AutoFlow - AI Ops Manager for MSMEs

AutoFlow OSS - FOSS Hack

An open-source, local-first AI automation engine for Indian MSMEs

Talk to your business in plain language. Let local AI run the operations. Zero cloud. Zero API keys. Zero technical knowledge.

🎯 The Problem

India has 63 million MSMEs. 97% use WhatsApp as their primary business tool. Yet 86% have zero digital operations system no CRM, no inventory automation, no customer follow-up. Every day, shop owners manually handle hundreds of WhatsApp messages, track stock in notebooks, chase udhaar payments by memory, and lose customers to Blinkit and Zepto because they can't respond fast enough.

Existing automation tools like Zapier, n8n, and Make require APIs, technical knowledge, and cloud subscriptions that small business owners simply don't have. Over 2 lakh kirana stores shut down in 2024 alone not because of bad business, but because operations became unmanageable.

✅ The Solution

AutoFlow OSS lets a business owner type:

"When someone asks about stock on WhatsApp, check my inventory and reply automatically"

And in seconds a complete, working automation is built, visualised, and deployed. Entirely on their own machine. No internet required after setup.

✨ Key Features

🗣 Natural Language to Workflow Type your automation need in plain English or Hindi. Local AI (Ollama + qwen3:8b) parses intent and builds the complete workflow graph triggers, conditions, actions - automatically. No coding. No drag-and-drop required.

🧠 AI Explains Every Workflow After building, the AI explains in plain language exactly what the automation does and why. Full transparency - the owner understands every step before activating. Never seen in Zapier or n8n.

📊 Visual Drag-and-Drop Editor React Flow powered canvas. See what the AI built, modify any node, add conditions, reconnect logic. Full visual control after AI generation.

📦 Inventory-Aware Agents Connect your inventory CSV directly. Automations check stock levels, trigger reorder alerts, and reply "out of stock" automatically - no Google Sheets API, no cloud dependency.

📱 WhatsApp Local Bridge Uses whatsapp-web.js (Apache 2.0, open source) - no Meta API keys, no paid tier. Owner scans QR once. Automation responds to customers on WhatsApp instantly.

📋 Template Marketplace 5 pre-built MSME workflow templates - stock query, udhaar reminder, payment follow-up, order confirmation, customer broadcast. Fork, customise, deploy in under 2 minutes.

📜 Full Audit Log Every automation execution logged locally - what triggered it, what it decided, what it sent, success or failure. Complete transparency and control.

👥 Target Users

Small retail shops, restaurants, clinics, salons, tutors, freelancers, repair services - any MSME owner managing daily customer queries, orders, or payments on WhatsApp without a technical team.

🔧 How It Works - 3 Phases

Phase 1 - Natural Language Input Owner types their automation need in plain English. No forms, no dropdowns, no technical setup.

Phase 2 - Local AI Builds the Workflow Ollama (qwen3:8b) running entirely on the local machine parses intent, extracts triggers, conditions, and actions, and assembles a complete node graph automatically. The AI then explains what it built in plain language.

Phase 3 - Activate and Run One click automation goes live. WhatsApp bridge activates. Every incoming message is checked against active workflows. Matching messages trigger the full execution pipeline automatically.

⚙️ Tech Stack

Layer Technology License AI Engine Ollama + qwen3:8b MIT / Apache 2.0 Backend API Python + FastAPI MIT Workflow Executor Custom Python DAG engine MIT Database SQLite + SQLModel Public Domain Frontend React + Vite + Zustand MIT Node Editor React Flow MIT Messaging Bridge whatsapp-web.js (Node.js) Apache 2.0 Deployment Docker Compose Apache 2.0

Every component is free and open source. No cloud services. No API keys. No vendor lock-in.

🔓 FOSS Compliance

AutoFlow OSS has zero proprietary dependencies. The entire system runs on a single docker-compose up command on any machine with Docker installed.

The messaging layer uses whatsapp-web.js a community open source library requiring no API keys or Meta developer accounts. The core workflow engine, AI parser, executor, and audit log function completely independently of any messaging transport. WhatsApp can be replaced with Telegram, email, or any webhook with no changes to the core.

AI Attribution: Portions of this codebase were developed with AI assistance (Claude by Anthropic, Cursor IDE). All generated code has been reviewed, understood, and tested by the team.

License: MIT

🚀 Quick Start

git clone https://github.com/SIBAM890/AutoFlow
cd autoflow-oss
docker-compose up --build

• Frontend: http://localhost:3000

• API Docs: http://localhost:8000/docs

• WhatsApp Bridge: http://localhost:3001/status

👨‍💻 Team Nexis

Sri Sri University - FOSS Hack 2026 - KLH Localhost