Back to challenges
Challenge

Deploy Real-time Anti-Scam Voice Intelligence with Pydantic AI and VAPI

Citizens face increasingly sophisticated voice scams targeting personal bank accounts. Build a real-time scam threat analysis agent using Pydantic AI for type-safe structured validation and VAPI for low-latency conversational audio transcript ingestion.

Workflow AutomationHosted by Vera
Sign in free to start
Challenge brief

What you are building

The core problem, expected build, and operating context for this challenge.

Implement a structured Pydantic AI agent integrated with VAPI to analyze live call transcripts and flag impersonation scams with typed risk metrics.

Delivery guide

How work is evaluated

Evaluation

Evaluates Pydantic AI agent speed and typed detection accuracy on scam transcript streams.

Datasets

Shared data for this challenge

Review public datasets and any private uploads tied to your build.

Evaluation rubric

How submissions are scored

These dimensions define what the evaluator checks and which criteria separate a passable run from a strong one.

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

pydantic_schema_validation

Ensures output strictly conforms to ThreatAssessment model schema

Binary check

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

Dimension 2

scam_detection_recall

Recall metric on identifying scam indicators in test transcripts • target: 0.95 • 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

What you should walk away with

  • Build type-safe AI agents with Pydantic AI

  • Process streaming voice transcript events from VAPI audio webhooks

  • Extract coercive persuasion and government impersonation markers

  • Generate typed alert payloads for immediate bank transfer interventions

Resources and assets

Reference links and supporting material

Dataset notes

Synthetic audio transcripts of legitimate public agency calls and scam impersonation attempts.

How this agent runs

Evaluates Pydantic AI agent speed and typed detection accuracy on scam transcript streams.

Challenge input

JSON object representing transcript event frame with caller_id and transcript_segment

Pydantic AI

Guarantees strict schema outputs for safety critical decisions

VAPI

Provides real-time conversational streaming ingestion

Evaluated output

JSON object strict matching ThreatAssessment model (is_scam, threat_category, urgency_score)

Checks for
  • Ensures output strictly conforms to ThreatAssessment model schema
  • Recall metric on identifying scam indicators in test transcripts • target: 0.95 • range: 0-1
Proof of success
  • Benchmark: ScamDetect-2026
  • Scam Detection Recall target: 0.95
  • 1 public reference case
Runtime evidence
  • Python execution harness
  • Python sandbox (unavailable on Versalist)
View technical recipe

Configured tools

Action Space
  • Pydantic AI · Required
  • VAPI · Optional
  • LangWatch · Optional
Orchestration
  • Pydantic AI · Required

Evaluation contract

  • pydantic_schema_validation · Weight 1
  • scam_detection_recall · Weight 1

Recipe state

This is a preview. The configuration can change before the evaluation recipe is locked.

Frequently Asked Questions about Deploy Real-time Anti-Scam Voice Intelligence with Pydantic AI and VAPI