Spheres – Life Simplified
Here’s the feedback from the client about this project. See Client Feedback
How it Works
The AI driven life manager app that actually knows you.
Built using Flutter, Firebase & OpenAI (with function calling using GPT4.1 model)
— Effortlessly stay on top of work, family, projects and personal goals. All with the help of a personal assistant in your pocket
— The more it gets to know you, better the results get. It automatically creates a tasks or a meeting list based on your conversation with the app, generates missing data using AI with the percentage of accuracy that the user can accept or reject.
— Helps with the AI-based reminders from the tasks and meetings. The user can say something like I have a deadline on Spheres project tomorrow at the day end and that they want to get reminders every 3 hours. It automatically sets those reminders every 3 hour until the deadline. AI or the users themselves can set different reminder types i.e. once, repeating (i.e. daily, weekly, twice a week, only on a specific day, every 15 minutes etc) and nudge
(reminder every hour, reminder only in the morning and night or a reminder 30 minutes before the task due date).
— Extensive predefined and AI-based onboarding to know you better.
— Extensive in-app and AI-based search across different topics.
— Function calling to make the app knowledgeable about the existing internal functionality
The Challenge:
A lot of productivity apps these days are not smart enough to understand the context and reasoning of tasks along with the backing of user’s data. The more you know about the user, the better decisions making you can do when presenting him the options. The challenge was developing something insmall iterations and incorporating user’s feedback to make corrections along the way to build this relatively complex app.
- Productivity chatbot apps don’t value what they know about the user enough (their
behaviours, interests, liking/disliking, preferences, hobbies and choices etc) - Smart auto creation of missing fields (like description, reminders, time estimations,
complexity etc) while creating tasks based on the context - AI-based search and planning the user’s day based on his past interactions. For example, we know this user likes to workout on specific days at a fixed time or they go out on dates weekly on a Friday evening, the app creates an entry in their schedule with the reminders.
- There can only be a few onboarding questions to understand the user, right? Well not
entirely. With out AI-based onboarding questions generator, it keeps on generating questions
from the prior user’s data to understand him better. - The architecture supports function or tool calling which follows a similar MCP protocol. The app is aware of all the functionality that exists in the app and calls out those functions based on the user’s input or prompt.
The Solution:
We built this innovative yet very complicated app called Spheres – a cross-platform solution thatenables users to:
— Build their personal profiles, go through the onboarding process with predefined questions and smart AI-based questions to follow the predefined ones.
— Build their personal profiles, go through the onboarding process with predefined questions and smart AI-based questions to follow the predefined ones.
–AI-based reminders (with a lot of customizations i.e. once, repeating (every x minutes or hours, different times of the day, daily, weekly, on specified days of the week, on a specific date of the month or year etc. and letting users make changes if they want)
— AI-based schedule the user’s day based on his preference (for example, if it’s a work day for the user, the schedule generated for that day would be based on his work or professional life and if it’s an off day of the user, the activities included would be based on his interests for off days etc)
The following attachment shows all of the onboarding process discussed above. Majority of the components in the predefined questions have options that the user can choose from. The AI-questions that come after these are mainly open-ended question giving users the ability to say anything they like.
The following screen shows the custom question types like activities. User can create unlimited activities, set their days of the week and time that they do it on. There are atleast 10 different custom question components
The following attachment shows the glimpses of the most complex part of this application which was to generate the AI-based reminders using a whole lot of these options. This results in user’s getting the push notifications at the specified time.
The following attachment shows how we extract the tasks or the meeting information from the chat in the first image on top left and let user edit any information they want before storing it. The second image shows the listing down of all the tasks that are active or completed of the logged-in user. Third image indicates the level of filtering in the task.
The following screen dives into editing capability of the task’s title & time estimation. By default, the app creates the estimations based on the AI along with the description, category and sub category.
The following screens shows different areas of the app including the chat on the schedule planning, the filtering of tasks by the categories from the menu directly, and some Q&As about how well the app knows you based on it’s past predictions. This last screen is exactly the reason why the app was created with the goal to know you.
The following attachment shows the scheduling feature with the AI-based chat and some filters to assist on how the generated plan should look like. The user can add or remove any activity through the chat and the app saves multiple versions of the schedule. Once user finalises the plan, they can set it through the “Set Schedule” button which brings up a dialog listing all the generated versions and letting the user pick the final one. 