Skip to content

crunchydosa123/ProjectKaarigar

Repository files navigation

Project Kaarigar:

PLEASE USE GUEST ACCOUNT IN THIS WEBSITE

DEMO WEBSITE

https://frontend-new-557742533869.asia-south1.run.app/

📘 Index

Some Points to be considered

ALERT: FALL BACK DEMO WEBSITE LINK(PLEASE USE THIS LINK FOR TESTING FRONTEND INCASE THE ONES IN PPT DONT WORK BECAUSE OF DOMAIN ERROR OF GCP)

https://frontend-new-557742533869.asia-south1.run.app/ PLEASE use guest account 1,2

In case guest accounts dont work use these Mock accounts : username: surajchavan99886@gmail.com pass:123456

Mock account2: username: raju.deo@gmail.com pass:123456

Problem statement:

image

Our Team:

image

Video Demo

Watch the demo

Camera Agent Walkthrough

Please take a look at this video in case u missed out to get a detailed overview of how our specialized camera agent helps artisans navigate complex photo and video shoots with ease and get the best possible results.

Video Link - Watch the demo

Whatsapp Bot Walkthrough

Please take a look at this video to get an overview of the user side for our personalized Whatsapp Bot.

Video Link - Watch the demo


ALERT: FALL BACK DEMO WEBSITE LINK(PLEASE USE THIS LINK FOR TESTING FRONTEND INCASE THE ONES IN PPT DONT WORK BECAUSE OF DOMAIN ERROR OF GCP)

https://frontend-new-557742533869.asia-south1.run.app/

Introduction:

Local artisans possess incredible skill and create unique products, but often face significant hurdles in navigating the complex digital world.
From building an online brand and creating engaging social media content to optimizing product listings and analyzing sales, the technical barriers can be overwhelming.

Project Karigar is an end-to-end solution that acts as a digital companion for these artisans.
We replace complex forms and software with an intuitive, conversational AI that understands their language and needs.

Our platform automates:

  • Content creation
  • Marketing
  • Marketplace listings
  • Intelligent insights

…allowing artisans to focus on what they do best: creating.


Product Summary

image

Who is it for?

Project Kaarigar is built for:

  • Artisans (handcrafted makers, solo creators)
  • Small businesses (micro-brands wanting to go online)
  • Homemakers / Hobbyists who want to sell their creations

Mobile-first, multilingual, and designed for users with minimal technical experience.


Areas of User Impact

  • Original, consistent brand identity — conversational onboarding builds a polished brand persona without forms.
  • Optimized marketplace listings — SEO-ready titles & descriptions tailored to each platform.
  • Improved digital content quality — AI-generated product photos, backdrops and marketing videos.
  • Data simplified into actionable insights — sales & trend analytics converted into pricing, content and growth recommendations.

What does Project Kaarigar do today

Conversational Brand Onboarding

Streamlines artisan onboarding via a conversational interview that generates a personalized brand document and profile, ensuring consistent branding across the app and listings.

AI Content Generation & Natural Language Editing Support

Generates platform-specific AI photos and videos for product shoots and marketing.
Supports natural-language editing commands (e.g. “shorten this video and add retro music”) executed via our FFmpeg + Gemini modules to produce ready-to-post media.

AI-based SEO-optimized Listing with Marketplace Integrations

Transforms product images and descriptions into high-ranking, SEO-optimized listings, then automatically posts them to multiple marketplaces and channels (WhatsApp, Amazon, Flipkart, Instagram, etc.).

Multi-channel Commerce & Social Media Engine

Analyzes social and marketplace trends to suggest content, automates posting and ad campaigns, and manages direct sales & payments using an integrated WhatsApp bot and other commerce connectors.

Direct Social Media Sales & Order Management with Agentic Insights

Converts sales and engagement data into smart pricing advice and data-driven content suggestions to increase conversions and revenue.
Provides end-to-end support from discovery → listing → sale → fulfillment.


How we differ from existing solutions:

image

As the comparison shows, the ecosystem today is fragmented — marketplaces, social platforms, and editing tools each provide pieces of the value chain.

Project Kaarigar combines the missing pieces into a single productised flow, adding conversational onboarding, end-to-end marketplace listing + outreach, agentic insights, and natural-language media generation.


Unique End-to-End Onboarding

Project Kaarigar is the only solution in this comparison that provides conversational onboarding that outputs a consistent brand document and profile.

Unified Media Workflow

We combine AI Camera Assist, Veo/FFmpeg natural-language editing, and Gemini-driven content ideation so artisans can capture, generate, and edit platform-ready media without switching apps.

Marketplace-First Distribution

While some platforms may offer bits of media tooling or analytics, Project Kaarigar is the only product that automates SEO-ready listing generation, posting, and outreach, closing the loop from creation → listing → sale.

Agentic Insights

We turn sales and trend signals into actionable, artisan-specific recommendations (pricing, content, promotions) powered by Gemini.


Project Kaarigar provides support to artisans across marketplaces, social channels, and editing workflows, enabling even the smallest businesses to reach customers with professional media, optimized listings, and data-driven growth.

In short: We replace complexity with a single, intelligent assistant , designed exclusively for the artisan.


List Of Features:

image

Architecture

image

Use Case Diagram:

image

Google Cloud & AI Tools Usage

image

Project Kaarigar leverages a suite of Google Cloud services and generative AI models to power an end-to-end pipeline for onboarding, media generation, listing creation, analytics, and distribution.


Core Platform Services

  • Cloud Functions — Event-driven glue that automates preprocessing, content generation, and posting via serverless, asynchronous workflows.
  • Cloud Run — Runs decoupled microservices for AI inference, video generation, and trend analysis with scalable reliability.
  • Firestore — Stores artisan, product, and brand data with integrity and enables semantic search using Vertex AI embeddings.
  • Cloud Storage — Hosts all media assets (images, videos, creatives) for scalable access, analytics, and real-time content delivery.

Generative AI Models & Media Tooling

  • Gemini 2.5 Flash-Lite — Conversationally generates brand documents, product descriptions, captions, and short marketing content from artisan inputs.
  • Gemini 2.5 Pro — Produces long-form creative and video-ready scripts from brand documents and campaign ideas.
  • Gemini Audio Native Dialog — Powers voice-based onboarding and natural spoken interactions for multilingual artisan experiences.
  • Google Veo 3 — Creates and enhances AI-driven videos of artisan products and marketing shoots.
  • Imagen 4 — Generates and edits high-quality product photos for consistent visual branding and marketplace listings.
  • Google Cloud Text-to-Speech — Converts generated or conversational text into natural voice for audio content and hands-free interaction.

Intelligence & Personalization

  • Vertex AI (Embeddings + AI Services) — Provides semantic understanding, recommendations, and personalization through embeddings and model APIs that enable RAG, similarity search, and tailored suggestions.

Tech Stack Deep Dive:

image

Our stack is now fully serverless + container-first on Cloud Run, with Firestore + Cloud Storage as the primary data layer and a set of Cloud Run microservices for each AI capability.


🖥️ Frontend

  • React — Mobile-first, component-driven UI.
  • Tailwind CSS — Fast, consistent styling.
  • Hosting: Frontend runs as a container on Cloud Run.

⚙️ Backend & Media

  • Python (+ Flask / microservices) — Lightweight APIs and orchestration.
  • FFmpeg — Media encoding, trimming, and edits driven by natural language commands (deployed inside the video/image editing Cloud Run service).
  • Container Image Registry: Artifact Registry for storing Docker images.

🗂️ Database & Storage

  • Google Firestore — Stores artisan, product, and brand metadata; used for fast lookups and user/profile data.
  • Google Cloud Storage — Hosts all media assets (images, videos, creatives) for scalable delivery and processing.

☁️ Deployment & Hosting

  • Cloud Run — Primary runtime for decoupled microservices:
    • Conversational Agent
    • Camera Agent
    • Video/Image Editing
    • Listings Generator
    • AI Analytics
    • WhatsApp Campaign
    • Frontend
      Each service runs as a container for easy CI/CD and independent scaling.
  • CI/CD — GitHub (or equivalent) → Artifact RegistryCloud Run for rolling updates and automated deployments.

🤖 Google AI Tools (Vertex AI + Models)

  • Gemini 2.5 Flash-Lite — Conversational generation: brand docs, short captions, product descriptions, and on-the-fly content edits.
  • Gemini 2.5 Pro — Long-form creative, campaign ideas, and video-ready scripts.
  • Gemini Audio / Native Dialog — Voice-based onboarding and multilingual spoken interactions.
  • Google Veo 3 — AI video generation & enhancement for product shoots and promos.
  • Imagen 4 — High-quality product photo generation and editing for consistent marketplace visuals.
  • Google Cloud Text-to-Speech — Converts generated text into natural voice for camera assist and hands-free UX.
  • Vertex AI (Embeddings + Services) — Semantic understanding, RAG workflows, personalization, and recommendation pipelines.

Hosting and Rolling Out Updates

Multiple Cloud Run services each handle a focused responsibility:

  • Conversational Agent — Gemini-powered onboarding.
  • Camera Agent — Voice + camera instructions / capture assist.
  • Video & Image Editing — FFmpeg + Veo + Imagen workflows.
  • AI Analytics — Trend analysis and pricing signal generation.
  • Listings Generation — SEO titles, descriptions, and marketplace integrations.
  • WhatsApp Campaign — Automated outreach and direct-sales flow.

The Frontend is deployed as a Cloud Run container that communicates with backend microservices.
CI/CD pipelines push container images to Artifact Registry and perform rolling updates to Cloud Run services.


User Experience

image ---

Project Kaarigar is designed to make digital commerce effortless for artisans by combining conversational AI, multimodal media tools, and marketplace automation into a simple, mobile-first experience.


AI-Assisted Seamless Journey

Delivers a consistent, guided workflow from conversational onboarding through listing management and marketing, minimizing cognitive load and manual steps for the artisan.

Conversational Platform

Offers an interactive, natural-language platform that uses voice-based dialogs (Gemini Audio / Native Dialog) to simplify complex tasks like branding and sales management.

Multilingual & Responsive Design

Ensures an intuitive, adaptable interface using React and Tailwind CSS that works flawlessly across all devices (mobile-first) and supports multiple regional languages.

Effortless Content Creation

Artisans can generate professional, platform-specific product photos (Imagen 4) and marketing videos (Veo 3) using simple natural-language commands — reducing time and producing consistent creative assets.

Guided Workflow & Market Integration

Provides a structured, step-by-step journey that guides artisans from raw craft to automatically posted, SEO-optimized listings across multiple marketplaces.

Agentic Hands-Free Interaction

Leverages voice-based Camera Assist for natural spoken interactions and in-context content editing, making the digital process accessible even while the artisan is actively working.


Market & Adoption

image

30–90 Day Playbook (Try / Launch / Measure)


Day 0–30 — Test & measure with select users

  • Release the current version of Project Kaarigar to selected audiences for usability testing and feature validation.
  • Run focused usability sessions and in-person workshops to test content creation flows, onboarding, and listing posting.
  • Collect feedback on pain points, UX friction, and most-valued features.

Day 31–60 — Iterate features & scale cloud deployment

  • Prioritize and double down on features users love; remove or rework parts that don’t serve the core value.
  • Harden cloud deployment for scale and cost efficiency (Cloud Run autoscaling, CI/CD, Artifact Registry).
  • Optimize media generation pipelines and monitor AI usage patterns to control costs.

Day 61–90 — Launch app on Play Store

  • Publish the mobile app to Play Store (and App Store if planned).
  • Open broader beta, enable onboarding campaigns, and measure activation → listing → sale funnel metrics.
  • Ramp up partner outreach and community acquisition programs.

First Users — Who & How We’ll Approach Them

Artisans Registered with NGOs

  • Partner with 2–3 key NGOs for direct access and trust.
  • Run in-person workshops to onboard the first ~100 users, gather qualitative feedback, and build advocacy.

Hobbyists / Homemakers Struggling to Sell

  • Target online communities (Facebook, Instagram groups, Reddit) with an exclusive beta offer.
  • Provide a free “Launch Your Shop” guide and lightweight onboarding to drive low-cost digital acquisition.

Women’s Savings & Local Community Groups

  • Work with local community leaders in a target geographic cluster.
  • Run small, trusted group meetings to demonstrate financial benefits and drive local adoption.

Implementation and WireFrames

image

Conclusion

Project Karigar is not just a tool it’s an ecosystem.
We empower artisans by:

  • Preserving authenticity
  • Providing AI-driven content & insights
  • Ensuring scalability and ease of adoption

Crafted in India, scaled for the world 🌍

ALERT: FALL BACK DEMO WEBSITE LINK(PLEASE USE THIS LINK FOR TESTING FRONTEND INCASE THE ONES IN PPT DONT WORK BECAUSE OF DOMAIN ERROR OF GCP)

https://frontend-new-557742533869.asia-south1.run.app/

Please use guest account on the website on pc, non incognito mode for this app to work for the time being.

Releases

No releases published

Packages

 
 
 

Contributors