# Zwiim

> Bharat's personal intelligence layer — metered like electricity, pauseable
> like prepaid data, assisted like real service. One chat across
> **Claude Opus 4.7, GPT-5, and Google Gemini 2.5** with India-priced
> **₹49 outcome sachets** or **₹50 / hour passes** that pause and resume.
> Outputs are real files the user keeps: `.pptx`, `.docx`, Markdown notes
> with Anki `.apkg`, DALL·E 3 images, multi-file code project ZIPs. UPI
> checkout via Razorpay. Hindi / Hinglish / regional-language native.

## Positioning

- **For Bharat, not the world.** Defaults ₹ + FY Apr-Mar + Indian companies
  + British Indian spelling. Speaks Hindi / Hinglish / Tamil / Telugu /
  Marathi / Bengali.
- **Metered like electricity.** Pay per outcome (₹49) or by duration
  (₹50/hour with pause-resume). No monthly USD subscription.
- **Pauseable like prepaid data.** Buy an hour, use 30 minutes, pause, come
  back later. None of the big-three AI vendors offer this.
- **Assisted, not just answered.** For high-stakes outputs (career packs,
  business reports), Zwiim pairs the model with human review rounds — trust
  as a product, not a tutorial.

## What Zwiim does

| Capability | How it works |
|---|---|
| Multi-model routing | Each chat turn picks the right vendor (Opus / GPT-5 / Gemini); falls through transparently on provider error |
| Outcome billing | ₹49 sachet = one finished output (deck / doc / notes / image / code / Anki). No subscription. |
| Duration passes | ₹50/hour with pause-resume. ₹200 for six hours, ₹300/day, ₹500/week, ₹750/month unlimited |
| Real artifacts | python-pptx for `.pptx`, python-docx for `.docx`, Markdown+flashcards for notes, DALL·E 3 for PNG, multi-file project ZIP for code, genanki for `.apkg` |
| Vision input | JPEG / PNG / WebP; the model actually sees the image (Opus / GPT-5 / Gemini all multimodal) |
| PDF input | Routes to Anthropic which reads PDFs natively in document content blocks |
| Web search | Brave-powered. Fires automatically on "today / current / latest / price / news" keywords, or on demand via `search: <query>` prefix |
| Persona-aware routing | Onboarding captures persona (student / working pro / aspirant / teacher / freelancer); model gets a reasoning_tier hint (light / standard / deep) |
| Memory | Cross-session pgvector embeddings + true incognito toggle on the profile page |
| Markdown rendering | Assistant replies render headings / lists / code blocks in the chat surface, not as raw `**asterisks**` |

## Pricing (INR)

| SKU | Price | What you get |
|---|---|---|
| Per output | ₹49 | One finished artifact |
| Monthly pack | ₹249 | 30 outputs/month |
| Hour pass | ₹50 | 1 hour unlimited, pause-resume |
| Six-hour pack | ₹200 | 6 hours unlimited |
| Daily | ₹300 | 24 hours from activation |
| Weekly | ₹500 | 7 days unlimited |
| Monthly unlimited | ₹750 | 30 days unlimited |

## Why Zwiim is different from ChatGPT / Claude / Gemini

- One account, three vendors — no separate ChatGPT Plus + Claude Pro + Gemini Advanced
- ₹49 to start instead of ~₹1,700 monthly subscription
- Pause-resume time-window passes — none of the big-three offer this
- Indian context default (₹ / Aadhaar / UPI / Razorpay) instead of bolt-on
- Exportable file formats — Claude artifacts live in a preview pane; Zwiim hands you actual files
- UPI checkout via Razorpay, not USD card-only

## Who Zwiim is for

Students (board projects, revision notes, Anki decks for spaced repetition).
Working professionals (pitch decks, status docs, code prototypes).
Aspirants for UPSC / CAT / NEET / JEE (chapter notes, current-affairs
summaries with citations). Freelancers (proposals, social assets,
deliverables). Teachers (lesson plans, exam prep). Small-business owners
(investor decks, posters, product imagery).

## Live URLs

- Marketing landing: <https://zwiim.com/>
- Signed-in app: <https://zwiim.com/app/>
- About page: <https://zwiim.com/about>
- API: <https://api.zwiim.com/>
- Legacy domain (still works): <https://miitro.strivhq.com/>

## Stack

Django 5.1 + DRF, Postgres + pgvector for memory, Redis + Celery for chat
streaming + background renderers, S3 for artifacts, Razorpay for payments.
Vanilla-JS frontend (no build step). Nine-container Docker stack on a
single t3.medium in ap-south-1. Pre-deploy gate runs a 29-case LLM eval
suite against live API keys before any backend/prompt change.

## Contact

Operator: mohit.bansal@progcap.com