尚未偵測到更新。GitHub 發布版本後會顯示在這裡。
個人 GitHub 帳號下公開儲存庫的擁有者可以登入並驗證所有權,自行管理介紹,無需管理員核准。
登入並產生驗證檔案,在 30 分鐘內將其提交至此儲存庫的預設分支。其他分支或分叉儲存庫中的檔案無效。然後在下方檢查檔案,即可無需管理員核准直接管理專案。請保留檔案以維持管理權限。
如需登入,將前往登入頁面。登入後請回到此處,再次點擊所選的驗證按鈕。
在專案資料夾中執行 /nomorevibe verify,即可驗證所有權並自行管理介紹。
一分鐘完成驗證AI tools for researchers: summaries and a chat that answer from the full text of papers (arXiv, bioRxiv, medRxiv, journals), collections with sourced comparison tables, daily digest, saved searches, rising topics, weekly letters, scoop watch, citation alerts, press kits, literature reviews and journal club slides.
翻譯準備中
取自儲存庫 README
正在載入完整 README。
檢視完整 README ↗AI tools for researchers, from reading a paper to presenting it. Summaries and a chat that answer from the full text of a paper, collections compared cell by cell, alerts that follow a field, and tools to write, submit and present research. Papers from arXiv, bioRxiv, medRxiv, journals and other preprint servers, in English and French.
Live at www.summarizepaper.com · All tools · Guides
Every answer rests on the paper itself: the full text is read when its license allows it (CC BY, BY-SA, BY-NC, BY-NC-SA, CC0), the abstract otherwise, and claims are checked against the passages they come from, with a link to each passage.
Free tools are open to every account; :star: marks Premium and :star::star: Premium+.
| Tool | What it does | Guide |
|---|---|---|
| Summaries | Essentials, detailed summary, key points, lay summary and blog article of any paper (arXiv ID, DOI, bioRxiv, medRxiv, HAL, OSF, PubMed, Europe PMC link), with its figures; rewritten for another reader or in 16 languages | guide |
| Chat with papers | Questions to one paper or several at once, answered from the full text with the passage behind each claim, figures read too; comparison tables with a source per cell; conversations kept and shared | guide |
| Collections | Papers kept together with notes and tags, edited with others, chatted with as a whole; preprints matched with their published versions, an email when a paper is published or retracted; :star: comparison table and open questions | guide |
| Tool | What it does | Guide |
|---|---|---|
| Daily digest | New arXiv, bioRxiv, medRxiv and journal papers of your fields every morning, ranked by your research profile and votes, with a line for researchers and a web page per email | guide |
| Saved searches | Every new paper of the daily announcements matching a search (plurals, "quotes", -word, OR, prefix*), ranked, by email, in the digest or by RSS. One watch free, five with Premium | guide |
| Rising topics | Topics rising each week, field by field, from several teams, with their history; "Is my topic rising?" for your own words | guide |
| Weekly letters | An article per field each week on the papers that matter beyond it, every number checked against the abstracts; web, RSS, email | guide |
| Scoop watch :star: | Describe a project in progress: prior art, competitors, and an email the day a new paper comes close to it | guide |
| Citation alerts :star: |
| Tool | What it does | Guide |
|---|---|---|
| Press kit :star::star: | By role: a draft release and the questions journalists will ask (authors), strengths and weaknesses (commentators), angles and wording (journalists), every quote checked against the paper, outside commentators suggested, PDF and Markdown | guide |
| Literature review :star::star: | From a research question: 15 to 30 founding and recent papers grouped by approach, with a referenced synthesis | |
| What did I miss? :star: | Papers close to your draft that it does not cite | |
| Bibliography check :star: | Retracted papers, preprints published since, outdated versions, citations the cited abstract does not support (.bib and .tex) | |
| Zotero, Notion, Obsidian | BibTeX, RIS and Markdown exports; papers sent to Zotero with their summary |
| Tool | What it does |
|---|---|
| Journal club :star::star: | Slides with the paper's open-licensed figures, strengths, limitations and discussion questions; edited in the browser, exported to PDF or PowerPoint |
| Conference companion :star::star: | The talks of a conference program that matter for your research, and the people to meet |
Literature reviews, bibliography checks, uncited-paper searches, topic maps and conference programs can be shared with a supervisor or co-authors through a read-only link (90 days, stopped by the owner).
Django 5.2 served by Daphne (ASGI), Channels websockets over Redis, PostgreSQL, on Heroku. Python 3.12. OpenAI models for text, embeddings and speech; paper metadata and search from arXiv, OpenAlex, Crossref, Semantic Scholar, NASA ADS, OpenReview, ORCID, bioRxiv and medRxiv. Stripe for Premium and Premium+.
| Area | Modules (summarizer/) |
|---|---|
| Pages, search | views.py, homepage.py, unified_search.py (arXiv and journals at once), search_filters.py, sitemaps.py, guides.py, tool_list.py (the tools page and menu) |
| Paper text and sources | paper_source.py (arXiv HTML, then PDF), sources.py (DOI, bioRxiv, medRxiv, OSF, HAL, Europe PMC, OpenAlex), publication_status.py, bibliography.py |
| Summaries | consumers.py (websocket), assistant.py, keypoints.py, summary_text.py, reading.py (other reader or language), evidence.py (claims found in the text), generated.py (background results polled by the page) |
| Chat | chat_page.py, chat_table.py (comparison tables), chat_figures.py, chat_vision.py, comparison.py |
| Collections | library.py, collection_tools.py, questions.py, publication_alerts.py, exports.py (BibTeX, RIS, Markdown, Zotero) |
| Following a field | digest.py, digest_web.py, digest_bundle.py, saved_searches.py, trends.py, trends_web.py, trend_sources.py, letters.py, audio.py (read aloud), synthesis.py |
| Research tools | scoops.py, citations.py, changes.py (versions), teams.py, students.py, student_papers.py, student_sources.py, missed.py, refcheck.py, literature.py, presskit.py, presskit_pdf.py, slides.py, slides_pptx.py, conference.py, result_share.py |
| Accounts and billing | accounts.py (email links, Google sign-in, bot checks), (quotas, never trusts the browser), (Stripe), , |
python3.12 -m venv venv
venv/bin/pip install -r requirements.txt
export SECRET_KEY=dev OPENAI_KEY=sk-...
venv/bin/python manage.py migrate
venv/bin/python manage.py runserver
Websockets need Redis on 127.0.0.1:6379 (brew services start redis).
Tests: venv/bin/python manage.py test summarizer and node --test tests/*.test.cjs.
Translations: French strings live in locale/fr/LC_MESSAGES/django.po; compile them with
python manage.py compilemessages. Guides are Markdown in content/guides/{en,fr}/, copied into the
database with python manage.py import_guides.
| Variable | Purpose |
|---|---|
SECRET_KEY, ON_HEROKU, SITE_URL | Django secret key, production settings, public address |
DATABASE_URL, REDISCLOUD_URL | Set by the Postgres and Redis Cloud add-ons |
OPENAI_KEY | OpenAI API key |
SECRETCMTA | SMTP password (outgoing email) |
GOOGLE_CLIENT_ID, GOOGLE_CLIENT_SECRET | Google sign-in |
OPENALEX_API_KEY | Optional, free: raises the OpenAlex daily budget and reads full texts OpenAlex keeps |
ADS_TOKEN, S2_API_KEY | NASA ADS and Semantic Scholar, sources of the citation alerts |
SCOOP_DRAFT_KEY | Encrypts the scoop watch's kept drafts |
STRIPE_SECRET_KEY, STRIPE_WEBHOOK_SECRET, STRIPE_PRICE_* | Premium and Premium+ (monthly, yearly, EUR and USD); closed while unset |
UMAMI_WEBSITE_ID, SENTRY_DSN | Optional analytics and error reporting |
Umami events (static/summarizer/js/funnel.js) carry only allowlisted language, period, placement and tool
identifiers, never queries, manuscript text or email addresses. Stripe webhooks and UserPlan remain the
source of truth for subscriptions.
git push heroku master; the release phase of the Procfile applies migrations.
One hourly Heroku Scheduler job, python manage.py send_digest, sends the digests that are due, then
counts the day's terms for the rising topics, writes the weekly letters and sends scoop, citation,
version, publication, saved-search and topic-map alerts and research placement reminders, within what is
left of the hourly email budget.
Other commands:
| Command | Purpose |
|---|---|
letters_write | Write the weekly letters of a week now, or print what they would hold |
trends_backfill, trends_calibrate | Count past weeks for the rising topics; print rising terms with their statistics |
scoop_calibrate, citations_calibrate, teams_calibrate, students_calibrate | Run a tool on sample inputs and print what it counts and why, to check thresholds (they write nothing) |
digest_stats | Click-through rate by rank, votes and unsubscriptions of the digests |
embed_papers | Embed summarized papers that have none; --rescore recomputes closest papers |
import_guides | Copy the Markdown guides into the database |
send_reactivation, suppress_emails, announce_tools | Account activation reminders, bounced addresses, announcement of new tools |
Stripe webhook: https://www.summarizepaper.com/stripe/webhook/ with checkout.session.completed and
customer.subscription.created, .updated and .deleted.
Developed by :sparkles: Quentin Kral :sparkles:. Questions and suggestions are welcome.
MIT License, see the LICENSE file.
| Who cites your papers and what they do with them, citing sentence included, self-citations apart |
| guide |
| Who works on it? :star: | People and institutions signing recent papers on a topic (first, last, corresponding author), their approaches, referees without your co-authors, new names monthly | guide |
| Research placements :star::star: | For students: who signs recent papers on a topic, their team and funding, readings, a first email and reminders | guide |
access.pybilling.pypremium_tools.pyreactivation.py| Operations | usage.py (AI tokens and cost by task, model, day and user), activity.py (daily staff page), loadshed.py (503 past 14 requests in flight), threadwatch.py, cache.py, canonical_host.py, mailcheck.py |