Product case study · 2026

AI Sales Agent | Facebook Messenger & Store Website

An Arabic-language sales agent deployed for a real store. It answers in Jordanian Arabic, recommends catalog products, recognizes product photos, and collects order details.

318 automated tests, GitHub Actions CI, and Docker / Render deployment.

Try the test admin account

Use these credentials in the store admin sign-in.

Email
user@test.com
Password
123456
AI Sales Agent | Facebook Messenger & Store Website feature screenshot
AI Sales Agent | Facebook Messenger & Store Website interface overview
( 01 )
About Project

An Arabic-language sales agent deployed for a real store. It answers in Jordanian Arabic, recommends catalog products, recognizes product photos, and collects order details.

Built with Python, FastAPI, Supabase, PostgreSQL, Gemini API, Jinja2.

The project includes jordanian arabic customer conversations.

The project includes catalog product recommendations and photo recognition.

The project includes deterministic jod pricing and order state handling.

The project includes arabic rtl storefront and admin dashboard.

The project includes hmac-sha256 webhook verification, csrf protection and rate limiting.

The project includes prompt-injection tests for llm guardrails.

The project includes telegram daily order report and bot simulator.

( 02 )
Tech Stack & Features

Technology Stack

Python, FastAPI, Supabase, PostgreSQL, Gemini API, Jinja2

PythonFastAPISupabasePostgreSQLGemini APIJinja2

Key Features

The core capabilities that shape this project's experience and workflow.

01

Jordanian Arabic customer conversations

02

Catalog product recommendations and photo recognition

03

Deterministic JOD pricing and order state handling

04

Arabic RTL storefront and admin dashboard

05

HMAC-SHA256 webhook verification, CSRF protection and rate limiting

06

Prompt-injection tests for LLM guardrails

07

Telegram daily order report and bot simulator

( 03 )

User Interface Showcase

A selection of key interfaces from the project, showcasing core user workflows, essential features, and responsive design.

01

Arabic Storefront

The public RTL storefront introduces Hajj and Umrah gifts and Messenger ordering.

02

Product Catalog

Customers can browse product cards and start a Messenger inquiry from the live store.

03

Admin Dashboard

The dashboard brings bot status, orders, and store actions into one view.

04

Product Management

Store admins can maintain product information, prices, categories, and availability.

05

Telegram Reports

Daily Telegram report delivery is visible and manageable from the admin area.

( 04 )

Project Workflow

Customer messages are validated and routed through deterministic order logic, with the store catalog and order data managed through Supabase.

Let's Work Together

Get in Touch

Let's talk about everything!

Have a project in mind, need full-stack or mobile development, or looking for an AI integration? Drop me a message and let's build something exceptional.