ZigsawAi

Use AI Models in a Node-based, No code platform to create AI powered projects and workflows.

Comment 2

Inspiration

The inspiration behind ZigsawAI was the realization that there are many niche problems in various industries that could be solved with AI, but these solutions often require the integration of multiple models and technologies. However, creating such solutions can be challenging, as they require extensive coding and a disjointed workflows. We wanted to make it easy for anyone to access and experiment with different AI models and tools, regardless of their technical expertise. That's why we created ZigsawAI - a No-code platform that allows users to easily connect and experiment with different AI models, empowering them to create innovative solutions and drive the AI revolution.

What it does

ZigsawAI allows users to treat AI models as building blocks. Users can construct workflows using these AI building blocks on a Node-based interface, connecting inputs (text, sound, etc) to AI models that process it. The system then allows you to chain the output of one model to another model as an input for solving a problem, just like a jigsaw puzzle.

How we built it

One of the core functionalities of our project was the Node based interface. To tackle this task in a limited time we used ReactJs and ReactFlow to create the frontend interface. Our backend is powered by FastAPI which uses the API enpoints from AssemblyAI, Huggingface, OpenAI to process the user inputs.

Challenges we ran into

1) We had a global team with members based out of India, Spain, Germany and US. Although it helped us work round the clock on the project, there were comedic conflicts between git commits and 4 sleep deprived developers. 2) Also because we wanted to save time and work fast, simultaneously working on mac, windows and linux, there were several dependency issues which ate away quite a chunk of time. 3) Deploying to the cloud took some time and had some issues with SSL / CORS,** our webapp works fine on Google chrome but there are still some issues on Microsoft edge **, and some of the formatting of data between UI and api.

Accomplishments that we're proud of

We are amazed to see a single interface running so many AI models. Our goal of enabling anyone to harness the power of AI to make their creativity real is a few steps closer. The Node based interface is not only intuitive but easy to use and understand. Supporting multiple AI workflows in single pipeline makes it even more powerful It was really awesome to see the team get together and make our ideas real. We just met a few days ago, forming friendships while hacking was surreal.

What we learned

Simple is always better and even though a problem seems very daunting, it can be broken down into smaller parts and conquered. Time was very crucial to us to instead of finesse and complicated models, we chose to keep it simple and solve the larger problem we set out to solve. To do that we learnt to manage our time, split our tasks and help each other if we were stuck.

What's next for PlugNPlay AI

There many ideas we have in mind: 1) AI workflow templates for specific tasks (like Speech => Text => Image) with all the blocks already in place for the user to use. 2) Integrations with whatsapp, discord, google sheets, zapier etc. The users can message an input and we will run the workflow in the backend and send them the output. 3) Even more models and blocks for users to use. Video input, stl outputs etc for creators.

Built With

Share this project:

Updates