tulasiram2005/HealthRoute-AI-Agentic-RAG-for-Medical-Desert-Intelligence

HealthRoute AI is an agentic IDP + RAG system that extracts, validates, and reasons over Ghana healthcare facility data using LangGraph, LLMs, Text2SQL, MLflow, and citations. It identifies medical deserts, flags suspicious clinical claims, and simulates specialist/patient routing to improve healthcare access decisions. # Virtue Foundation IDP Agent Agentic Intelligent Document Parsing project for the Databricks x Accenture healthcare coordination challenge. It extracts structured facts from messy medical facility text, validates claims, scores medical deserts, recommends interventions, and exposes both API and Streamlit planner UI. ## What Is Included - LangGraph-compatible five-step agent pipeline with deterministic fallback - CrewAI role definitions for extractor, validator, analyst, planner, and reporter agents - Strict Pydantic v2 models for facilities, NGOs, extractions, anomalies, care index, and doctor matching - Rule-based anomaly detection for ICU, surgery, imaging, capacity, NICU, trauma, and specialty contradictions - Care Access Index and medical desert prioritization - Doctor-facility matching engine with citations and patient impact estimates - Grounded query engine for the hackathon must-have questions - Specialist deployment scenario modeling with projected access-index lift - Field verification queue for suspicious facility claims before patient routing - Facility readiness ranking and specialty heatmap for non-technical planners - Board-ready HTML audit report with source rows, evidence, and intervention rationale - Offline-first Plotly Ghana impact map that does not depend on web map tiles - Tech Stack Proof page showing how the requested challenge stack is implemented - Credential-ready Databricks LLM, Databricks Genie-style, MLflow, and LanceDB adapters - LanceDB-ready RAG ingestion with JSON fallback and hybrid retrieval - FastAPI backend with query, facility analysis, care index, anomaly, map, matching, scenario, verification, and report endpoints - Streamlit command center with executive overview, agent Q&A, medical desert map, deployment planner, verification queue, gap lab, and audit report - Sample Ghana facility CSV and pytest suite ## Quickstart ```bash cd virtue_foundation_idp python3 -m venv .venv source .venv/bin/activate pip install -r require …