Docvoxia AI Agents for RealTime Multilingual OPD Environment

A Healthcare AI Agent that uses ADK with Gemini model and MongoDB MCP to transform doctor-patient conversations into intelligent clinical workflows, prescriptions, and patient care actions.

Comment

Inspiration

Healthcare providers spend significant time handling repetitive documentation tasks instead of focusing on patient care.

During OPD consultations, doctors must collect patient history, understand symptoms, document findings, prepare prescriptions, and maintain medical records manually.

This creates workflow delays, language barriers, and documentation errors.

We built Docvoxia to create an AI healthcare agent that does more than answer questions — it understands context, plans actions, and executes real clinical workflows.

What We Built

Docvoxia is an autonomous healthcare workflow agent powered by Gemini.

The agent listens to doctor-patient conversations, understands clinical context, retrieves relevant patient information, generates structured medical outputs, and assists with follow-up workflows.

The system uses multiple specialized capabilities:

Patient Intelligence Agent

Retrieves patient history and previous encounters using MongoDB-backed healthcare data.

Clinical Reasoning Agent

Analyzes symptoms, observations, and medical context to assist doctors.

Prescription Agent

Creates structured prescription drafts including medication, dosage, duration, and instructions.

Follow-up Agent

Maintains patient context and helps answer post-consultation questions.

Agent Workflow

Patient Conversation → Gemini Analysis & Reasoning → Patient Context Retrieval (MongoDB) → Clinical Decision Workflow → Prescription & Medical Record Generation → Patient Encounter Storage (MongoDB)

MongoDB MCP Integration

MongoDB provides the agent with secure access to healthcare data through MCP tools.

The agent can:

  • Retrieve patient history
  • Search previous consultations
  • Store new encounters
  • Update treatment information
  • Maintain long-term patient context

This enables the agent to complete multi-step healthcare tasks instead of only generating responses.

Challenges

The biggest challenge was designing reliable agent workflows.

Healthcare requires accuracy, safety, and human oversight.

We learned that effective AI agents need:

  • Clear tool boundaries
  • Structured outputs
  • Context management
  • Human approval for critical decisions

Future Vision

Docvoxia aims to become an AI workforce for hospitals, helping healthcare professionals reduce administrative workload and deliver faster, safer care.

Built With

  • fastapi
  • fhir-healthcare-standard
  • google-cloud-agent-builder
  • google-cloud-run
  • google-cloud-vertex-ai
  • google-gemini
  • model-context-protocol-mcp
  • mongodb-atlas
  • mongodb-mcp-server
  • multi-agent-architecture
  • python
  • rag
  • snomed-ct
  • speech-to-text
Share this project:

Updates