MOOT AI (Vercel)

MOOT AI debate practice tool.

Deploy

1. Set environment variables on Vercel:

   • OPENAI_API_KEY
   • VECTOR_STORE_ID (optional, but required if you want file_search grounding)

2. vercel to deploy, or push to a Git repo connected to Vercel.

Endpoints

• POST /api/send-message — chat via Responses API (+ TTS)
• POST /api/transcribe — uploads webm audio and transcribes with Whisper (whisper-1)
• POST /api/upload-document — uploads a file and attaches it to the Vector Store
• GET /api/vector-store — lists files currently attached to the Vector Store

Notes

Deployment & Environment

Set these environment variables (locally in .env or in Vercel project settings):

• OPENAI_API_KEY (required)
• VECTOR_STORE_ID (optional, needed for file search)
• OPENAI_MODEL (optional, defaults to gpt-4o-mini)

The app uses the Responses API (/v1/responses) with input + instructions, and enables file_search via tool_resources.file_search.vector_store_ids.

Vector store files are uploaded via /v1/files then attached using /v1/vector_stores/{id}/files.

CORS is enabled for all API routes to support local + Vercel.