aarushg025/call-center-compliance

Call Center Compliance API

Description

A RESTful API built for Track 3 of the Compliance Hackathon. The API is designed to accept Base64-encoded audio directly from JSON payloads. It leverages a blazing-fast Multi-Stage AI Pipeline:

1. 1.ASR (Speech-to-Text): Faster-Whisper transcribes Hinglish/Tanglish seamlessly using PyAV audio manipulation.
2. 2.NLP & Data Extraction: Gemini 2.5 Flash / Llama 3.3 dynamically evaluate transcripts against specific SOP criteria. Using Pydantic Schema-enforced prompts, the LLM maps the conversational data precisely to required specific categorical enumerations (EMI, BUDGET_CONSTRAINTS).
3. 3.Semantic Indexing: ChromaDB indexes conversation transcripts and metadata for immediate downstream context retrieval and semantic relevance validation.

Tech Stack

• —Language/Framework: Python 3.12, FastAPI
• —Transcription Library: Faster-Whisper, PyAV
• —Vector Database: ChromaDB
• —LLM / AI Models: Google Gemini 2.5 Flash (Primary, structured JSON generation), Groq Llama-3.3-70b (Fallback robust NLP generation)

Setup Instructions

1. 1.Clone the repository:

git clone <your-repo-link>
cd your-repo

1. 1.Install dependencies:

python -m venv .venv
# Activate Environment depending on OS
.venv\Scripts\activate
pip install -r requirements.txt

1. 1.Set environment variables: Create a .env file matching .env.example:

API_SECRET_KEY=sk_track3_987654321
GEMINI_API_KEY=your_key
GROQ_API_KEY=your_key

1. 1.Run the application:

uvicorn src.main:app --host 0.0.0.0 --port 8000 --reload