About this project
Description from the program archive
VCell-AI is an AI-powered platform for discovering, analyzing, and exploring biological models from the VCell database. While the platform already provides a functional chatbot interface, it requires additional capabilities to support real-world usage — including authentication, usage control, and broader integration with scientific data sources.
This project aims to extend the system across these dimensions. Auth0 authentication will be integrated using the existing VCell Auth0 tenant, allowing existing VCell users to log in seamlessly. On top of this, a per-user token tracking and enforcement system will be introduced, along with an admin dashboard for managing limits.
The prompt handling layer will be refactored into a centralized registry to improve consistency and reduce fragmentation. Local LLM support will be exposed as a user-facing feature. The system will also be expanded with PubMed search, complementing the already implemented BioModels DB integration.
To improve usability, BNGL-based molecular visualization will be integrated. Support for SBML data will be completed, and BioPAX pathway data will be introduced as an additional source for biological knowledge.
Prior to the coding period, 8 pull requests have already been raised covering API standardization, local LLM bug fixes, and BNGL visualization integration as a proof of concept, demonstrating familiarity with the codebase and commitment to the project.