Skip to main content
This repo supports a link-based flow for sending meeting transcripts to ChatGPT without embedding the full transcript in the ChatGPT URL (which can exceed browser / proxy limits). Instead, the Vexa Dashboard calls an authenticated Vexa API endpoint to mint a short-lived public URL. Then ChatGPT is opened with a query like: Read from https://api.your-domain.com/public/transcripts/<share_id>.txt so I can ask questions about it.

​
How it works

  1. Dashboard calls:
    • POST /transcripts/{platform}/{native_meeting_id}/share?meeting_id=<internal_meeting_id>
    • Requires X-API-Key
  2. API Gateway:
    • Fetches the transcript from transcription-collector
    • Formats it into plain text
    • Stores it in Redis under a random share id with a TTL
    • Returns a url like:
      • /public/transcripts/<share_id>.txt
  3. ChatGPT reads the public URL (no auth headers required).

​
Configuration (API Gateway)

Environment variables (optional):
  • PUBLIC_BASE_URL
    • Example: https://api.vexa.ai
    • If unset, the gateway derives the URL from X-Forwarded-Proto and X-Forwarded-Host (recommended to set these in your reverse proxy).
  • TRANSCRIPT_SHARE_TTL_SECONDS
    • Default: 900 (15 minutes)
  • TRANSCRIPT_SHARE_TTL_MAX_SECONDS
    • Default: 86400 (24 hours)

​
Security model

  • The public link is unguessable (random id) and short-lived (TTL in Redis).
  • The transcript is not stored permanently by this feature; it expires automatically.
  • You should still treat transcripts as sensitive data:
    • Keep TTL short (15–60 minutes)
    • Ensure HTTPS
    • Consider limiting request rate on /public/transcripts/*

​
Reverse proxy requirements

ChatGPT must be able to fetch the URL from the public internet. Your reverse proxy should:
  • Terminate TLS (HTTPS)
  • Forward requests to the API Gateway upstream
  • Set:
    • X-Forwarded-Proto
    • X-Forwarded-Host
In your Nginx, expose the API Gateway publicly (HTTPS) and ensure the hostname you want ChatGPT to read from points to the gateway. For this repo’s dev setup we use share.dev.vexa.ai → :8056.