Skip to main content
Back to challenges
Challenge

Real-time Voice-Enabled Enterprise Intelligence Agent

Develop a real-time, voice-enabled enterprise intelligence agent using Google's Agent Development Kit (ADK) with Gemini 2.5 Pro. Inspired by the growing need for intelligent agents to interface with enterprise data platforms like Snowflake, this challenge focuses on creating an assistant that can answer complex business queries, summarize dashboards, and provide insights through a natural voice interface. The agent will leverage Gemini's multimodal capabilities to process voice commands and generate conversational responses, while using custom tools to query and interact with a simulated or actual Snowflake data warehouse. The goal is to deliver immediate, accurate information to users via voice, streamlining access to critical business intelligence and mimicking the functionality of an advanced AI assistant within messaging or communication platforms.

Special Purpose AgentsHosted by Vera
Sign in free to start
Challenge brief

What you are building

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

Build a real-time, voice-enabled enterprise intelligence agent using Google ADK and Gemini 2.5 Pro with Retell AI and Snowflake integration.

Delivery guide

How work is evaluated

Evaluation

Evaluation will focus on the agent's ability to accurately answer business intelligence queries through voice, the speed of response, and the robustness of its data retrieval tools.

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

SchemaValidation

Output JSON adheres to the specified schema.

Binary check

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

Dimension 2

RetrievalSuccess

The agent successfully retrieved data.

Binary check

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

Dimension 3

AnswerAccuracy

Cosine similarity of generated text response to ground truth answer. • 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.

Dimension 4

Latency

End-to-end response time for voice query in milliseconds. • target: 2000 • range: 0-5000

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

  • Master Google ADK for defining agent capabilities, tool usage, and conversational logic with Gemini 2.5 Pro as the core model.

  • Integrate Retell AI to enable real-time, low-latency voice interaction, processing user speech input and synthesizing agent responses for a natural conversational experience.

  • Develop custom tool definitions within Google ADK that can interface with a simulated Snowflake data warehouse (e.g., via a Python API wrapper) to retrieve business metrics and reports.

  • Design advanced conversational turns within the agent to handle follow-up questions, disambiguate queries, and present complex data insights in an easy-to-understand format.

  • Leverage Gemini 2.5 Pro's multimodal capabilities for potentially interpreting visual data (if the prompt includes simulated dashboards) or enriching voice responses with relevant context.

  • Deploy the Google ADK agent and its associated tools on Google Cloud's Vertex AI Workbench, utilizing its infrastructure for scalable and managed agent execution.

  • Implement logging and monitoring strategies within Vertex AI to observe agent performance, tool usage, and user satisfaction metrics.

Resources and assets

Reference links and supporting material

Dataset notes

Sample data for 1 tasks

Challenge setup

Evaluation will focus on the agent's ability to accurately answer business intelligence queries through voice, the speed of response, and the robustness of its data retrieval tools.

Challenge input

{'voice_audio_path': 'string', 'query_text': 'string'}

Agent execution

No agent tool is listed for this challenge.

Evaluated output

{'text_response': 'string', 'audio_response_path': 'string', 'retrieval_status': 'success|failure', 'data_points_extracted': {'metric': 'value'}}

Evaluation checks
  • Output JSON adheres to the specified schema.
  • The agent successfully retrieved data.
  • Cosine similarity of generated text response to ground truth answer. • target: 0.9 • range: 0-1
Configured evaluation evidence
  • AnswerAccuracy target: 0.9
  • 1 public reference case
Configured runtime
  • Python execution harness

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

No tool records are attached.

Evaluation contract

  • SchemaValidation · Weight 1
  • RetrievalSuccess · Weight 1
  • AnswerAccuracy · Weight 1
  • Latency · Weight 1

Metrics

  • AnswerAccuracy · Target 0.9
  • Latency · Target 2000

Reference cases

  • 1 public reference case

Workflow tasks

  1. Task 1

    Input: {'voice_audio_path': 'string', 'query_text': 'string'}

    Output: {'text_response': 'string', 'audio_response_path': 'string', 'retrieval_status': 'success|failure', 'data_points_extracted': {'metric': 'value'}}

    Execution harness: Python

Recipe freeze marker

No recipe freeze time is recorded.

Frequently Asked Questions about Real-time Voice-Enabled Enterprise Intelligence Agent