User's guide
Run a literature review, end to end.
CERULEON takes you from a PubMed query to an audit-ready literature search report — search, screen, classify, synthesise, and export. This guide walks each step using the app exactly as it ships. Everything runs in your browser; sign in to sync your work to the cloud.
Getting started
Open CERULEON and the Welcome to CERULEON dialog appears. You have four ways in:
- Sign in with your email and password, or request a one-time magic link by email.
- Create account — email and a password of at least 8 characters.
- Continue with Google — one-click sign-in.
- Continue without an account — guest mode; your data stays in this browser only.
There is no consent checkbox to tick — by continuing you accept CERULEON's Terms of Service and Privacy Policy, shown beneath the buttons.
Interface overview
The workspace has three regions: a left sidebar, a top toolbar, and the main area showing the active view.
Sidebar
- The CERULEON wordmark returns to the Welcome screen.
- Projects opens the project catalog; below it sits your project list and the New project button.
- The workflow nav — Dashboard, Search, Screening, Table, Similar devices, Synthesis, Bibliography, Obsidian, Export, Audit trail — appears once a project is selected. (Similar devices shows only in SOTA projects. No project selected? The workflow nav is hidden and you land on Welcome.)
- The sidebar can be collapsed to icons.
Toolbar
- A mode badge (SOTA or Device-specific), the project name, and the current view label.
- On the right: the AI configuration bulb (glows while AI is working), theme settings, the notebook, and your account menu — which holds your profile, subscription, Cloud sync, your Zotero connection (see Zotero integration), Support, and sign-in/out.
Creating a project
Click New project, give it a name, and choose a mode. The mode is set at creation and shapes the whole workflow — the table columns, the AI synthesis structure, and which regulatory fields appear.
| Mode | For | What it changes |
|---|---|---|
| SOTA (default) |
State of the art — broad clinical context, guidelines, systematic reviews / meta-analyses. | State-of-the-art evidence table and synthesis. Filter presets favour guidelines and reviews. |
| Device-specific | Device clinical evaluation. | Unlocks PICOTS scoring and GSPR mapping (Annex I MDR, sections 1–9) in the Table and screening panel, a device-oriented synthesis, and device-classification fields (type, risk class). |
You can switch a project's mode later from Edit project, but CERULEON asks you to confirm — the table format and AI structure change, and PICOTS/GSPR may need re-running.
Dashboard
The Dashboard is your project at a glance: the evidence pyramid (Oxford CEBM levels) and the spread of GRADE confidence across your included pool, plus pool counts. Drill into a tier to see the studies behind it.
Search
Build your pool two ways — query PubMed, or import references you already have.
PubMed query
- Enter your concepts; CERULEON expands each to
(term[Title/Abstract] OR term[MeSH Terms])and joins them with AND. The query is split into an editable semantic axis and auto-generated filter clauses. - As you type, the query is colour-coded — the
AND,ORandNOToperators, field tags and parentheses each get their own colour, so the logic is easy to read at a glance. - Pick the SOTA or Device-specific filter card for publication-type and date presets (SOTA defaults to the last 5 years), or set custom filters.
- Every search is saved to the project's query history — kept in full as part of the MDR/IVDR audit trail — with its date and result count.
Import
- Drag a file onto the drop zone, or browse for it. Supported: RIS (.ris), NBIB (.nbib, PubMed/MEDLINE export) and BibTeX (.bib). The format is detected automatically. Import is file-based.
- Tag the import with its source (Cochrane, ClinicalTrials.gov, Europe PMC, Embase, Web of Science, Scopus, CINAHL, or a custom label) for the audit trail.
- Import from Zotero. Once you've connected a Zotero account, an Import from Zotero Library button appears on this card — pull a whole collection (or your entire library) straight in, PDFs and all. See Zotero integration.
Records are de-duplicated by PMID across your pool; imports also check the DOI against what's already there. Load the results into Screening when you're ready.
Screening
Work through your records and decide Include, Exclude, or send to full-text review. Every manual decision is logged.
Auto Screen with AI
With AI enabled, Auto Screen with AI reads each pending abstract against your inclusion / exclusion criteria and returns a decision (include / exclude / full-text), a confidence level, and a rationale quoting the abstract — with per-criterion verdicts. You can override any AI decision; the override is timestamped.
The article editor
Open a record to edit it in the side panel:
- Study design — a free-text field.
- Evidence level — Oxford CEBM: Ia, Ib, IIa, IIb, III, IV.
- GRADE confidence — High, Moderate, Low, Very low.
- In Device-specific projects, PICOTS and GSPR sections appear here too.
GRADE Evaluation (AI) suggests a GRADE confidence and an Oxford level for the record, which you can accept or change.
Full-text PDFs & the reading view
Attach the full-text PDF to any record — drop files onto the batch-PDF tool, or, with a Zotero account connected, fetch missing PDFs for your included articles straight from your library (see Zotero integration). With AI on, CERULEON can then analyse the full text.
When AI analyses an article's PDF it produces a reading view — a concise AI digest with an overview and any reproduced results tables. Records that have one show a water-drop button (in Screening and in the Table); click it to open the reading view in a side panel.
Table
The Table view is your evidence table — every included article in one comparable grid, annotations and all. Columns follow the project mode: a state-of-the-art layout for SOTA projects, or PICOTS scoring with GSPR mapping for Device-specific ones. Toggle Show quotes to surface the supporting source quotes. This table is what feeds the Evidence-table export.
Rows for articles you've analysed show a water-drop button — open it for the reading view (an AI overview and reproduced results tables drawn from the PDF).
Similar devices
In SOTA projects, the Similar devices view lists devices similar to the one under investigation, extracted from your included articles by AI — it scans abstracts (and PDFs when available) to build the table.
- Edit any row, or add a device manually.
- Enrich with web research to fill in missing manufacturer, class, mechanism or IFU details. Web research is part of the higher plans; where it isn't available, the view tells you.
Synthesis
Generate a structured synthesis of your included pool with AI. The sections follow the project mode:
- SOTA — summary, discussion, limitations & gaps, conclusion.
- Device-specific — summary, discussion, safety & performance, benefit / risk, GSPR coverage, conclusion.
Edit any section inline, or regenerate a single section without touching the rest. With a pool of three or more included articles, CERULEON discovers themes before writing. With AI off, it falls back to a non-AI narrative built from your annotations.
Bibliography
A numbered bibliography of your included articles. Switch the citation style — Vancouver, APA, or Harvard — and every reference re-formats at once. The selected style carries through to the bibliography export.
Obsidian vault
The Obsidian vault turns your included articles into an interlinked knowledge base — the same idea as the Obsidian note-taking app, but built for you automatically and running right in the browser (nothing to install). It's assembled from work you've already done: your articles, annotations and synthesis themes.
Notes
Every included article becomes a Markdown note — YAML frontmatter, an extraction table, verbatim supporting quotes, [[wikilinks]] and #theme tags. Alongside the article notes you get a note per synthesis theme, a note per synthesis chapter, and an index note that maps the whole vault.
Graph
Toggle Graph to see your vault as a live, force-directed graph — a global graph of every note, plus a local graph around the note you're reading. Nodes are draggable and sized by how connected they are; hovering lights up the neighbourhood; the layout animates into place and then settles. (It honours your reduced-motion setting.)
Links, backlinks & citations
- An article-to-article edge appears when you write a
[[wikilink]]between two notes — otherwise links flow through the shared theme and chapter notes. - Every note lists its backlinks (the notes that point to it) and an outline of its headings.
- The References button fetches the real bibliographic references between your included articles from PubMed and draws them as directed citation edges: when article A cites article B and both are in your pool, A's note gains a References in corpus section. It's fail-soft — an offline lookup just leaves the graph as it was.
Finding & writing
- Search notes… filters as you type; type
#for a ranked list of the vault's tags, and the one you pick becomes a removable filter chip. - Add your own notes with +, and edit any note block by block. Type
@to insert a[[wikilink]]to an included article, or#to complete a tag from those already in the vault. - Once you edit a generated note it stops auto-updating from the data (it shows an edited badge); Regenerate rebuilds it from scratch, and notes you created yourself can be deleted.
- The Notebook's Push to Obsidian button drops the current notebook note into the active project's vault, while keeping its copy in the notebook.
To take it outside the app, export the whole vault as a ZIP of .md files (Export → Obsidian vault) and unzip it into a real Obsidian vault — the wikilinks, tags and structure all carry over.
Export
Export reflects your current screening pool. Each deliverable is its own card with its own formats:
| Deliverable | Formats |
|---|---|
| Database queries | PDF, XLSX |
| PRISMA flow | PDF, XLSX |
| Evidence table — included articles | PDF, XLSX |
| Excluded articles | PDF, XLSX, RIS |
| Narrative synthesis | PDF, TXT |
| Bibliography | PDF, TXT, RIS |
| Full report | PDF (all sections combined, audit-ready) |
| LaTeX source | ZIP (.tex + .bib — compile on Overleaf) |
| Obsidian vault | ZIP (interlinked .md notes — see Obsidian vault) |
The full report bundles cover, introduction, PRISMA, evidence levels, exclusion criteria, search keywords, evidence table, excluded articles, discussion and bibliography — designed to annex into your Clinical Evaluation Report.
With a Zotero account connected, the Bibliography card also shows → import to Zotero, which pushes your included articles straight to your Zotero library. See Zotero integration.
Zotero integration
CERULEON connects to your Zotero library so references and PDFs can move both ways. The connection is tied to your account, so you'll need to be signed in.
Connect
Open your account menu → Zotero → Connect Zotero account, then paste a Zotero API key. Create one at zotero.org → Settings → Keys (“Create new private key”) with library and write access; CERULEON works out your numeric user id from the key. Once connected, the row reads Connected and the Zotero actions light up across the app. Manage or disconnect from the same place at any time — disconnecting makes CERULEON forget the key, and your Zotero library itself is left untouched.
What you can do once connected
| Action | Where | What it does |
|---|---|---|
| Import from Zotero | Search → Import from Zotero Library | Pull a collection (or your whole library) into the project as an import group — the same path as an RIS import. Optionally download the items' PDFs into CERULEON (your storage quota and the 25 MB per-file cap still apply). Activate the import in the search list to load it into Screening. |
| Fetch missing PDFs | The batch-PDF tool → fetch missing PDFs | For included articles that don't have a PDF yet, look each one up in your Zotero library and attach any PDF found. It enriches your existing records only — never creating new articles or duplicates. |
| Push to Zotero | Export → Bibliography → → import to Zotero | Send your included articles to your Zotero library. It's dedup-safe — items already there are matched and filled in, never duplicated. Optionally attach each article's PDF, and file everything into a collection named after your project. |
Audit trail
Every meaningful action — search runs, screening decisions and overrides, AI runs, and exports — is recorded with a timestamp. The audit trail gives you the regulatory traceability (MDR / IVDR) behind your deliverables, so a reviewer can reconstruct exactly how the review was built.
Settings
The Settings dialog holds a single preference: Theme — Light, Dark, or Auto (Auto follows your operating system).
AI is configured separately, from the AI configuration bulb in the toolbar:
- AI features — a master on/off switch. With it on, CERULEON uses AI for screening, extraction, GRADE and synthesis; turn it off to work entirely by hand or to hold quota.
- Model for deep tasks — choose Balanced Mode or Heavy Mode on Pro+ and Pro Max plans. Quick passes (auto-screening, GRADE) always run in Balanced Mode for speed; the model choice applies to synthesis and full-text analysis.
- Your plan and monthly AI usage are shown here too.
Tips & best practices
- Name and re-run your searches. The full query history is your audit trail — keep it meaningful.
- Use Device-specific mode for a clinical evaluation. It unlocks PICOTS and GSPR mapping you'll need for the CER annex.
- Always read the AI rationale. Auto-screening quotes the abstract for every decision — verify it in seconds, and override anything you don't trust.
- Classify as you screen. Set Study design, Evidence level and GRADE in the article panel so the Dashboard pyramid and exports stay accurate.
- Sign in to protect your work. Guest data stays in one browser; signing in syncs projects, articles, annotations and PDFs.
- Export early, export often. Deliverables always reflect the current pool — regenerate after each screening pass.
- Connect Zotero to move both ways. Pull collections in, fetch missing PDFs for your included set, and push your final bibliography back — all dedup-safe.
- See your review as a graph. The Obsidian vault turns your included pool into interlinked notes and a live graph; hit References to draw the real citation links between them.
FAQ
Do I need an account?
No — you can continue without an account and work fully offline-in-browser. But guest data stays in that one browser and is not backed up; sign in to sync across devices.
Is my data synced to the cloud?
When you're signed in, your project data — projects, articles, annotations and uploaded PDFs — syncs automatically. App preferences (theme, AI on/off, model) stay local to the browser.
Which formats can I export?
PDF, XLSX, TXT, RIS, and a LaTeX ZIP (.tex + .bib) for Overleaf. The full report is PDF.
What's the difference between SOTA and Device-specific projects?
SOTA targets the broad state of the art (guidelines, reviews). Device-specific adds PICOTS scoring and GSPR mapping (Annex I MDR §1–9) and a device-oriented synthesis for a clinical evaluation. The mode is set at creation.
Do I have to use AI?
No. Turn AI features off and run the whole workflow by hand — search, screening, classification, and a non-AI narrative all still work. AI assistance is part of the paid plans.
Which AI model runs my tasks?
Balanced Mode by default. On Pro+ and Pro Max you can pick Heavy Mode for deep tasks (synthesis, full-text analysis). Quick passes like auto-screening and GRADE always run in Balanced Mode for speed.
How does evidence grading work?
In the article editor you set an Oxford CEBM Evidence level (Ia–IV) and a GRADE confidence (High–Very low). The AI GRADE Evaluation can suggest both, which you confirm or change.
Can I import references I already have?
Yes — drop in a RIS, NBIB, or BibTeX file. CERULEON detects the format and de-duplicates by PMID and DOI against your existing pool.
Can I connect Zotero?
Yes. Sign in, then account menu → Zotero → Connect Zotero account and paste a Zotero API key. You can then import collections, fetch missing PDFs for your included articles, and push your bibliography back to Zotero (dedup-safe). Your key is stored encrypted server-side and never kept in the browser. See Zotero integration.
What is the Obsidian vault?
An in-app, interlinked knowledge base built automatically from your included articles — Markdown notes with wikilinks, #theme tags, backlinks and a live graph. No install needed; you can also export it as a ZIP of .md files and open it in the real Obsidian app. See Obsidian vault.