Vercel AI SDK Real-Time Commercial Office Investment Dashboard
Institutional capital is returning to top-tier commercial office markets through selective target investments. Build a real-time streaming investment screening agent using Vercel AI SDK (`ai`) to process market sentiment, vacancy rates, and capital return signals live.
What you are building
The core problem, expected build, and operating context for this challenge.
Develop a real-time streaming office investment dashboard using Vercel AI SDK to filter high-conviction commercial acquisitions.
How work is evaluated
Evaluates schema accuracy and streaming response correctness for office screening.
Shared data for this challenge
Review public datasets and any private uploads tied to your build.
How submissions are scored
These dimensions define what the evaluator checks and which criteria separate a passable run from a strong one.
zod_schema_match
Ensures returned object validates against defined Zod target schema.
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
conviction_accuracy
Accuracy of conviction tier assignments against benchmark rules. • target: 0.95 • 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
Use Vercel AI SDK `streamText` and `generateObject` with TypeScript schemas
Implement structured dynamic UI streaming using React / Next.js hooks
Filter market signals based on physical vacancy vs effective post-tech occupancy
Construct dynamic investment conviction tags (BUY, HOLD, AVOID) based on capital return data
Reference links and supporting material
Quarterly market telemetry for 25 primary and secondary US office submarkets.
How this agent runs
Evaluates schema accuracy and streaming response correctness for office screening.
Challenge input
JSON market metrics (market_name, prime_vacancy, badge_swipe_rate, capital_inflow_m)
AI SDK
Provides TypeScript-first streaming and structured object generation.
Evaluated output
JSON with market_name, conviction_rating, rationale
- Ensures returned object validates against defined Zod target schema.
- Accuracy of conviction tier assignments against benchmark rules. • target: 0.95 • range: 0-1
- Benchmark: CREMarketScreeningBench
- Conviction Accuracy target: 0.95
- 1 public reference case
- JavaScript execution harness
- javascript sandbox (unavailable on Versalist)
View technical recipe
Configured tools
- Vercel AI SDK · Required
- Hume AI · Optional
- Arize AI · Optional
- Vercel AI SDK · Required
Evaluation contract
- zod_schema_match · Weight 1
- conviction_accuracy · Weight 1
Recipe state
This is a preview. The configuration can change before the evaluation recipe is locked.