AutoAuth Backend

The AutoAuth Backend is an AI-powered system designed to automate the Prior Authorization (PA) process in healthcare. It leverages agentic workflows, Retrieval-Augmented Generation (RAG), and modern LLM frameworks to analyze patient data against insurance policies and determine eligibility.

🚀 Key Features

• Agentic Workflows: Multi-agent system comprising Gap Analysis, Eligibility Check, and PA Document Generation agents.
• Structured RAG: Efficiently retrieves specific policy rules (Required Documents, Eligibility Criteria) from a ChromaDB vector store.
• Patient Data Integration: Processes EHR records and patient-uploaded evidence to identify clinical gaps.
• Observability: Fully instrumented with Arize Phoenix for LangChain trace analysis and debugging.
• Admin Dashboard: Integrated SQLAdmin for easy management of Users, Cases, and EHR records.
• Automated Payer Submission: Tools for generating and submitting structured PA packets to payers.

🛠 Tech Stack

• Framework: FastAPI
• ORMs/DB: SQLAlchemy with SQlite (Configurable)
• Vector Store: ChromaDB
• AI Framework: LangChain
• LLMs Supported: Google Gemini, OpenAI, Hugging Face
• Observability: Arize Phoenix
• PDF Processing: pdfplumber, pypdf, reportlab

📁 Project Structure

backend/
├── agents/          # AI Agent logic (Eligibility, Gap Analysis, PA Packet)
├── api/             # FastAPI routers and endpoints (v1)
├── core/            # Configuration and global settings
├── db/              # Database session and base model definitions
├── models/          # SQLAlchemy data models (User, Case, EHR)
├── scripts/         # Data migration and seeding scripts
├── services/        # Business logic and external service integrations
├── tools/           # Custom LangChain tools (PDF extraction, Policy retrieval)
├── orchestrator/    # Workflow management and case status transitions
└── main.py          # Application entry point

⚙️ Setup & Installation

1. Prerequisites

• Python 3.10+
• Virtual Environment (recommended)

2. Install Dependencies

pip install -r Requirements.txt

3. Environment Variables

Create a .env file in the backend/ directory:

OPENAI_API_KEY=your_key
GOOGLE_API_KEY=your_key
PHOENIX_ENABLED=true
DATABASE_URL=sqlite:///./data/autoauth.db

4. Database Setup

The system automatically handles migrations and seeding on startup (via main.py). To manually seed:

python scripts/seeder.py
python scripts/ingest_policies.py

5. Running the API

uvicorn main:app --reload

The API will be available at http://localhost:8000.

• Swagger Docs: http://localhost:8000/docs
• Admin Panel: http://localhost:8000/admin
• Phoenix Dashboard: http://localhost:6006

📊 Observability

This project uses Arize Phoenix for tracing AI agent performance. When enabled, all LangChain runs are automatically logged, allowing you to visualize retrieval steps and LLM outputs in real-time.