MerchFlow-AI / README.md
Gaurav vashistha
Docs: Complete README rewrite for production Glassmorphism UI and google-genai migration
f4ae8bb
metadata
title: MerchFlow AI
emoji: ๐Ÿ‘Ÿ
colorFrom: blue
colorTo: purple
sdk: docker
pinned: false

๐Ÿš€ MerchFlow AI

Autonomous E-Commerce Catalog Intelligence

Python FastAPI Google Gemini Pinecone Docker

Proprietary & Confidential โ€” Bhishaj Technologies


A multi-agent AI pipeline that converts raw product imagery into enterprise-grade, SEO-optimized e-commerce catalogs in seconds.


๐ŸŽฏ Core Value Proposition

MerchFlow AI eliminates the manual bottleneck of product catalog creation. By orchestrating Computer Vision, Retrieval-Augmented Generation (RAG), and Large Language Models in a seamless autonomous pipeline, it delivers production-ready product listings โ€” from a single image upload โ€” with zero human intervention.


๐Ÿ—๏ธ System Architecture

The system employs a high-performance, event-driven architecture orchestrated by FastAPI:

๐Ÿ‘๏ธ Visual Analyst Agent

  • Function: Zero-shot product image analysis
  • Process: Extracts granular visual attributes โ€” dominant colors, material composition, design style, branding elements, and product classification
  • Engine: Gemini 2.5 Flash via the unified Google GenAI SDK (google-genai)

๐Ÿง  Semantic Memory Agent

  • Function: RAG-based keyword retrieval with intelligent fallback
  • Process: Vectorizes visual attributes to query a high-dimensional index, retrieving historically high-converting SEO keywords and market trends. When the database has no match for a niche, the Intelligence Fallback system autonomously generates keywords via Gemini โ€” ensuring 0% empty results
  • Engine: Pinecone Vector DB with gemini-embedding-001 embeddings (768 dimensions)

โœ๏ธ Writer Agent

  • Function: High-conversion copy synthesis
  • Process: Fuses visual intelligence with retrieved market data to generate persuasive, conversion-optimized titles, descriptions, and feature bullet points
  • Engine: Meta Llama 3.3 70B (via Groq Cloud)

โš™๏ธ Pipeline Orchestrator

  • Function: Async pipeline management & delivery
  • Process: Handles non-blocking agent execution, error propagation, and API lifecycle management. Results are delivered instantly through the Premium Glassmorphism UI
  • Engine: FastAPI with async/await architecture

๐Ÿ–ฅ๏ธ Production Interface

MerchFlow AI ships with a Premium Glassmorphism UI built for instant catalog generation:

  • ๐ŸŽจ Frosted-glass aesthetic with dynamic gradient backgrounds
  • ๐Ÿ“ค Drag-and-drop image upload with real-time processing feedback
  • ๐Ÿ“Š Structured JSON output display for visual data, SEO keywords, and generated listings
  • ๐Ÿ“ฑ Fully responsive design across desktop, tablet, and mobile

๐Ÿ› ๏ธ Technology Stack

Layer Technology Purpose
Runtime Python 3.10+ Core language
Framework FastAPI Async API orchestration
AI SDK google-genai (Unified) Vision & embedding inference
Vision Model Gemini 2.5 Flash Product image analysis
Embeddings gemini-embedding-001 768-dim vector generation
Vector DB Pinecone (Serverless) Semantic keyword retrieval
LLM Llama 3.3 70B (Groq) Copywriting synthesis
UI Glassmorphism / Tailwind CSS Production dashboard
Deployment Docker / Hugging Face Spaces Containerized hosting

๐Ÿ“‹ System Updates & Technical Milestones

Date Milestone
Feb 2026 โœ… Full migration from deprecated google-generativeai to unified google-genai SDK
Feb 2026 โœ… Vision model upgraded to Gemini 2.5 Flash for industry-leading latency
Feb 2026 โœ… Pinecone vector alignment to 768 dimensions with gemini-embedding-001 for mathematical precision
Feb 2026 โœ… Intelligence Fallback system deployed โ€” guarantees 0% empty SEO keyword results
Feb 2026 โœ… n8n webhook decoupled โ€” pipeline relies strictly on the Glassmorphism UI for delivery
Feb 2026 โœ… Production Glassmorphism dashboard launched

๐Ÿš€ Quick Start

1. Environment Configuration

Create a .env file in the project root with the following keys:

GEMINI_API_KEY=your_google_genai_api_key
GROQ_API_KEY=your_groq_cloud_api_key
PINECONE_API_KEY=your_pinecone_api_key

2. Installation

pip install -r requirements.txt

3. Launch

python main.py

The production dashboard will be available at http://localhost:7860.


๐Ÿ“ Project Structure

MerchFlow-AI/
โ”œโ”€โ”€ main.py                  # FastAPI orchestrator & pipeline logic
โ”œโ”€โ”€ dashboard.html           # Glassmorphism production UI
โ”œโ”€โ”€ Dockerfile               # Container deployment config
โ”œโ”€โ”€ requirements.txt         # Python dependencies
โ”œโ”€โ”€ agents/
โ”‚   โ”œโ”€โ”€ visual_analyst.py    # Gemini 2.5 Flash vision agent
โ”‚   โ”œโ”€โ”€ memory_agent.py      # Pinecone RAG + embedding agent
โ”‚   โ””โ”€โ”€ writer_agent.py      # Llama 3.3 copywriting agent
โ””โ”€โ”€ .env                     # Environment variables (not tracked)

โš–๏ธ Legal Notice

PROPRIETARY & CONFIDENTIAL

This software, including all source code, documentation, algorithms, and associated intellectual property, is the exclusive proprietary property of Bhishaj Technologies (UDYAM-UP-02-0108589).

Unauthorized copying, distribution, modification, reverse engineering, or any form of reproduction of this software โ€” in whole or in part โ€” is strictly prohibited and may result in legal action under applicable intellectual property laws.

ยฉ 2026 Bhishaj Technologies. All Rights Reserved.


Built with precision by Bhishaj Technologies ๐Ÿ‡ฎ๐Ÿ‡ณ