Inspiration
Small workshops and independent makers often keep purchase records scattered across invoices, receipts, and other documents. Connecting those records to a specific product, job, or production batch can become a time-consuming task.
I created CostTrace to make this process easier.
What it does
CostTrace is an AI NAS Skill designed to gather and cross-check purchase evidence related to a specific job or production batch.
It helps users locate relevant documents, extract available information, identify missing details and possible discrepancies, and organize the evidence into a reviewable record.
It does not assume that every purchased material was consumed or present uncertain information as verified accounting results.
How I built it
I developed CostTrace as a new Skill using the official UGREEN AI NAS Skill Kit.
The proposed workflow uses documented NAS capabilities for file discovery, document extraction, evidence collection, and information review. The package includes SKILL.md and supporting Markdown references.
I designed the workflow to request human confirmation before saving information.
Challenges
The main challenge was designing a useful workflow without assuming that the available tools guarantee exact accounting calculations or complete document traceability.
What I learned
I learned that useful AI automation is not just about extracting information. It must also identify uncertainty, show missing evidence, and keep people in control of important decisions.
What's next
My next goal is to validate CostTrace on a real UGREEN NAS and evaluate its document extraction capabilities using real workshop records.
Built With
- agents
- ai
- document
- knowledge
- markdown
- ocr
- ugreen
