Nida - A simple, yet powerful Contact Center Analytics Solution

Nida is a contact center analytics solution that allows you to analyze customer calls using AI. The solution supports easy creation of multiple persons to evaluate the calls and thus allows to extract department specific information. For example:

• Sales Quality Manager --> Able to easily extract insights about the agents' sales qualities, e.g., up-selling attempts, empathy, etc.
• Marketing Manager --> Extract sentiment, awareness regarding promotions, etc.
• Product Manager --> Extract product related issues and pain points of customers

The project consists of a multi-stage pipeline to process calls:

• Transcription via whisper or gpt-4o-audio-preview, subsequent diarization (if required) via gpt-4o
• Insight extraction via gpt-4o, based on user-defined personas
• Analytics based on PowerBI (optional)

Getting Started

Prerequisites

• Docker
• Azure CLI
• Azure Developer CLI
• Available quota for Azure OpenAI in your Azure subscription (rates can be adjusted in infra/app/src.bicep):

Model	Deloyment Type	Quota
gpt-4o (2024-11-20)	GlobalStandard	30k TPM
whisper-001	Standard	1 RPM
gpt-4o-audio-preview (2024-12-17)	GlobalStandard	80k TPM

For example swedencentral

Deployment to Azure (via azd up)

1. Clone the repo
2. az login
3. azd up

Then visit the azurecontainerapps URL that is returned. In the UI, you can upload the sample calls from samples/audios/*.wav and create a first persona from samples/marketing_sentiment_details.txt.

Local deployment (manually)

1. cd src
2. cp .env.sample .env and update .env with your valid keys, endpoint, and settings
3. Make sure you have the relevant permissions on the Storage Account (Storage Blob Data Contributor and Storage Queue Data Contributor), Azure OpenAI resource (Cognitive Services OpenAI User) and Azure Search (Search Service Contributor, Search Index Data Contributor and make sure to enabled RBAC-access under Keys)
4. pip install -r requirements.txt
5. streamlit run main.py
6. Then head to the Diagnostics page and make sure all tests pass.

Overview

This project consists of:

• infra/ – Bicep files for Azure infrastructure deployments.

  • app/src.bicep is the main file to modify.
  • Refer to azure.yaml for basic service definitions, such as container app settings, Dockerfiles, and any optional resource group info (commented).

• src/ – Main application code with a Streamlit app.

  • Main.py sets up Streamlit configuration, page title, and loads a logo.
  • Other Python files (e.g., transcriptions.py) import services like azure_storage and azure_transcription.

• samples/ – Sample person definition and sample audio call.