BMAD-METHOD/bmad/tools/serverless-samples

Serverless samples for the Journaling → AI → LinkedIn MVP

Overview

This folder contains minimal, well-documented serverless function skeletons you can deploy to Vercel/Netlify/AWS Lambda to support the mobile app. They are intentionally small and include placeholders for secrets and minimal error handling so you can iterate quickly.

Included files

• package.json — example dependencies
• .env.example — environment variables to populate
• functions/
  • signed-upload.js — returns signed S3 upload URL (short TTL)
  • transcribe-start.js — starts a transcription job for an uploaded audio file
  • transcribe-worker.js — worker that calls Whisper/OpenAI speech-to-text and returns transcript (invoked internally or via webhook)
  • generate-post.js — proxies sanitized text to OpenAI and returns post variants
  • publish-linkedin.js — performs LinkedIn OAuth callback handling and publishes UGC posts

How to use

1. Copy this folder into your deployment target (Vercel, Netlify, etc.).
2. Populate environment variables from .env.example.
3. Install dependencies (see package.json).
4. Deploy functions. Test with the example request payloads provided in the project documentation.

Security notes

• Never expose your OpenAI/Whisper/LinkedIn client secrets to the client. Keep them in server env vars.
• Signed upload URLs should be short-lived (TTL < 1 hour) and validate file size/content-type.
• Delete audio files immediately after transcription completes.

This is a starter scaffold; adapt to your provider and architecture as needed.