Start an Interview

API first:

Create the API route and +page.server.ts file.

• Install Open AI API

• Configure environment variable

• Import SDK

• Call to GPT to generate a JSON list of questions that contain the question and an accurate answer to that question. The goal is to 1. Generate a list of questions and 2. Generate a list of answers to each question that when we ask GPT to compare the question and answer then it will be able to accurately determine if the questions "wrongness or rightness level".

Depending on if the user opts into multiple choice or not we will need to be able to generate not only one question but three for each. One question will be the AI generated "right answer" and the other two will be "wrong".

The form action should:

• Accept FormData stuff as parameters
• Data passed into the FormData should include:

1. Job description
2. Challenge level
3. Keywords
4. Years of experience

Should execute the following logic/functionality:

IF (user wants text to speech functionality in interview)

Loops through questions and generate MP3 audio files from AWS Polly via AWS Polly SDK. You can create a shell function and return a hard coded mp3 audio file.

END IF:

Loops through questions and answers and stores them both in DB using Prisma ORM.

RETURNS:

• Interview: id, all questions, all answers, mp3 files in necessary.

Page Route next:

Create the +page.svelte

Create a form that uses the use:action attribute and submits to the form action in the +page.server.ts file

Please add wireframes and AC for dashboard page.

Acceptance Criteria:

The landing page will consist of the following UI components:

Product Logo

Navigation Bar

• Home
• Account
• Dashboard

Hero Section:

Headline: Conquer your next interview with AI-powered preparation.

Form

• The form will allow users effortlessly start the interview process weather they are
• creating an interview for themselves or as a recruiter.
• The form will use the Shadcn-Svelte Tab Component.
• This tab component will be used to allow the user to toggle between the form tailored for the interviewee and the form tailored for the recruiter.
• The form will have a clickable component that starts an instant interview using random questinos. (Think google's "I'm feeling lucky feature").
• Users will be able to select keywords to tailor the interview further. Selecting keywords like Javascript, React, System Design, AWS .etc using the Shadcn-Svelte Toggle Component.
  
  Benefits Section:
• Highlight key features like:
• Practice with real-time AI feedback
• Personalized interview question selection
• Improve your communication skills
• Build confidence and reduce interview anxiety

How it Works Section:

• Explain the platform's functionalities in a step-by-step manner with visuals.
• Use clear and concise language.

Stretch Goals:

• Leverage the typewriter animation in the form to show examples of what users can put inside of the form.

Animation and badge when you create your first interview
Animation when you invite someone to partake in an interview you created
Animation and badge when you complete an interview

Progress bar while interview is in progress

Green animation when providing the right answer
Red animation when providing the wrong answer
Yellow animation when answer is not quite right but close

Acceptance Criteria:

A user can interact with a clickable UI component that does the following:

• On click, start listening. Prompt the user with a confirmation (I think this is required now to get it to work).
• Record the audio until timeout or the user clicks the UI component again.

Once the user submits their recording as an answer then execute the following logic:

• Convert/Create audio data in a format that can interface with AWS Transcribe. Transcribe is a cloud service for Speech to Text.
• Store that generated audio transcription into a format that can be forwarded to Open AI API.
• Store the interview question and answer pair in the database. See ERD to determine the schema if Prisma types is not sufficient.

The request take the question being answered and the recorded answer and return a "rightness score". How this score is calculated could be as simple as prompting the model with "score this answer in regards to this specific question and return a JSON object with the level of rightness. Rightness will be on a scale of 0 - 100 or 0 -1.

Pseudo code:

import AI from 'ai';
import Polly from 'aws/polly';
import Transcribe from 'aws/transcribe';

const record = () => return true; // this will start recording audio from the user after permission prompt.


while (user wants to record) {
 // keep recording
 const file = record(true)
// create auido file to send to AWS Transcribe
 const data = convert(data)
}

if (user wants to submit recording) {
 record(false)
 // Transcribe audio data
// Get and  return text from speech
const text = Transcribe(data)
// Then store the interview question in the database for future reference
Prisma.interview.update( { text, ...data) } )
}

As a user I want to be able to authenticate on the platform using:

• Pasword-less authentication w/ Supabase Magic Email
• Linkedin SSO
• Google SSO
• Email and Password (Stretch goal)

You can follow this documentation to get started.

Acceptance Criteria:

• A user will be able to register via methods listed above. Only one is required for MPV however, the rest should be technically feasible and attainable within a reasonable amount time.

We need our database designed and ready for development for the hackathon.

Acceptance Criteria:

1. Implement Supabase databse w/ Prisma integration.
2. Create ERD diagram.
3. Create Prisma Schema.
4. Run migration to create and seed database.
5. Query the database to prove it is working as expected.

It would be cool to have a real payment gateway set up for users to subscribe to the platform

A user can analyze each interview submission by ID and receive an analysis object that ranks/scores the interview submission

Acceptance Criteria:

API Route first:

The API Route should:

Accepte an interview id as a parameter in FormData.

Get that interview submission and related data like interview using Prisma ORM

DOES SOME AI MAGIC

In Prisma create an InterveiwAnalysis object. (Check ERD or Prisma Schema for this)

Returns InterviewAnalysis object.