AIEcommerce - Autonomous PC Assembly & Marketplace Pipeline
Architected a distributed system using LangGraph agents to automate the end-to-end process of scraping hardware components, assembling optimized PC builds, and publishing enriched listings to MercadoLibre.
// class lineage
Read this project as a specialized implementation built on a reusable engineering base.
Autonomous Agent Orchestration Engine
A stateful, graph-driven execution runtime that coordinates specialized AI agents through defined workflows, handles failures gracefully, and maintains a single source of truth for downstream consumers.
AIEcommerce - Autonomous PC Assembly & Marketplace Pipeline
A specialized e-commerce agent pipeline that autonomously assembles, validates, enriches, and publishes hardware products at scale.
Inherited Traits
Overrides
The Challenge
Managing a high-volume hardware catalog required manual intervention for compatibility checks, pricing updates, and SEO-optimized descriptions, leading to significant operational bottlenecks.
The data source (Tecnomega) provided raw PDF and web data that lacked the structured attributes necessary for high-conversion listings on platforms like MercadoLibre.
Synchronizing stock and volatile pricing across a decoupled architecture (Django backend + MedusaJS storefront) while maintaining a multi-agent assembly logic was a complex orchestration hurdle.
The Solution
I designed a multi-agent orchestration layer using LangGraph that treats PC assembly as a stateful graph. Agents act as specialized roles: an Inventory Architect audits stock, a Bundle Creator validates component compatibility (CPU sockets, RAM types, Power requirements), and a Creative Director generates marketing assets.
The system utilizes a 'Sentinel' pattern for health monitoring and a centralized product 'Master' record in Django that acts as the single source of truth for the MedusaJS store and external marketplace synchronizers.
Autonomous PC Builder Agent
LangGraph-powered logic that selects compatible parts from the database to create high-margin computer 'Towers' based on real-time stock.
AI Content Enrichment
Automated pipeline that extracts technical specs from raw text/PDFs and uses LLMs to generate SEO-optimized titles and descriptions for MercadoLibre.
Cross-Platform Sync Engine
Robust synchronization service that manages inventory state between the internal Django DB, the MedusaJS storefront, and MercadoLibre API.
Intelligent Image Upscaling
Integrated processing pipeline that fetches low-res supplier images and enhances them for high-quality marketplace display.
Technical Implementation
The architecture follows a modular service-oriented pattern. The 'Agents' service communicates with the 'Core' Django API via structured Pydantic schemas. LangGraph manages the state of assembly tasks, ensuring that if a component goes out of stock during the process, the graph self-heals by selecting an alternative.
# Example of the Inventory Architect Node in LangGraph
async def inventory_architect_node(state: AgentState):
# Audit available components for a specific category
components = await aiecommerce_service.get_available_components(
category=state["target_category"],
min_stock=5
)
if not components:
return {"status": "error", "error_message": "Insufficient stock"}
return {
"available_components": components,
"next_step": "validate_compatibility"
}Results & Impact
The transformation from manual entry to an agentic pipeline allowed for the rapid expansion of the product catalog. The system now autonomously maintains thousands of listings, ensuring that every PC 'Tower' offered is technically compatible, profitably priced, and marketed with AI-generated precision.