dutoannha.com is a Vietnamese-language construction cost estimator for people planning to build a house. It turns project details into a cost breakdown by item and a payment schedule, and it also reviews contractor quotes.
The problem
People planning to build a home need a rough budget early, before they talk to a contractor. They also need to know at which stage each part of the money will be due.
How it works
According to the project's README, the design rule is that the AI does not calculate money and does not invent prices. A language model reads uploaded drawings, PDFs and spreadsheets and turns them into structured lines. Material names are matched to standard items. Unit prices come from a PostgreSQL database and the totals come from a calculation engine written in Python, not from the model. The model then explains the result in words.
My role
Nguyễn Hải Minh designed and built the system together with AI, working with Claude and Codex and coordinating several AI agents.
Tools and stack
- Next.js (App Router) for the website
- FastAPI (Python) for the API and the calculation engine
- PostgreSQL 16 with pgvector for prices and embeddings
- Docker Compose for the database, API and web containers
- An LLM for reading documents and explaining results
Checked
- The public site at dutoannha.com opens.
- It produces an estimate table from the data entered.
Not checked
- How close the estimates are to real construction costs.
- Visitor and customer numbers, which are not published here.
What can be observed
On the public site, entering project details returns an estimate table split by item, together with a payment schedule.
FAQ
What is dutoannha.com?
dutoannha.com is a Vietnamese web tool that estimates the cost of building a house from project details, with a breakdown by item and a payment schedule.
Does the AI calculate the prices?
No. By the project's own design rule, prices come from a database and totals from a calculation engine. The language model reads documents and explains the numbers.
What is dutoannha.com built with?
Next.js for the website, FastAPI for the API, PostgreSQL 16 with pgvector for data, and Docker Compose to run it.
What was checked?
The public site opens and produces an estimate table from the data entered. The accuracy of the estimates against real construction costs was not checked.