Multifamily Demand-Handling Voice Agent using CrewAI and VAPI Integration
Multifamily operators struggle with lead conversion due to slow response times and inefficient inquiry handling. Build a multi-agent CrewAI orchestration with VAPI voice interface integration that handles prospective tenant inquiries and schedules property tours with under 2-second response latency.
What you are building
The core problem, expected build, and operating context for this challenge.
Design a collaborative CrewAI agent ensemble paired with VAPI real-time voice streaming to capture prospect preferences, check lease management DBs, and book automated tours.
How work is evaluated
Tests CrewAI task completion for extracting tenant preferences and booking tours via simulated VAPI call payloads.
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.
Qualification Evaluation
Checks if qualification decision matches financial parameters.
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
Tour Booking State
Verifies that tour booking succeeded without errors.
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
intent_extraction_accuracy
Accuracy of unit target and timestamp extraction from call transcript • target: 1 • 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
Configure CrewAI agents with dedicated roles for lead intake, lease database lookup, and tour scheduling.
Integrate VAPI Webhooks to stream audio transcriptions to the agent ensemble in real time.
Enforce tenant pre-qualification parameters (income-to-rent ratio, move-in timeline).
Achieve zero booking conflicts across automated scheduling calendar endpoints.
Reference links and supporting material
Dataset of 80 real estate inquiry call transcripts with annotated prospect financial attributes, preferred move-in dates, and target property specs.
Challenge setup
Tests CrewAI task completion for extracting tenant preferences and booking tours via simulated VAPI call payloads.
Challenge input
JSON transcript event from VAPI
CrewAI
Mandatory framework for agent team orchestration.
VAPI
Required interface for voice transcript input and audio stream integration.
Evaluated output
JSON with qualification_status, target_unit, requested_time, and tour_booked
- Checks if qualification decision matches financial parameters.
- Verifies that tour booking succeeded without errors.
- Accuracy of unit target and timestamp extraction from call transcript • target: 1 • range: 0-1
- Benchmark: Voice Conversational Agent Benchmark
- Intent Extraction Accuracy target: 1
- 1 public reference case
- Python execution harness
- Python 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
- VAPI · Required
- CrewAI · Optional
- Vijil · Optional
Evaluation contract
- Qualification Evaluation · Weight 1
- Tour Booking State · Weight 1
- intent_extraction_accuracy · Weight 1
Metrics
- intent_extraction_accuracy · Target 1
Benchmarks
- Voice Conversational Agent Benchmark
- general
Reference cases
- 1 public reference case
Workflow tasks
- Task 1
Input: JSON transcript event from VAPI
Output: JSON with qualification_status, target_unit, requested_time, and tour_booked
Execution harness: Python
Evaluation tools
- CrewAI · Agent Framework
- VAPI · Voice Interface Runtime
Recipe freeze marker
No recipe freeze time is recorded.