github.com/botaruibo/MyVoiceTyping/blob/main/docs/LICENSE_DECISION.zh-CN.md

What it clarifies:

• App code, the local text-polishing model, and the dataset should be treated as separate assets with separate license / usage boundaries.
• For the current growth goal — easier Star, fork, PR, weekly digest inclusion, awesome-list inclusion, and product-directory review — the most practical default choices are MIT or Apache-2.0.
• MIT is the simplest low-friction option for adoption.
• Apache-2.0 is also a strong option if we want clearer patent / contributor boundary language.
• GPL-3.0 / AGPL-3.0 remain possible, but they would likely add more friction for early adoption and directory / enterprise evaluation.
• Keeping the App repo as No license is still the biggest blocker for community submissions and reuse.

Current recommendation for the 0 → 1000 stars plan:

1. Prefer MIT if the priority is maximum adoption and low-friction community growth.
2. Use Apache-2.0 if explicit patent language is more important.
3. After the maintainer confirms the final choice, add the root LICENSE, update README / landing / release notes / press kit / llms.txt, then follow up with HelloGitHub, ruanyf/weekly, GitHubDaily, and relevant directories.

This comment does not finalize the license. Final choice is still pending maintainer confirmation.

[botaruibo]

Update: the App repository now has an explicit project-level license: MIT.

This completes the App-side license clarification.

• License commit: aa34dd0
• App license: https://github.com/botaruibo/MyVoiceTyping/blob/main/LICENSE
• README now includes the MIT badge and App / Model / Dataset usage boundary notes.
• Model page: https://modelscope.cn/models/botaruibo/MyVoiceTyping-1.5b-q4
• Dataset usage boundary: https://github.com/botaruibo/MyVoiceTyping-Dataset/blob/main/DATA_LICENSE.md
• Dataset sources: https://github.com/botaruibo/MyVoiceTyping-Dataset/blob/main/docs/SOURCES.md

GitHub API now reports license.key = mit for the App repo.

Remaining related work: keep verifying upstream licenses for dataset sources, and continue making the model / dataset usage boundaries visible in submissions and docs.