QueueLess Agent

Inspiration :

In healthcare coordination, AI agents cannot afford to hallucinate, skip policy checks, or expose private patient details. Traditional LLM chats often act unpredictably when handling appointment changes or cancellations. We built QueueLess Agent to demonstrate how agents can operate with strict runtime governance and deterministic control while maintaining fast, natural multi-turn conversations.

What It Does :

QueueLess Agent acts as an intelligent healthcare assistant that handles patient scheduling queries, appointment lookups, and cancellations:

Patient Profile Lookup: Safely retrieves account preferences (e.g., patient morning preferences) using custom patient lookup tools.
Deterministic Clarification: Refuses to process actions based on ambiguous inputs, prompting patients for required details (like doctor name or appointment ID) before fetching or modifying slots.
Policy Enforcement: Enforces strict step-by-step verification checklists before processing cancellations.
Privacy & Safety: Redacts sensitive PII (such as phone numbers and emails) from internal logs prior to output storage.
Session Continuity: Retains multi-turn conversation context across session restarts.

How We Built It :

AWS Strands SDK: Used as the open-source agent harness to manage tools, plugins, and execution loops.
Groq API (Llama 3): Powers low-latency, high-speed multi-turn chat responses.
Strands Skills Plugin: Defines structured Markdown workflows (skills/appointment_cancellation.md) that guide the agent through mandatory verification steps.
Strands Hooks: Implements runtime hooks for PII redaction and tool rate-limiting guardrails.
Session Persistence: Manages state persistence across user sessions without forcing re-verification.

Challenges We Overcame :

Ensuring Deterministic Execution: Preventing the LLM from making assumptions about appointment times required crafting clear tool schemas and pairing them with Strands Skills plugins.
Stream Cleanup & API Alignment: Handling multi-turn stream parameters gracefully while maintaining clean terminal output.

Accomplishments We're Proud Of :

Achieving instant, natural responses using Groq while maintaining strict safety guardrails.
Successfully implementing end-to-end Strands SDK primitives (Tools, Skills Plugins, Hooks, and Session Persistence) in a functional Python environment.

What's Next for QueueLess Agent :

Integrating real-world EHR (Electronic Health Record) API endpoints via FHIR standards.
Adding automated SMS/Voice reminders powered by AWS SNS and Twilio.