README file from
GithubResearch-based Learning Community
An Obsidian plugin for research-based learning: turn literature into a traceable reading chain — domain → AI literature review → verified sources → skimming → deep-read cards → optional publishing.
中文简介:面向研究性学习的 Obsidian 插件。把「领域 → 文献综述 → 核验文献 → 略读 → 精读卡片」 做成一条可追溯的链路,并支持把学习成果改写成面向特定读者的科普图文。
Features
- Domain-level literature review — AI drafts a structured review; candidates carry DOIs and are verified with Crossref before they can be added.
- Literature workspace — each paper gets its own folder: metadata, skim report, deep-read cards, optional PDF, all named consistently.
- Deep-read card — guided five-part card (quote, experience, Aha, inspiration, meaning) plus concept value and an executable action.
- PDF full-text organizer — OCR (DeepSeek OCR) turns a PDF into Markdown, repairs layout, and adds paragraph-level Chinese translation for English papers.
- Poster generator — writes a short post for a reader profile you define, draws an illustration with seedream, and composes a 1080×1440 poster locally.
- Practice cards — plan, daily-note progress log, step checkboxes and a summary, with a calendar and a today list in the sidebar.
- Optional publishing — push approved cards and posters to a self-hosted community server.
Install
Community plugins
Search for Research-based Learning Community in Settings → Community plugins.
Manual
- Download
main.js,manifest.json,styles.cssfrom the latest release. - Put them in
<vault>/.obsidian/plugins/research-learning-community/. - Enable the plugin in Settings → Community plugins.
Setup
| Setting | Purpose |
|---|---|
| AI endpoint / key / model | Any OpenAI-compatible chat endpoint. Used for reviews, cards, translation, posters. |
| Reader profiles | Name + free-form description; the poster/post generator adapts length and tone to it. |
| Image endpoint / key / model | Volcengine Ark seedream for poster illustrations (optional). |
| Sync server + device token | Optional; only needed if you publish cards to a community server. |
Privacy
- AI calls go directly from your machine to the endpoint you configure; there is no middle server.
- The plugin only uploads:
- the cards you explicitly publish,
- the PDF you explicitly choose to OCR (to your own AI provider),
- the poster you explicitly generate.
- Notes, drafts and unpublished cards never leave your vault.
- API keys are stored in the plugin's local
data.json(plain text, local only).
Data layout
领域/
└─ <domain>/
├─ 文献综述.md
└─ 文献/<paper>/
├─ <paper>.md metadata
├─ <paper>-略读.md skim report
├─ <paper>-dr-xxxxxxx.md deep-read card
├─ <paper>-全文.md organized full text
└─ assets/ figures and posters
Status
Early but usable. The optional server component is not required for local use.
License
MIT
Build from source
pnpm install
pnpm build # typecheck + bundle to main.js
Layout:
src/ plugin source
packages/contract/ card & API schemas
packages/vault-core/ vault layout and rendering
main.js in the repository root is the built bundle, so the plugin can also be installed without building.