Otto.ai

Otto, your personalised AI therapist available at anytime, through SMS.
Otto uses an Airtable Database to store chat logs. These logs are used to derive analytical insights into user behaviour and chat summaries.
Upon invoking the /e command, the user will get an email summary of past journal entries. Can be used to generalize user behaviour trends.
The /j command makes journaling, easier than ever. With Otto, write reflective entries and receive helpful follow up questions.

Inspiration

When Hetav was in grade 7, Hetav’s best friend lost their grandmother due the covid pandemic. He was grieving greatly and he needed someone to talk to. Due to cultural influences and societal stigma surronding mental care support, Hetav’s friend did not opt for a therapist. Even if he wanted to, these services were in shortage and had long waitlists.

What it does

Otto is a personalised therapist that is available through SMS anytime. He is approachable, empathetic and shares authentic, heart-felt conversations with those who need it the most. Otto breaks the stigma around mental care support by seamlessly integrating into youth culture through texting, a common comfort zone for students.

He has many features such as authentic conversation; journaling, a popular mental health briefing strategy; a summary feature, offering the user an opportunity to step back and reflect on all the past conversations; an analytical tool, providing a way for professional therapists, doctors and parents to gain insights into user behaviour; and an emaling feature to email journal summaries.

Target Audience : Students/Youth (However, all ages are able to use Otto)

How we built it

Otto was developed with Flask (Python), LangChain, FlowiseAI, OPENAI API, the Airtable Database, Twilio, Resend API and ngrok. We utilized Flowise, LangChain and OPENAI API to build the logic behind all of Otto. The three different API endpoints helped us create the different commands for journaling, summarizing, analytics and emailing. Airtable aided us with storing all the journal entries. With the help of Flask we were able to differentiate between commands and normal therapist otto without any commands. Twilio was used to setup the SMS and Resend API was used to setup the emailing. ngrok was used to put the localhost server on the internet. Otto's SMS interface worked on a $15 Twilio Free Trial Account; we were only able to add 2 phone numbers for this project.

Challenges we ran into

implementing ngrok and establishing a network server to run Twillio on
LLM buffer memory with LangChain to maintain past conversation message context
Invalid Resend & Twillio API key configurations
making the perfect prompt for ChatGPT (prompt engineering)
switching logic and functions to handle different edge cases of user behaviour (e.g. switching between /j and normal conversations)

Accomplishments that we're proud of

We implemented different commands such as /j (for journaling), /s (summary), /a (analytics) and /e (emailing).
Otto's communicative interface is natural, authentic and very human-like. With the help of prompt engineering, we were able to achieve this goal.
We were able to complete the project in time.

What we learned

Learned to use FlowiseAI to make API endpoints with low code. Implemented Twilio SMS in Python for the first time and learned create and read operations on Airtable. Additionally, we discovered the massive problem of cultural influences and societal stigma surronding mental care support.

What's next for otto

We hope to make otto available to the masses. It can really help solve stigma surrounding mental health and that is why we hope to bring awarness about it through marketing. We'll try to seek financial assistance to grow Otto and make it viable solution in the frontier of AI+Mental Health