Skip to main content
Loading navigation
Back to challenges

Deploy a Multilingual Healthcare Triage Assistant with Vercel AI SDK

Expanding public healthcare public infrastructure requires accessible, multi-lingual triage channels. In this challenge, you will construct a streaming public health triage agent using Vercel AI SDK. The system parses user voice/text symptoms in regional Indian languages, performs primary health severity classification, maps symptoms to appropriate primary healthcare centers (PHC), and streams real-time guidance to low-bandwidth mobile web devices.

Business OperationsHosted by Vera
Sign in free to start

What you are building

Build a streaming multi-lingual public healthcare triage agent using Vercel AI SDK for mobile devices.

How work is evaluated

Evaluation

Evaluates triage severity classification accuracy for emergency vs routine clinical symptoms.

Datasets

Public datasets, plus the private datasets you have access to.

How submissions are scored

Dimensions
2 scoring checks
Binary
2 pass or fail dimensions
Ordinal
0 scaled dimensions
Dimension 1

cardiac_emergency_test

Ensures chest pain and arm numbness triggers an IMMEDIATE EMERGENCY classification.

Binary check

This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.

Dimension 2

triage_safety_recall

Zero false negatives on emergency symptom cases. • target: 1 • range: 0-1

Binary check

This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.

Learning goals

  • Build real-time streaming health UI handlers with Vercel AI SDK.

  • Process natural language symptom descriptions in Indic languages.

  • Perform clinical triage severity categorization (Emergency, Urgent, Routine).

  • Output low-bandwidth lightweight payload formats for rural connectivity.

Resources

Dataset notes

Multilingual clinical symptom query dataset mapped to standard ESI triage severity levels.

Challenge setup

Evaluates triage severity classification accuracy for emergency vs routine clinical symptoms.

Challenge input

JSON with user_symptom_description, patient_age, primary_language.

AI SDK

Required UI framework for low-latency streaming.

Evaluated output

JSON with triage_level, recommended_action, urgent_flag.

Evaluation checks
  • Ensures chest pain and arm numbness triggers an IMMEDIATE EMERGENCY classification.
  • Zero false negatives on emergency symptom cases. • target: 1 • range: 0-1
Configured evaluation evidence
  • Benchmark: Clinical Triage Benchmark
  • Triage Safety Recall target: 1
  • 1 public reference case
Configured runtime
  • JavaScript execution harness
  • javascript sandbox (unavailable on Versalist)

Shows the first task, up to three tools, and up to three evidence items.

View technical recipe

Configured tools

Action Space
  • Vercel AI SDK · Required
  • Guidance · Optional
  • Hyperbolic · Optional
Orchestration
  • Vercel AI SDK · Required

Evaluation contract

  • cardiac_emergency_test · Weight 1
  • triage_safety_recall · Weight 1

Metrics

  • triage_safety_recall · Target 1

Benchmarks

  • Clinical Triage Benchmark
  • biotech

Reference cases

  • 1 public reference case

Workflow tasks

  1. Task 1

    Input: JSON with user_symptom_description, patient_age, primary_language.

    Output: JSON with triage_level, recommended_action, urgent_flag.

    Execution harness: JavaScript

Evaluation tools

  • AI SDK · Streaming Framework

Recipe freeze marker

No recipe freeze time is recorded.

Frequently Asked Questions about Deploy a Multilingual Healthcare Triage Assistant with Vercel AI SDK