This is an archived repository. The new Noora code can be found here.

Noora

Use AI to improve your social conversation.

A platform utilizing conversational AI to improve the social conversation of individuals with Autism Spectrum Disorder (ASD).

---

Overview

Noora is is a research project led by Prof. Monica Lam, from Stanford University, with the goal of improving the social conversation of people with Autism Spectrum Disorder (ASD). The project is a collaboration between Stanford's Open Virtual Assistant Lab and Dr. Lynn Koegel from Stanford's Dept. of Psychiatry and Behavioral Sciences.

This project built with Next.js, Tailwind CSS, OpenAI's GPT-3, and Microsoft Azure.

Developer Setup

To run Noora locally for development, install the dependencies from your package manager and start the Next.js app.

yarn install
yarn run dev

Create a .env.local file for the following environment variables.

For the speech-to-text and text-to-speech functionality, create an Azure speech resource (and a corresponding resource group).

OPENAI_API_KEY=[your API key]
SPEECH_KEY=[your API key]
SPEECH_REGION=[your API key]

Repo Structure

pages: all of the Next.js app's paths (e.g., noora.tsx, _404.tsx). These files simply return components, typically wrapped by the <Page> component.

components: the bulk of the website code. Folders under this directory reference the pages of the website, with the exception of components/global and components/interfaces.

• components/interfaces contains the code behind the Noora Chat feature and the Ask Noora feature.

data: all static or pre-written data, such as GPT-3 prompts, statement banks, routes.

scripts: code that is called by components.

• /scripts/noora-chat (and, more specifically, get-reply.ts) contains the logic and flow for Noora's replies, and is used in MessageBox.tsx (where users submit their message in the Noora Chat).

pages/api: the REST API endpoints. Rate limiting is in place for endpoints (implementation found in scripts/rate-limit.ts).

• api/openai calls OpenAI's Completion endpoint and returns the resulting text and logprobs.
• api/get-speech-token.ts retrieves an ephemeral authentication token for Microsoft Azure's speech service.

public: the public assets including images and static files.