What is Rhinal?
Rhinal is a private memory tool for clinical researchers. You talk to it like a colleague — describe a case, log a decision, capture an observation — and it structures that thought, timestamps it, and writes it to a database you own in Notion.
The longer you use it, the more it accumulates a longitudinal record of your thinking that nothing else can replicate. It runs on models you control — either locally on your machine (completely private) or through an API key you provide (leaves your machine, but is still your key, your account).
Why it exists
Clinical memory is lost
Observations made at 2am, differential diagnoses reconsidered on day 3, follow-up patterns noticed after 6 months — these rarely make it into structured records. They live in heads, WhatsApp threads, and paper.
Existing tools don't compound
Notes apps are flat. LLMs forget. EMRs are structured for billing, not longitudinal clinical thinking. Rhinal is built specifically to let one thought connect to the next, across months.
Privacy is non-negotiable
Clinical data cannot go to a third-party server unless it's been de-identified and you've verified compliance. Rhinal runs locally by default. Nothing is sent anywhere except to the model you chose and to your own Notion.
It should work for anyone
You don't need to be a researcher to benefit from structured thinking over time. Architects, engineers, students — anyone who thinks across long timescales benefits from the same pattern.
Words to know
Setting it up
01
Sign in with Google
What you do Click 'Continue with Google' on the landing page.
Why it matters Your identity is separate from your storage. Google handles authentication; Rhinal never sees your Google password.
What you get A secure session tied to your email. Your records follow you across devices.
02
Choose your model
What you do Pick a local Ollama model (free, private) or a cloud model (faster, needs a key).
Why it matters Different researchers have different hardware and privacy needs. Rhinal routes to whichever engine you chose.
What you get A working inference engine. You can switch at any time — your history isn't tied to one model.
03
Connect your Notion
What you do Click 'Authorize Notion' and grant access to your workspace.
Why it matters Rhinal needs write access to create your memory database. Read access is not requested.
What you get A 'Rhinal Vault' database auto-deployed inside your Notion, with the correct schema for all six fields.
04
Think out loud
What you do Type anything — a case you're working on, a decision you made, an observation. Hit enter.
Why it matters The canvas is where capture happens. There's no form to fill, no template. Just talk.
What you get A structured record in Notion within ~30 seconds. Mode-tagged, timestamped, indexed to your vault.
What happens inside
1
You type
Your text enters the canvas. It can be raw — a case summary, a half-formed observation, a direct question.
2
The model converses
The AI asks one sharp clarifying question if your thought is ambiguous. It never assumes; it asks. When you've answered, the record is ready.
3
Click 'Save to vault'
When you decide the thought is complete, you press Save. The model doesn't decide for you — you do.
4
Distillation
A second (silent) model call extracts concept, summary, domain tags, and source quotes. Mode is taken directly from your selection — the model cannot override it.
5
Written to Notion
The structured record is POST'd to your Notion database via the API. Six fields: Concept, Summary, Mode, Domain Tags, Timestamp, Full Thread. Nothing is stored on Rhinal's server.
Four modes
You select the mode before you type. The model cannot override your choice — it's the tag that organises your vault.
Capture
Log an observation as it happens.
Use when you've noticed something — a clinical finding, a result, an event. No analysis yet, just the fact.
Decision
Record a choice and its reasoning.
Use when you've decided something: a treatment change, a protocol deviation, a diagnostic pivot. Include the why.
Spec
Define a plan, protocol, or structure.
Use when you're designing something: a study design, an intervention protocol, a workflow. Precision matters.
Recall
Surface and connect past knowledge.
Use when you're reviewing: looking back at a case, revisiting a decision, making sense of a pattern over time.
Your data & safety
Maximum privacy — nothing leaves your machine
- The AI model runs on your CPU/GPU via Ollama. No network call for inference.
- Your text never reaches any external server during the thinking/distillation steps.
- The only outbound call is the final Notion write — to your own Notion account.
- Your API keys (if any) are AES-256-GCM encrypted at rest in the database.
- No analytics, no telemetry, no logging of your content on Rhinal's side.
⚠ Before using with real patient data
Real patient data enters legal territory: DPDP Act (India), GDPR (EU), HIPAA (US), and institutional ethics requirements. Using a cloud model with identifiable patient data is almost certainly not compliant without a signed data-processing agreement with the provider. De-identify first, or use a local model only. This tool has not been reviewed for clinical use — that sign-off must come from your institution, not from the software.
A real example
LEPTOSPIROSIS CASE · 3-MONTH FOLLOW-UP
Week 1 · Capture
Presentation: 34yo male, pyrexia 5 days, myalgia, conjunctival suffusion, mild jaundice. Weil's suspected. Started doxycycline empirically pending serology.
Week 1 · Decision
Switched to IV penicillin G after serology confirmed. MAT titre 1:800 (L. interrogans). Decision: add empiric steroids — debated, went against given absence of severe pulmonary haemorrhage.
Week 3 · Capture
Follow-up: bilirubin trending down. Creatinine still elevated (2.1). Patient mobilising. Echo: trace pericardial effusion, clinically silent — plan watchful waiting.
Month 3 · Recall
Reviewed full case longitudinally. Pattern: biphasic course with leptospiraemic → immune phase transition at day 7. Creatinine normalised by week 6. Effusion resolved. Discharge functional. MAT 1:200 at 3 months — expected.
De-identified illustrative example. Not a real patient record.
FAQ
Does Rhinal read my existing Notion content?+
What happens if I close the browser mid-thought?+
Can I use it without Ollama (local model)?+
Can I delete records from Notion?+
Is this a medical device?+
Why Notion and not a built-in database?+
Ready to start?
Four steps. Takes about five minutes. The first thought you save is the start of a record nothing else can hold.
Get started →