This content originally appeared on DEV Community and was authored by Chris Yaowen Zhang
Introducing Enthusiast v1.4
We tend to think of AI as operating on clean, structured data. E-commerce teams know better: most of their information comes as PDFs from vendors, spreadsheets with inconsistent formatting, or scanned purchase notes from clients. Embedding these unstructured sources into automated workflows has long been a challenge. With the latest release of Enthusiast 1.4, we’re closing that gap.
Enthusiast is our open-source, agentic AI toolkit for e-commerce — enabling engineering teams to build custom AI agents grounded in product data and workflows, with full control over infrastructure, integrations, and models.
In version 1.4, agents can now interact directly with uploaded files — from invoices and product manuals to catalogs and datasets — enabling richer, more contextual automation across e-commerce systems.
This release also introduces two new pre-built agents that utilize the new file capabilities in tangible e-commerce scenarios.
New in Enthusiast v1.4
1. File Uploads in Agent Conversations
Enthusiast v1.4 introduces file upload support, allowing agents to receive and process files directly within a conversation. Supported file types include images, PDFs and CSVs. Developers can also build their own extensions to make other file types accessible to the AI.
Uploaded content becomes part of the agent’s reasoning context. Agents can parse, extract, and cross-reference information across multiple files — without leaving the chat interface or building custom ingestion scripts.
This feature enables engineering teams to build workflows that require contextual interaction with files.
2. New Pre-Built Agent: Product Catalog Enrichment
Building on the new file upload capability, Enthusiast 1.4 introduces a pre-built agent for catalog enrichment — designed to extract structured product data from unstructured sources.
Many e-commerce teams know the problem well: receiving product sheets from multiple vendors, each with different layouts and inconsistent structure. Transforming this data into a customer-friendly catalog is a time-consuming manual process, prone to error.
The Catalog Enrichment Agent automates that process. You define which product attributes your catalog should contain — for example, name, SKU, dimensions, material, or price. Then, you upload supplier PDFs, spreadsheets, or photos. The agent processes each file, extracts the relevant information, and maps it to your catalog structure automatically.
This allows engineering teams to standardize data onboarding across vendors without writing custom parsers or cleaning scripts.
Example: Upload a vendor’s 200-page product catalog PDF. The agent identifies each product, extracts all defined attributes, and returns a structured, enriched dataset ready to merge with your existing system.
3. New Pre-Built Agent: Purchase Order OCR
Also built on the new file upload functionality, Enthusiast 1.4 introduces an agent for automated order creation — designed to translate unstructured order inputs into structured draft orders.
In some industries, order processing still relies on manual workflows. Customers send purchase requests as paper letters, handwritten lists, or email attachments in inconsistent formats. Each one requires manual interpretation and data entry before it reaches the Order Management System.
The Order Creation Agent automates this process end-to-end. You forward or upload the scan, photo, or document then the agent reads the content, identifies product names or SKUs, matches them to your existing catalog, and builds an order in your e-commerce system (such as Medusa.js).
This allows teams to integrate legacy order channels into automated workflows. Unlike traditional OCR tools, using an AI Agent that also internalizes the full product catalog, makes it possible to automatically process a wide variety of inputs with a high confidence rate.
Example: A wholesale client sends a photo of a handwritten order list. The agent parses each item, finds the corresponding SKUs in your Medusa catalog, and generates a ready-to-review draft order.
Native Medusa Integration
Enthusiast v1.4 ships with a native Medusa connector, so your agents can:
→ Sync and index catalogs directly from Medusa
→ Access and enrich product data using AI-driven workflows
→ Automate customer-facing or back-office operations
This connector makes it possible for Medusa developers to easily extend existing e-commerce logic with AI capabilities instantly, using the same familiar infrastructure and conventions.
Built for Engineering Workflows
As with every Enthusiast release, v1.4 prioritizes transparency, flexibility, and full control across the stack:
→ Run locally with Docker, deploy to the cloud with Kubernetes, or host on your own servers.
→ Model-agnostic architecture - works with OpenAI, Gemini, Mistral, or Ollama
→ E-commerce connectors for Medusa, Shopify, Shopware, and Solidus
→ LangSmith integration for execution tracing and debugging
You stay in control of your data, deployment, and models – while leveraging an open, extensible AI stack.
Get Started
Deploy your first file-aware, workflow-connected agent in minutes.
GitHub: upsidelab/enthusiast
Docs: upsidelab.io/tools/enthusiast
Contributions and ideas are always welcome. Let’s keep building open, extensible AI systems solving e-commerce challenges!
This content originally appeared on DEV Community and was authored by Chris Yaowen Zhang
Chris Yaowen Zhang | Sciencx (2025-10-23T15:09:04+00:00) Announcing Enthusiast 1.4: AI Agents Meet E-commerce Workflows. Retrieved from https://www.scien.cx/2025/10/23/announcing-enthusiast-1-4-ai-agents-meet-e-commerce-workflows/
Please log in to upload a file.
There are no updates yet.
Click the Upload button above to add an update.