Skip to main content
Back to challenges
Challenge

OpenAI Agents SDK Conversational Voice Desk with Bland AI

Public infrastructure citizen support services struggle with high caller volume and regional dialects. Build an automated multi-dialectic voice agent using OpenAI Agents SDK and Bland AI to resolve citizen queries regarding identity cards and welfare distributions.

Business OperationsHosted by Vera
Sign in free to start
Challenge brief

What you are building

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

Develop an autonomous multilingual citizen support voice workflow using OpenAI Agents SDK integrated with Bland AI.

Delivery guide

How work is evaluated

Evaluation

Evaluates conversation completion rates and accuracy of service registry dispatch.

Datasets

Challenge datasets

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

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

correct_service_dispatch

Checks if the correct service action was chosen.

Binary check

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

Dimension 2

resolution_rate

Percentage of correctly resolved intents • target: 0.9 • 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

  • Implement OpenAI Agents SDK tool invocation for service registry lookup

  • Integrate Bland AI telephony API for low-latency audio streaming

  • Handle multi-turn context retention across spoken query sessions

Resources and assets

Reference links and supporting material

Dataset notes

Multilingual citizen query transcripts with labeled intent codes.

Challenge setup

Evaluates conversation completion rates and accuracy of service registry dispatch.

Challenge input

Audio transcript JSON

OpenAI Agents SDK

Mandatory requirement

Bland AI

Assigned voice tool

Evaluated output

Agent action JSON

Evaluation checks
  • Checks if the correct service action was chosen.
  • Percentage of correctly resolved intents • target: 0.9 • range: 0-1
Configured evaluation evidence
  • Benchmark: Voice-Service-Bench
  • Resolution Rate target: 90%
  • 1 public reference case
Configured runtime
  • 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

Action Space
  • Bland AI · Required
  • OpenAI · Optional
  • Alibi Detect · Optional

Evaluation contract

  • correct_service_dispatch · Weight 1
  • resolution_rate · Weight 1

Metrics

  • resolution_rate · Target 0.9

Benchmarks

  • Voice-Service-Bench
  • general

Reference cases

  • 1 public reference case

Workflow tasks

  1. Task 1

    Input: Audio transcript JSON

    Output: Agent action JSON

    Execution harness: Python

Evaluation tools

  • OpenAI Agents SDK · Core Framework
  • Bland AI · Voice Automation Tool

Recipe freeze marker

No recipe freeze time is recorded.

Frequently Asked Questions about OpenAI Agents SDK Conversational Voice Desk with Bland AI