J M .
Jaydeep Mayani
Ahmedabad, IN Available for
Consultation
Back to the eCommerce Case Study

Free AI Product Descriptions Using a Local LLM

Case Study AI Automation Jaydeep Mayani

Every cloud AI API charges per request. For a catalog with hundreds of products needing descriptions, that adds up - and it means sending your product data to a third-party server every single time. I wanted a version of the same automation that cost nothing per run and kept everything on my own machine.

I'm Jaydeep Mayani, a Project Manager and AI Automation Builder based in Ahmedabad, India. This is how I built a product-description generator into my Shopify catalog pipeline using a locally-hosted open-source model instead of a paid API - zero marginal cost, and no product data ever leaves my computer.

Why Local Instead of Cloud?

My existing catalog pipeline already parses product spec sheets into Shopify-ready CSVs. But not every spec sheet includes a finished description - sometimes it's just a title, price, and a table of variants. Rather than writing those by hand or paying per-call for a cloud AI API, I wanted the pipeline to fill the gap itself, for free:

  • —No per-request cost - the model runs on hardware I already own
  • —No data leaves the machine - pricing, unreleased products, and supplier details stay local
  • —No rate limits - the pipeline can process an entire folder without waiting on API quotas

How It Works

1. Running a Model Locally with LM Studio

LM Studio is a desktop application that downloads and runs open-source language models directly on your computer, and exposes a local server with the same request format as OpenAI's API. That last part matters: it means automation scripts written for a cloud API work almost unchanged - you just point them at your own machine instead of a remote server.

2. Detecting What Needs a Description

As the pipeline parses each product's spec sheet, it checks whether a written description already exists. If one is missing, instead of leaving the field blank or flagging it for manual entry, it triggers a local generation request automatically - the pipeline never stops to wait on a human.

3. Generating a Structured, On-Brand Description

The script sends the product's known details - fabric, color, category, price point - to the local Qwen2.5 model with a strict prompt: a fixed word count, a consistent premium tone, and formatting rules that match the rest of the catalog. The result comes back in the same shape every time, so it drops straight into the CSV without manual cleanup.

from openai import OpenAI client = OpenAI(base_url="http://localhost:1234/v1", api_key="not-needed") response = client.chat.completions.create( model="qwen2.5", messages=[{"role": "user", "content": prompt}] )

See how this fits into the full Shopify catalog pipeline

The Trade-off, Honestly

Local models are genuinely free to run and fully private, but they're not a drop-in replacement for the largest cloud models on every task - output quality depends on which open-source model you run and how much hardware you have. For a fairly narrow, well-defined task like short product descriptions with a strict format, a mid-sized local model like Qwen2.5 performs well. For more open-ended writing, a cloud model may still be the better tool. Matching the model to the task matters more than defaulting to either option.

Who This Is For

  • —Sellers who want AI-assisted copywriting without an ongoing API bill
  • —Anyone handling unreleased products or sensitive catalog data who'd rather not send it to a third-party server
  • —Automation builders evaluating local models as part of a larger pipeline

Want a similar zero-cost pipeline built for your catalog?

Get in Touch

FAQ

Can AI product descriptions be generated without paying for API costs?

Yes - by running an open-source language model locally through tools like LM Studio, product descriptions can be generated with zero per-request API costs, since the model runs on your own hardware instead of a paid cloud service.

What is LM Studio and how does it work for automation?

LM Studio is a desktop application that runs open-source language models locally and exposes an OpenAI-compatible API on your machine, so existing automation scripts can call it exactly like they would call a cloud AI API, with no code changes needed.

Is local AI safer for private product data than cloud APIs?

Yes - since the model runs entirely on local hardware, product data, pricing, and descriptions never leave the machine or get sent to a third-party server, which matters for sellers with unreleased products or sensitive catalog data.

Which AI model was used for this pipeline?

This pipeline uses Qwen2.5, an open-source language model, run locally through LM Studio's OpenAI-compatible API.

Who is Jaydeep Mayani?

Jaydeep Mayani is a Project Manager and AI Automation Builder based in Ahmedabad, India, with 7+ years of experience across project management, digital marketing, and AI-powered workflow automation for eCommerce and content operations.