ArgusQA

Support Call QA Agent.

AI call QA agent for transcript scoring, policy detection, escalation misses, and coaching notes.

What It Does

• Classifies workflow intent from natural language.
• Extracts IDs, owners, severity, and amount-like values.
• Routes requests through a LangGraph workflow.
• Applies domain policy and human-review rules.
• Persists every turn and decision for QA.
• Exposes queue and analytics APIs.
• Runs locally with Docker Compose, PostgreSQL, and LiveKit.

Architecture

User / Operator
      |
      v
FastAPI Backend -------- PostgreSQL
      |
      v
LangGraph Support Call QA Agent
      ^
      |
LiveKit Voice Worker --- LiveKit Server

Quick Start

git clone https://github.com/SarthakShrivastav-a/ArgusQA.git
cd ArgusQA
copy .env.example .env
docker compose up --build

API health: http://localhost:8000/api/health

Run tests:

python -m pytest -q

Docker Image

sarthak73/argusqa:latest
sarthak73/argusqa:main-<short_sha>
sarthak73/argusqa:dev-latest
sarthak73/argusqa:dev-<short_sha>

Configuration

All runtime values live in config/app.yaml. Do not hardcode provider names, model names, LiveKit agent names, policy keywords, thresholds, or Docker image names.

API Overview

Method	Route	Purpose
GET	/api/health	Health check
POST	/api/workspaces	Create workspace
GET	/api/workspaces/{actor_id}	Read workspace
POST	/api/qa/sessions	Start workflow session
PUT	/api/qa/sessions/{session_id}/turn	Advance workflow
POST	/api/qa/sessions/{session_id}/escalate	Manual escalation
GET	/api/agent/queue	Human-review queue
GET	/api/analytics/summary	Metrics
POST	/api/voice/connect	Create LiveKit token

License

MIT. See LICENSE.