Jun 7, 2026

Aura - Your Personal AI Assistant

ai llm voice ai avatar productivity agentic ai langgraph gemini trugen

Aura is a full-stack AI Chief of Staff that lets you talk to a real-time conversational video avatar and get actual digital work done across your apps — not just chat, but execution.

The core idea: you speak naturally to a hyper-realistic AI avatar (powered by TruGen's Huma model), and it performs real tasks — drafting and sending emails, scheduling meetings, creating GitHub repos, writing PRDs, building Notion pages, setting up Linear backlogs, researching competitors, and generating launch checklists — all from a single conversation.

The hero demo: tell Aura "I have an idea for an AI SaaS" and watch it ask clarifying questions, write a full PRD, create a Notion page, open a GitHub repo, populate a Linear backlog, schedule Calendar milestones, and generate a launch checklist — all while talking through the avatar in real time.

How it's built: TruGen handles the avatar's face, voice, and ears. The brain is a LangGraph agent running on Gemini 3.1 Flash Lite, exposed as an OpenAI-compatible streaming endpoint that TruGen calls as its LLM. All tool execution — Gmail, Google Calendar, GitHub, Notion, Linear — happens server-side with the user's OAuth tokens stored encrypted in Supabase. The same agent also powers a text chat fallback in the UI.

Every integration is declared in a capability registry. If a service isn't connected, the avatar says "that service isn't linked — connect it in Settings" instead of failing. Users manage their connections from a Settings page in the app.

Stack: Next.js · FastAPI · LangGraph · Gemini 3.1 Flash Lite · TruGen · Supabase · Framer Motion

This build was uploaded as a hackathon project

Hackathon

Hack Days in Delhi

View All Projects
Give a star to encourage!Discussion
Start a new conversation!
Login to join the discussion