Prasaar / architecture ← Product
Under the hood

How Prasaar keeps creators consistent

Four services on Google Cloud, one event bus, and a learning loop. Pick a flow below and press play to watch a request move through the system.

4deployables (Cloud Run)
7event topics (Pub/Sub)
8end-to-end flows
0servers to manage
The product loop

Plan → Create → Post → Learn → Reward

Each stage feeds the next. Tap a stage to play its flow.

Flow player

Watch a request move through the system

Highlighted nodes are the ones the flow uses. Numbers show the order.

    Event bus

    Everything after "post" is event-driven

    Pub/Sub decouples posting, analytics, coaching and rewards. Every consumer is idempotent: it writes processed/{eventId} first, so a duplicate delivery changes nothing.

    Data model

    Live state in Firestore. Analytics in BigQuery.

    Firestore drives the real-time UI. BigQuery holds every metric snapshot, so baselines, attribution, leagues and journeys run as SQL.

    Scoring, live

    The maths behind the scores

    The numbers are computed in code, and Gemini explains them. Try the sliders.

    Idea Score

    0
    score = 25·demand + 20·(1−competition) + 20·fit + 15·goal + 10·timing + 10·hook

    Post Coach: baseline comparison

    Each post is compared to your own rolling median of the last 10 posts of the same format (black line).

    7-day views
    CTR %
    Retention %
    Gamification engine

    A weekly streak, against your own plan

    Every Monday at 00:00 IST, Cloud Scheduler compares each creator's published posts with their planned cadence. Tap the weeks to try it.

    Kept 🔥cadence met
    ⇄
    Frozen ❄️missed, freeze used
    ⇄
    Brokenno freeze left
    →
    Comeback 💪3 easy quests

    Plan an idea +10Publish +50New platform +20New language +30Act on Coach tip +40Reply to comments +5

    The XP ledger is append-only (xp/{uid}/events/{eventId}), and the level is derived from total XP. If a publish fails because of us, the streak is protected.

    Scale

    Built for 100k creators from day one

    PressureDesign choiceGoogle service
    Failure modes

    What happens when something breaks

    FailsAutomatic behaviourWhat the user sees
    Ship loop

    From laptop to production in minutes