Inspiration
Began as an idea of Bucket List & break down tasks for users to have visual for better understanding with random inspirational quotes to enhance user's confidence (17~25 years old). Then I pitched this idea to my design class professor from standford which were super interested and asked, "Do people actually understand what they understand?". And he went through a whole question to tell my plan of my long-term goal I'm trying to achieve in 3~5 years. Which comes out, in most time I did had some sort of plan, it just needed a bit of time or good question to trigger myself to have the vision what to do for this dream come true which I wanted to spread this feeling to many other students who feel lost.
What it does
It helps users uncover core values by listening and asking counter-questions like why or how they feel a certain way. This enhances critical thinking, letting users reach their own conclusions. Rooted in counseling psychology, it offers easier access than traditional therapy. Interstly, I have interviewed some of my colleague, data shows most of them were open about sharing concerns or secrets with AI chatbot than with family or friends, especially males. Which I wanted to create a place that it is always here for the user just to listen and support them with their insane inner thoughts goal.
How we built it
I'm yet currently at front-end development process, but the rough plan is to integrate exisitng AI services since I don't have enough time to build or train a LLM or NPL. I made a workflow chart to see what will be the main UX design and then see what's the most difficult one which is the 1st priority to complete, the AI contribution. I wanted to make an AI that has a natural voice which feels like talking to humen. There were several services such as sesame, ElevenLab, Hume AI and more. But I decided to go with Deepgram since it shows to be the most fast real-time STT (Speech-To-Text). When the user input voice goes in, the Deepgram will process it into a context, send it to Exa ai (optional step if needed) for undertanding certain context/topics, if not will skip to the NLP & Response Generation (Claude 3.5 or 4.0 if available) since it is strong at understanding complex instruction and multiple context at once, but also gives a clear/simple answer which I believe it can lead the converation smooth but deeply. When Claude finishes process, it will be sent to a TTS (Text-To-Speech) nari Dia ai as an output which has the most realistic emotional voice ai service I found so far. Other features are depended on the User Interface, so we'll see how that goes later.
Challenges we ran into
For now on the early stage, I'm have a hard time integrating all of these AI platform API into this service since I'm a intro level developer, has 0 experience on app or web development, yet decided to challenge since it's worth the work if I can motivate one more person out the world being lost as I was back then. So basically, everything is new to me. I'm here to challenge, not do something easy.
Accomplishments that we're proud of
I done a bit of testing through creating a UI design on a paper sheet, testing the UX with colleages or random person around college campus or my neighborhood which almost everyone single one had walkaway with one goal they had but didn't know how to do so. They went though couple paper of the MVP and
What we learned
Developing my Clarté AI chatbot has been a rewarding yet challenging journey, and I’ve gained valuable insights along the way. I started by setting up an AI agent workflow in Activepieces, beginning with a Telegram bot to manage user inputs, which marked my initial success. I then explored the Anthropic API with Claude Opus 4 to generate reflective questions, but I encountered issues like credit balance errors and invalid API versions. Through troubleshooting, I adjusted the headers and body configurations, which helped me overcome those hurdles. There were several moments I faced a deadend, yet consistently kept solving it by not giving up.
What's next for LifeQuest
I have focused on individuals, especially 17~25 y/o male audiences, yet planing to expand for both gender (or more) and scale from individual goals to Team/Organization goals such as Group Projects/Events/Research and more that tend to be a long-term work. I believe AI is a virtual teammate like iron man and jarvis. A standford professor Jeremy Utley mentions "AI should be a tool to It will become a tool that reinvents the practice of brainstorming.
Built With
- activepieces
- bolt.new
- deepgram
- elevenlabs
- exa
- firebase
- grok
- javascript
- perplexity
- telegram
Log in or sign up for Devpost to join the conversation.