Positive Posture Preserve Me (PPPM)

It is a multimodal memory-to-confidence platform.
The core idea is simple: people often forget their strongest moments right when they need confidence the most. PPPM helps users capture those moments in voice, text, image, and video, then transforms them into an uplifting capsule they can revisit before exams, interviews, presentations, and high-pressure days.

Inspiration

We were inspired by a common pattern: people are capable and resilient, but stress makes them forget their own evidence of progress.
Traditional journaling tools are useful, but often feel passive and text-heavy. We wanted to build something emotionally engaging and immediate: a “confidence replay” experience that feels more like reliving momentum than reading notes.

What it does

PPPM lets users:

create memory capsules from text, voice, images, and videos
chat with AI for clarifying prompts to deepen context
generate interleaved multimodal output (story text + visuals + audio/video segments)
store media locally by default, with optional Google Cloud Storage credentials
run capsule generation asynchronously with job progress/status updates
track generation state in Firestore (with safe in-memory fallback)
use secured auth flows with server-side CAPTCHA checks and verified Google ID tokens

How we built it

We built PPPM as a full-stack TypeScript project:

frontend: React + Vite + Zustand + i18n
backend: Node.js + Express + Prisma
AI layer: Google GenAI SDK for multimodal pipeline orchestration
storage: local filesystem fallback, optional GCS
realtime/state: Firestore-backed generation state service

On the generation side, we moved from a synchronous path to an async job pipeline:

enqueue generation job
update stages (UPLOADING -> UNDERSTANDING -> PLANNING -> GENERATING -> FINALIZING)
compose final multimodal segments
persist capsule and return complete result to UI

Challenges we ran into

balancing cost and reliability: cloud-only flows were expensive early, so we implemented local-first storage fallback
turning generation into a robust async workflow without breaking existing client behavior
hardening auth beyond UI-only checks (server-side CAPTCHA verification and proper Google token signature verification)
handling multimedia assembly constraints across environments (especially video toolchain/runtime restrictions)
keeping user experience smooth while adding security and backend complexity

Accomplishments that we're proud of

delivered a practical local fallback architecture that still supports cloud upgrade paths
implemented a real async generation pipeline with progress/status visibility
integrated multimodal story composition with text, image, audio, and video segments
upgraded security posture with stronger auth verification flows
shipped a git-ready, deployable project structure with clear environment templates

What we learned

We learned that emotional-product UX and backend architecture must evolve together.
If generation quality improves but latency and reliability degrade, trust drops quickly.

We started modeling product trade-offs as:

\[ \text{User Trust} \propto \frac{\text{Quality} \times \text{Reliability}}{\text{Latency} + \text{Friction}} \]

That mindset helped us prioritize async jobs, transparent progress, and fallback paths.