Skip to content

README.md

Latest commit

 

History

History
174 lines (130 loc) · 6.1 KB

README.md

File metadata and controls

174 lines (130 loc) · 6.1 KB

Healthcare Booking Application

A comprehensive healthcare booking and management system built with autonomous agents for insurance verification, medical triage, and voice interactions.

Project Structure

/documentation

Project documentation and guides. Contains the main README and other supporting documentation files.

/insurance-agent

Insurance verification and benefits checking agent powered by Infinitus API.

  • infinitus/ - Core insurance agent implementation
    • va_mcp.py - Model Context Protocol integration for Insurance API with voice-agent communication
    • README.md - Detailed insurance agent API - MCP documentation

/triage_agent

Medical symptom triage system that assesses patient symptoms and provides preliminary medical assessment and provides recommendations.

  • infermedica/ - Symptom triaging powered by Infermedica API

    • agntcy/ - Identity and Observability components
      • common/ -
        • tbac.py/ - implementation of identity on triage
        • triage_client/ - triage a2a client
      • service/ -
        • triage_service/ - Triage a2a service
      • utilities/ - Testing utilities
        • tbac_bidirectional.py - Identity service bidirectional test client
        • tbac_unidirectional.py - Identity service unidirectional test client
        • test_api_util.py - Triage service test client
      • main.py - Identity wrapped A2a service with observability hooks

  • medical-triage/ - Standalone triage service

    • well-known/ -
      • agent-card.json - Agent metadata and capabilities manifest
    • medical_triage_agent.py - Core triage logic and API endpoints
    • Dockerfile - Containerization configuration
    • docker-compose.yml - Service orchestration
    • requirements.txt - Python dependencies

/voice-agent

Voice interaction agent for patient communication via phone or voice interface.

  • agntcy/ - Voice agent implementation with identity and observe
    • agent/
      • healthcare_agent.py - voice-agent orchestrator
    • models/
      • session.py - session management and persistence
    • services/
      • a2a_client.py - A2A protocol client
      • audio_service.py - Audio input and output service
      • insurance_client.py - Insurance MCP client
      • llm_client.py - LLM processing client
    • main.py - voice-agent entry point and initialization
    • README.md - Detailed voice agent implementation along with identity and observe

/common/agntcy/observe

Observability and metrics collection infrastructure.

  • deploy/ - Docker Compose configurations and deployment scripts

    • docker-compose.yml - Production stack configuration
    • docker-compose-build.yml - Build from source configuration
    • Custom metrics plugin and MCE integration

  • custom_metrics_plugin/ - Custom metrics implementation

    • TokenUsage, AgentErrorCount, ToolCallCount, NumberActiveAgents

  • unit-tests/ - Testing suite

    • a2a_client.py - Agent-to-agent communication test client
    • http_client.py - HTTP API test client

  • Core MCP Servers:

    • va_a2a_mcp.py - Agent-to-agent, Model Context Protocol clients with voice-agent communication
    • va_http_mcp.py - HTTP-based Medical triage, Model Context Protocol based Insurance clients with voice-agent communication

Prerequisites

Environment Variables

Set the following environment variables before running the application:

# Base paths
export APP_HOME=${APP_HOME:-/Users/xiaodonz/Documents/GitHub/cs1}
export TELEMETRY_HUB_HOME=${TELEMETRY_HUB_HOME:-/Users/xiaodonz/Documents/GitHub/telemetry-hub}

# Observability
export OTLP_HTTP_ENDPOINT="http://localhost:4318"

# LLM Configuration
export JWT_TOKEN="your-jwt-token"
export ENDPOINT_URL="your-endpoint-url"
export PROJECT_ID="your-project-id"
export CONNECTION_ID="your-connection-id"

# MCP Configuration
export MCP_URL="https://mcp.unstable.infinitusai.dev/mcp"
export X_INF_API_KEY="your-api-key"

# A2A Configuration
export A2A_SERVICE_URL="http://localhost:8887"
export A2A_MESSAGE_URL="http://localhost:8887"
export A2A_API_KEY="your-a2a-key"

Getting Started

1. Observability Stack

Start the observability services (ClickHouse, OTEL Collector, MCE, Grafana):

cd common/agntcy/observe/deploy
docker-compose up -d

Or use the build version if building from source:

docker-compose -f docker-compose-build.yml up -d

Access Grafana at http://localhost:3000 (admin/admin)

2. Voice Agent

cd voice_agent
source venv/bin/activate
python3 -m agntcy.main

See /voice_agent/README.md for detailed setup.

3. Triage Agent

See /triage_agent/README.md and /triage_agent/medical-triage/README.md

4. Insurance Agent

See /insurance-agent/README.md

Observability

The observability stack includes:

  • OpenTelemetry Collector - Telemetry ingestion
  • ClickHouse - Time-series database for traces and metrics
  • Metrics Computation Engine (MCE) - Computes 17+ metrics including:
    • Custom metrics: TokenUsage, AgentErrorCount, ToolCallCount, NumberActiveAgents
    • Judge metrics: GoalSuccessRate, Groundedness, Consistency, ContextPreservation, etc.
    • Confidence metrics: LLMAverageConfidence, LLMMaximumConfidence, LLMMinimumConfidence
  • Grafana - Dashboards for visualization

Metrics are automatically collected by the observe SDK. See /common/agntcy/observe/deploy/ for setup and configuration.

Key Technologies

  • MCP (Model Context Protocol) - Agent communication standard
  • Infermedica API - Medical symptom analysis
  • Infinitus API - Insurance benefits verification
  • OpenTelemetry - Distributed tracing and observability
  • ClickHouse - Analytics database
  • Grafana - Monitoring dashboards
  • Docker - Containerization and orchestration
  • Identity - Task-based Access Control and enforcing policies on agent communication

Agent Communication

Agents communicate with each other, enabling:

  • Model Context Protocol (MCP) tool calls
  • Seamless agent-to-agent (A2A) interactions
  • Standardized HTTP-based APIs
  • Identity management and authentication
  • Distributed tracing across agent boundaries