qp

A multi-tenant AI chat platform that hosts specialized AI assistants, each accessible via different subdomains. Built with React, TypeScript, and Vite on the frontend, with a Next.js API backend.

Chat Assistants

This repository supports nine specialized chat applications:

1. stan-assistant — https://stan-assistant.neurosift.app/

Technical assistant for Stan probabilistic programming. Provides guidance and code examples based on the Stan User's Guide and Reference Manual.

2. nwb-assistant — https://nwb-assistant.neurosift.app/

Assistant for Neurodata Without Borders (NWB). Helps with PyNWB and related tools. Includes Jupyter integration for running Python code.

3. neurosift-chat — https://chat.neurosift.app/

Explores and queries neuroscience datasets across DANDI Archive, OpenNeuro, and EBRAINS repositories. Features dataset search, NWB file analysis, and data visualization capabilities.

4. dandiset-explorer — https://dandiset-explorer.neurosift.app/

Specialized assistant for exploring specific Dandisets with Python/Jupyter integration.

5. test-chat — https://test-chat.neurosift.app/

Development testing assistant with Jupyter support.

6. figpack-assistant — https://figpack-assistant.neurosift.app/

Assistant for Figpack.

7. bids-assistant — https://bids-assistant.neurosift.app/

Assistant for the Brain Imaging Data Structure (BIDS) specification. Helps with organizing neuroimaging data according to BIDS standards.

8. hed-assistant — https://hed-assistant.neurosift.app/

Assistant for Hierarchical Event Descriptors (HED). Provides guidance on annotating events and data using HED tags, based on the HED specification and resources documentation.

9. repronim-assistant — https://repronim-assistant.neurosift.app/

Assistant for ReproNim (Reproducible Neuroimaging). Helps with reproducibility tools and best practices including HeuDiConv, DataLad, Neurodocker, and related projects for data conversion, version control, and containerization.

Architecture

• Frontend: React + Vite + TypeScript single-page application
• Backend: Next.js API (nextjs/qp-api/) handling chat persistence and LLM requests
• Routing: Subdomain-based assistant selection (e.g., stan-assistant.vercel.app)
• LLM Integration: OpenRouter for model flexibility
• Storage: MongoDB for chat history
• Tools: Custom tools per assistant (document retrieval, Python execution, dataset queries)

Development

# Frontend
npm install
npm run dev

# Backend API
cd nextjs/qp-api
npm install
npm run dev