English | 中文
Lumina is an information management workspace that combines a web app, FastAPI backend, and browser extension to help you capture web content, run AI insights, and manage reading efficiently.
- Browser capture workflow: one-click full-page or selection capture via popup/context menu, powered by Defuddle, with recent capture history and error logs.
- Structured article library: title search plus category / author / source / time filters, optional topic filter when enabled, and batch category, hide, and delete actions.
- Deep reading experience: detail page supports original/translated views, immersive mode, TOC, article notes, highlight annotations, topic chips, and rendering for code, math, and common media embeds.
- AI insight pipeline: generate summaries, key points, quotes, outlines, translations, auto-classification, and similar-article recommendations with monitorable background tasks that support continuation, repair rounds, cancellation, retry, and chain timelines.
- Columns workspace: publish curated columns, insert article / topic / quote-style references in the editor, and organize longer-form reading paths on top of your library.
- Optional topic knowledge: enable Topic settings, install Lumina CLI locally, connect Bridge to a knowledge project (default provider: llm_wiki), then sync entity/concept topics back to articles and topic detail pages.
- Comments and collaboration: article comments/replies, admin moderation (hide/delete), GitHub/Google OAuth sign-in, and sensitive-word filtering for public discussions.
- Admin control center: configure site basics and home copy, model APIs (general/vector), prompts, recommendation strategy, categories, columns-related settings, comments, optional topic/Bridge settings, and storage.
- Operations and observability: monitor AI task timelines, usage metrics (calls/tokens/cost), header notification center for failed tasks/API errors, and key backend health signals.
- Content lifecycle management: local media storage/compression/cleanup, detail-page Markdown export, public RSS feeds, plus background backup generation/download and strict incremental import for migration and recovery.
- Localized UI and access model: built-in Chinese/English UI, light/dark themes, guest browsing, and admin-authenticated management flows.
flowchart LR
A["Capture article in extension"] --> B["Backend stores content"]
B --> C["Create AI tasks"]
C --> D["Worker runs AI analysis"]
D --> E["Read and manage in web app"]
E --> F["Export / columns / RSS"]
B -.-> G["Optional: local CLI + Bridge"]
G -.-> H["Knowledge compile entities/concepts"]
H -.-> I["Write topics back to Lumina"]
I -.-> E
Android phone RSS reader can use readrops-lumina, download, to quickly collect content to Lumina.
- Normal mode (default): shows original content, full-text annotations, highlights, TOC, AI insights, topics, and recommendations.
- Immersive mode: wide reading mode, hides navigation for focused reading.
- Monitoring module: model usage/billing, AI tasks, and comment monitoring.
- Settings module: basic, categories, AI, comments, storage, and optional topic parsing settings.
- One-click full-page capture: capture via extension button or page context menu without selecting text.
- One-click selection capture: select text first, then capture via extension button or context menu.
- Theme switch: supports light/dark mode.
- Recent reading history: keeps the latest 5 articles for quick jump.
- Article comments: supports commenting on article detail pages for collaboration and feedback.
- Content export: export article title, cover image, and summary by category, and download the current article or column detail page as Markdown.
- Columns workspace: published column pages, template-assisted drafting, article/topic/content reference insertion, and column comments.
- RSS subscription: public RSS feeds for articles and columns; article feeds support category filters.
- Notification center: view AI task-chain failures and API error notifications in the page header.
- Backup operations: storage settings can start a “latest backup” background job, poll generation status, and download the archive when ready.
More features are evolving...
docker compose up -d
./scripts/docker_healthcheck.shURLs:
- Web: http://localhost:3000
- API: http://localhost:8000/backend
- API docs: http://localhost:8000/docs
- The Docker Compose file includes an API healthcheck for
/backend/. ./scripts/docker_watchdog.shcan run via cron or systemd and restart the API when the probe fails.- In production, prefer nginx serving
/backend/media/directly instead of proxying media through FastAPI. Seedeploy/nginx/lumina.conf.example.
# Frontend
cd frontend
npm install
npm run dev
# Backend
cd backend
uv sync
uv run uvicorn main:app --reload
# Extension
cd extension
npm install
npm run devBackend startup validation requires INTERNAL_API_TOKEN. Set it in env or Docker config.
On first use, open /login to set the admin password, then sign in normally.
Backend routes are only served under /backend/api/* (bare /api/* is not available).
Check API_BASE_URL. Same-origin setups usually use /backend; split local ports may use http://localhost:8000/backend.
Check the extension API base URL and confirm the browser can reach the backend.
Topic parsing is optional and off by default. Enable it in admin Topic parsing settings, install/run local CLI + Bridge + knowledge compiler, then sync. Public topic pages only show compiled entities and concepts for now.
MIT License