One creator IP → a researched article pipeline for five overseas platforms.
确定你的内容 IP，再围绕关键词持续写有证据的海外长文。
Longform Atlas is a local writing workspace for creators publishing on X Articles, Quora Answers, Medium, LinkedIn Articles and Substack. Enter keywords and an audience to suggest a creator positioning, or write your own IP brief. Then build distinct platform-specific article plans around that positioning. Add evidence to each plan; create an offline outline or connect your own OpenAI-compatible model endpoint to draft up to three selected articles at a time. Review and publish manually.
This is a workflow tool, not a ranking guarantee or an auto-posting bot. It is designed for evidence-led, original writing rather than keyword stuffing.
The built-in sample demonstrates the workflow; it is not a claim of published results. See the cross-platform article plan and the editable offline outline and export controls.
- Chinese and English interface; article language is a separate setting / 中英文界面与文章语言分别切换。
- Creator-IP brief: audience, role, positioning, voice, keywords and real experience / IP 定位、受众、表达方式与真实经验。
- Keyword-to-positioning suggestion and a small cross-platform article plan / 关键词建议定位，形成跨平台选题计划。
- Evidence per article; up to three selected drafts per deliberate batch / 每篇单独补证据，一次最多生成三篇初稿。
- Platform-specific prompts for five overseas publishing destinations / 五个平台的不同文章结构。
- Source ledger, first-hand evidence and original-angle fields / 来源台账、亲身经验和独特观点。
- Search phrase, reader intent and canonical-original URL fields / 搜索词、搜索意图与转载原文链接。
- Offline outline, optional full draft through your own model API / 离线大纲或自带模型 API 生成长文。
- Editable Markdown, copy/download and a basic structure check / 编辑、导出和结构检查。
- Local browser storage for the project; API key stays in the current page only / 项目存于本机浏览器，API 密钥不持久化。
- The hosted GitHub Pages demo loads Google Analytics only after consent for basic visit/referral measurement and clicks to LydiaTools GitHub pages. Automatic enhanced measurement (including scroll and generic outbound-click events) is disabled. The analytics code does not send draft, evidence, source-field, or API-key values. The downloadable local app does not load Google Analytics. / GitHub Pages 在线演示页仅在明确同意后加载 Google Analytics，用于基础访问、来源及前往 LydiaTools GitHub 页面的点击统计；滚动和通用出站点击等自动增强测量已关闭。统计代码不发送草稿、证据、来源字段或 API 密钥；本地版不加载 Google Analytics。
- No account connection, auto-publishing, scheduling or fabricated analytics / 不登录平台、不自动发布、不编造流量效果。
Try the browser demo — plan a creator IP, build platform-specific topics, download the plan and export an offline outline. The public demo has no model connection or article-generation button; use the local app for full AI drafts. Share a workflow observation.
Download v0.2.0 ZIP · Release notes
Unzip the download. Requires Python 3.9+; no Python packages or build step. From the extracted longform-atlas-v0.2.0 folder, run:
python3 app.pyOn Windows, run start-windows.bat with Python 3 installed. On macOS, run start-mac.command or the command above.
Open http://127.0.0.1:8765. Click Load a sample IP and brief, then Build article plan. Use Use in editor on a plan and Build offline outline to test without an API key. For a full article, add real evidence for that article, supply your own compatible HTTPS base URL, model and API key, then press Generate article. For multiple articles, select up to three evidence-filled plans and press Draft selected briefs. The server accepts localhost HTTP for a local model. The app sends each brief and key to that chosen provider only when you click generation; it does not save the key.
运行 python3 app.py，打开本机地址；点击「载入示例 IP 与选题」「生成长文选题计划」，选一篇放入编辑器，再「离线生成大纲」即可测试。生成完整初稿时再填写自己的兼容 API 地址、模型和密钥。
The generated text can still contain mistakes. Check all claims and source links, edit the prose in the app, then export Markdown and publish from your own account.
| Destination | Best use in this tool | Current publication detail |
|---|---|---|
| X Articles | Original analysis with a strong hook and sections | Publishing requires eligible X Premium/Premium+, Business or Organization access (X Help). |
| Medium | Evergreen story and republishing | Medium supports canonical links for imported/republished work (Medium Help). |
| Quora Answers | Direct answer to a specific reader question | Answer the question directly, disclose relevant affiliation, and avoid promotional filler (Quora policies). |
| Professional how-to or lesson | Article SEO title and description can be edited (LinkedIn Help). | |
| Substack | Newsletter essay and owned reader relationship | Publishing can start free; paid subscriptions have a platform fee (Substack Help). |
Platform features and eligibility change; verify the current account and platform documentation before publishing. The batch limit and per-article evidence fields favor substantive, distinct work; scaled low-value pages can violate Google's spam policy (Google Search Central). Indexing or ranking is never guaranteed.
app.py runs a loopback-only HTTP server with a per-process local request token. Static assets live in web/. The optional generation request is sent through the local server to the user-entered provider. The server does not log request bodies or keys. IP profiles, plans, evidence fields and drafts are saved in this browser's localStorage; the API key is not.
No platform accounts, cookies or publishing APIs are used. The source ledger does not scrape linked pages; users provide and verify evidence.
python3 -m unittest discover -s tests -v
node --check web/app.jsSee design notes, Chinese quick guide, and LICENSE.