Order groceries online with a local LLM skill
I keep finding more and more usecases for Q3.6 27b (same league) and the best performance is, when answers to my question is already in the context.The moment I'm trying something open-ended or ambitious, Claude/ChatGPT clearly take you to the goal quicker.For things, where there's a way to build a knowledgebase though, the local llm definitely can be a true contender. Task: The user built a custom grocery shopping skill for a local LLM (Qwen 3.6 27b) to automate online ordering. The system integrates loose fridge inventory, the last 10 representative orders for preference learning, and real-time stock data. Execution occurs via Chrome driven by the OS accessibility API, allowing the agent to navigate shopping sites and place orders autonomously.; tools: custom-built or unnamed agent. Human role: The original post does not say how a person stayed involved.; limitations: The source does not specify human approval steps, error handling for out-of-stock items, or payment security measures. Reliance on OS accessibility APIs for browser control may be fragile across different websites. The claim of 'fully solved' is anecdotal and lacks verification of order accuracy or financial impact..
Tools
custom-built or unnamed agent
Published on Agent Work Atlas:
Overview
I keep finding more and more usecases for Q3.6 27b (same league) and the best performance is, when answers to my question is already in the context.The moment I'm trying something open-ended or ambitious, Claude/ChatGPT clearly take you to the goal quicker.For things, where there's a way to build a knowledgebase though, the local llm definitely can be a true contender.
Workflow
The user built a custom grocery shopping skill for a local LLM (Qwen 3.6 27b) to automate online ordering. The system integrates loose fridge inventory, the last 10 representative orders for preference learning, and real-time stock data. Execution occurs via Chrome driven by the OS accessibility API, allowing the agent to navigate shopping sites and place orders autonomously.
Reported result
The author reports that the local LLM successfully ordered products online, solving a personal pain point regarding ingredient availability and waste. No quantitative metrics, error rates, or cost savings are provided. The user notes that while open-ended tasks favor cloud models, this specific knowledge-base-driven workflow was fully resolved by the local model.
Human role
The original post does not say how a person stayed involved.
Limitations
The source does not specify human approval steps, error handling for out-of-stock items, or payment security measures. Reliance on OS accessibility APIs for browser control may be fragile across different websites. The claim of 'fully solved' is anecdotal and lacks verification of order accuracy or financial impact.
Try it
Act as a local LLM agent configured for grocery ordering. My context includes: [replaceable: fridge inventory list], [replaceable: last 10 order history], and [replaceable: real-time stock API data]. Use Chrome via OS accessibility APIs to navigate [replaceable: target grocery website]. First, perform a dry run: identify 3 items needed for [replaceable: specific meal plan], check stock, and generate a cart summary. Do not place the order. Present the summary to me for approval. If approved, execute the checkout process. If any item is out of stock, pause and ask for substitution preferences. Ensure no payment is processed without explicit confirmation.
Source reports are not independently verified. Publication date is the catalog publication date, not the original article date.