Can I treat a top-ranked trial as eligible?

No. It never says eligible or qualify, only trials worth discussing; a trial with unchecked rules sits under Not fully checked, and missing items are not proof nothing was missed.

Why do some trials stop early with unchecked rules?

Exclusion rules are judged first; after a confident likely fails the rest stay not checked yet. Opening a trial uses its own budget to check the remainder.

What data leaves my browser during a search?

Parts go to ClinicalTrials.gov as typed condition plus city coordinates and distance, and to Jev as cancer type, stage, age, sex and notes without place; both keep data under their own terms.

Do the tests hit ClinicalTrials.gov and Jev live, or replay fixtures?

Fixtures. The tests have no live network: ClinicalTrials.gov results and Jev answers are replayed from fixtures, and the first run compiles Typia's Go plugin, which takes a couple of minutes before later runs use the cache. The live app needs Node 22.18 or newer with .nvmrc pinning 24, TYPESAFE_API_KEY in apps/worker/.dev.vars, make db-local for D1 plus GeoNames cities at about 3 MB once, then make -j2 dev for the API on 8787 and the web app on 5173.

What license does TrialScout use, and what is it built on?

MIT. It is a TypeScript stack: a Hono API on Cloudflare Workers with D1 via Drizzle, KV for cached verdicts, a Durable Object rate limiter and a daily refresh cron; the web app is Vite, React, Tailwind 4 and shadcn/ui served by a second Worker that forwards /api/*.

How does a search decide which rules Jev sees?

Plain code splits each trial's eligibility text into separate rules with no AI, and hard filters such as condition, status, age, sex and distance run before Jev sees a trial. Jev is the only AI component, answers one typed question per rule, and cannot write text, so all interface wording is written by hand.

How are verdicts shown, and what does it refuse to say?

Every verdict appears next to the exact quoted rule text, and when the profile lacks information the answer is ask your doctor with a question to take to the appointment, never an assumed pass or fail. It never says a patient is eligible or will qualify, and the red not-medical-advice strip is on every page and printed sheet.

CommunityJev Tools & CLIs

Trialscout

Demo that checks a plain-language cancer profile against recruiting ClinicalTrials.gov eligibility rules with one Jev question per rule, quoting source text.

See highlights

Open github.com

Stars: 0
Language: TypeScript
License: MIT
Published by: Community
Author: tushar-im
Last verified: Oct 2026

tushar-im/trialscout

User repository on GitHub

View on GitHub

TrialScout (trialscout.cc) reads a patient's plain-language profile, checks it against every eligibility criterion of recruiting solid-tumour oncology trials on ClinicalTrials.gov using TypeSafe AI's Jev, and shows which trials are worth raising with a doctor. Each criterion is quoted from the source.

Forks: 0
Last pushed: Updated Sep 2026

Repo stats from the GitHub API, cached Oct 2026.

About this project

A demo, not a medical product. It is not medical advice, and no clinician has checked its answers. TrialScout checks a patient's plain-language profile against the eligibility rules of recruiting cancer trials near them on ClinicalTrials.gov, and shows each answer next to the trial's own words. Live at trialscout.cc. You describe the cancer (type, stage, age, sex, where you live, and anything else you know), and it:

Quoted from the project's README. Read the full README.

Tagscommunity github

Works forJev for Classification

Used inJev for Software & DevTools

Languages: TypeScript

Highlights

Splits eligibility text into rules in plain code with no AI, then asks Jev one typed question per rule in a fixed budget.
Shows likely meets, likely fails, ask your doctor, or not checked yet beside exact quoted text and ranks trials.
Reads first 100 trials, caps at 300 questions in 30 requests; unchecked rules stay not checked yet, never passed.
Profile lives in sessionStorage and is dropped after use; Jev verdicts cached 7 days under hashed key with no profile.

Watch out

Demo, not medical advice; no clinician checked answers, and missing or unchecked items prove nothing-talk to a doctor. Copy reviewed against the repository README on 2026-10-04, not from running it.

Questions about this entry

Short answers drawn from the linked project or page.

More Jev resources like this

Jevest

GitHub Action that reviews pull requests with a read-only finder agent plus a refuting verifier, using Jev only to triage, route, label, and gate merging.