Agents & human interaction
From conversation to action.
A voice-first assistant prototype that connects natural conversation with tools for reminders, appointments, and personal records.
The problem
Everyday tasks often involve several screens and forms. This project explores a conversational interface for older adults, translating a spoken request into a structured action while keeping the conversation understandable.
How it comes together
Listen
Voice or text input enters through the web interface.
Understand
The agent uses conversation history to interpret the request.
Act
LangChain selects tools for reminders, records, and appointments.
Respond
Task results return to the conversation and structured data is stored.
My contribution
- Connected voice and text interaction with an asynchronous FastAPI and WebSocket service.
- Organized task capabilities as LangChain tools, with conversation memory and a LangGraph workflow implementation.
- Integrated persistent records and task queries so the assistant can act on stored information.
The result
An end-to-end assistant prototype that demonstrates conversation, tool execution, and data persistence in one flow. The repository includes both the agent implementation and the structured conversation workflow.