AI SDK Pediatric Rapid WGS Triage Agent with AutoGPT
Pediatric intensive care units demand rapid whole-genome sequencing interpretation to guide life-saving clinical decisions. Build a real-time diagnostic agent with Vercel AI SDK and AutoGPT to orchestrate variant filtering, HPO phenotype matching, and prioritized clinical summary generation. Reach full diagnostic report rendering in under three seconds.
What you are building
The core problem, expected build, and operating context for this challenge.
Construct a streaming pediatric WGS diagnostic agent using Vercel AI SDK integrated with AutoGPT autonomous search capabilities for rapid variant prioritization.
How work is evaluated
Evaluates stream response time, top-1 candidate gene diagnostic match rate, and report clarity.
Challenge datasets
Public datasets, plus the private datasets you have access to.
How submissions are scored
These dimensions define what the evaluator checks and which criteria separate a passable run from a strong one.
correct_top_gene
Verifies that the top predicted gene matches expected clinical diagnosis.
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
top1_accuracy
Diagnostic accuracy for top candidate gene recommendation. • target: 0.92 • range: 0-1
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
What you should walk away with
Build a streaming diagnostic web API with Vercel AI SDK generateText/streamText functions
Integrate AutoGPT autonomous sub-agents for literature validation queries
Match patient Human Phenotype Ontology (HPO) terms with genomic candidate genes
Output prioritized diagnostic reports formatted for intensive care clinicians
Reference links and supporting material
Anonymized diagnostic benchmarking suite containing pediatric phenotypes mapped to HPO codes and candidate gene variant shortlists.
Challenge setup
Evaluates stream response time, top-1 candidate gene diagnostic match rate, and report clarity.
Challenge input
JSON with hpo_terms array and variants array
AI SDK
Required framework for streaming clinical triage responses.
AutoGPT
Integrates autonomous verification tools for variant literature.
Evaluated output
JSON with top_candidate_gene, inheritance_mode, and diagnostic_confidence
- Verifies that the top predicted gene matches expected clinical diagnosis.
- Diagnostic accuracy for top candidate gene recommendation. • target: 0.92 • range: 0-1
- Benchmark: PediatricGenomics-Bench
- Top1 Accuracy target: 0.92
- 1 public reference case
- JavaScript execution harness
- javascript sandbox (unavailable on Versalist)
Summary shows the first task, up to three tools, and up to three evidence items. Open the technical recipe to view all configured items.
View technical recipe
Configured tools
- AutoGPT · Required
- Zed · Optional
- GLM-4 · Optional
Evaluation contract
- correct_top_gene · Weight 1
- top1_accuracy · Weight 1
Metrics
- top1_accuracy · Target 0.92
Benchmarks
- PediatricGenomics-Bench
- biotech
Reference cases
- 1 public reference case
Workflow tasks
- Task 1
Input: JSON with hpo_terms array and variants array
Output: JSON with top_candidate_gene, inheritance_mode, and diagnostic_confidence
Execution harness: JavaScript
Evaluation tools
- AI SDK · Agent UI & streaming framework
- AutoGPT · Autonomous research loop
Recipe freeze marker
No recipe freeze time is recorded.