-
Project poster
Inspiration
Understanding agents is hard, thousands of employees use agents daily without knowing what they are actually doing. Our aim is to provide a live visual interpretation of the agent's reasoning process, eliminating the blind reliance on AI.
What it does
Our platform visualises and manages AI reasoning workflows in real time, allowing users to explore, modify, and regenerate multi-agent decision-making processes with full traceability and audit logs.
How we built it
Our project is a real-time, multi-agent AI reasoning platform built on a modern tech stack:
Frontend: Next.js, React, ReactFlow, Tailwind CSS, Framer Motion, React-Markdown, Shadcn
Backend: Python, FastAPI, Valyu, Holistic AI, LangChain, LangSmith
Challenges we ran into
Ensuring communication between the backend and the frontend, parsing the output from the backend and converting it into an understandable JSON format for the frontend, creating interactive visual interpretations and implementing features to modify and disable a node.
Accomplishments that we're proud of
We built a fully interactive AI transparency and reasoning platform with real-time graph visualization, and the ability to give the user full control, which we see can be really useful for enterprise teams and organisations. We are proud to build a frontend that is both understandable and comprehensive, and integrating it seamlessly with our backend
What we learned
We learned how to coordinate multiple AI agents, handle complex data flows, and make their decision-making understandable and interactive for users.
What's next
Enable collaborative multi-user features, optimise performance for large graphs, and add advanced node editing and export/import capabilities.
Log in or sign up for Devpost to join the conversation.