Live Demo: https://crunchymon-callinsights-app-iasu4g.streamlit.app/
This is a full-stack web application that automates the analysis of sales call recordings. The system orchestrates multiple AI services to transcribe, enrich, and summarize call data, delivering actionable insights to a user-friendly dashboard and generating a professional PDF report. It's designed to help sales managers quickly understand call outcomes and agent performance without manual review.
-
Multi-Source Audio Ingestion: Accepts audio data via direct file upload (MP3, WAV, etc.) or from a public URL, handling temporary file storage and secure data transfer to the processing backend.
-
Asynchronous Transcription & Enrichment Pipeline: Leverages the AssemblyAI API to perform speaker diarization (identifying who is speaking), sentiment analysis, and key entity detection. Implements a robust polling mechanism to handle the asynchronous nature of the transcription job, waiting for the task to complete before proceeding.
-
Generative AI-Powered Summarization: Takes the structured, enriched transcript from AssemblyAI and pipes it into a Google Gemini model. A custom-engineered prompt instructs the model to perform a targeted analysis, including:
- Identifying key discussion points and customer objections.
- Scoring sales agent performance based on tone and flow.
- Defining actionable next steps and follow-up tasks.
-
Automated PDF Report Generation: Includes a data cleaning module using regex and Unicode normalization to format the AI-generated insights into a clean, professional PDF report, which is made available for users to download.
- Language: Python
- Web Framework: Streamlit
- Core Services: AssemblyAI API (for transcription & enrichment), Google Gemini API (for generative analysis)
- Libraries: Requests, FPDF
- Clone the repository:
git clone [https://github.com/yourusername/CallInsights.git](https://github.com/yourusername/CallInsights.git)
- Install dependencies:
pip install -r requirements.txt
- Run the app:
streamlit run app.py