Albetar AI Modules

Conversational bilingual product search experience
Hybrid query intent engine and semantic matching flow
Hardware-aware ranking and recommendation results
Automated review sentiment analysis and scoring
Arabic AI-generated business performance narrative
Query analytics and operational intelligence insights
Revenue forecast and anomaly detection dashboard

Production-grade bilingual AI engine for e-commerce intelligence

AlBetar AI Engine is a production-grade Retrieval-Augmented Generation (RAG) architecture built for Arabic and English e-commerce intelligence.

It powers conversational product search and recommendations through a multi-layer pipeline: hybrid intent parsing, embedding generation via self-hosted microservices, semantic retrieval over PostgreSQL + pgvector, and hardware-aware ranking for CPU, VGA, RAM, and storage.

Core Capabilities

  • Conversational product search with Arabic and English support
  • Hybrid query understanding (keyword engine for speed, LLM routing for complex intent)
  • Automated review sentiment analysis (positive / neutral / negative with confidence scoring)
  • AI business insights and recommendation narratives from real KPI and revenue data
  • Conversational business Q&A in Egyptian Arabic with safe read-only query execution
  • 14-day revenue forecasting with anomaly detection for unusual spikes and drops
  • Embedding caching layer that reduced LLM API cost by approximately 60 to 70 percent
  • Query analytics and session tracking for continuous model and UX improvement

Technology Stack

  • Next.js API Routes
  • TypeScript
  • LLM Providers: Gemini 2.0 Flash, GPT-4o Mini, Grok-4
  • OpenAI GPT-4 Turbo
  • OpenRouter
  • OpenAI embeddings (text-embedding-3-small, text-embedding-3-large)
  • Xenova multilingual-e5-base (self-hosted embeddings)
  • PostgreSQL + pgvector
  • Upstash Redis
  • Express.js microservices
  • Docker
  • Google Cloud Build
  • Recharts

Project Outcome

The platform transformed product discovery and admin decision-making by combining semantic retrieval, bilingual natural language interfaces, and operational intelligence in one system. Superadmins can ask business questions in Arabic, get grounded insights and forecasts, and take action faster with clear AI-assisted recommendations tied directly to live store data.

العودة للرئيسية