hamzaabduljabbar

AutoConst Drawing Takeoff Claude — Data skill for Claude Code

Data community

Index construction PDF drawing sets into a queryable database.

How to install AutoConst Drawing Takeoff Claude

This entry records only its repository, not the path inside it, so there is no exact command to give. Open hamzaabduljabbar/autoConst-drawing-takeoff-claude and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What AutoConst Drawing Takeoff Claude does

Index construction PDF drawing sets into a queryable database. Counts, dimensions, confidence scores, and sanity checks from the PDF vector layer. Claude Code skill + CLI.

Alternatives in Data

  • Opencode Mem — OpenCode plugin that gives coding agents persistent memory using local vector database 1.5k ★
  • Anthropic Retrieval Demo — Lightweight demo using the Anthropic Python SDK to experiment with Claude's Search and Retrieval capabilities 191 ★
  • FACT — FACT – Fast Augmented Context Tools: FACT is a lean retrieval pattern that skips vector search 161 ★

README

Drawing Takeoff

**Turn a construction PDF drawing set into a structured, queryable database — then answer takeoff questions from the database instead of making an LLM re-read the drawings every time.**

Contractors get flattened PDF drawings, not the BIM model. But a construction PDF still carries a hidden **vector text layer** with the real tags, schedules, and dimensions. This tool reads that layer once, builds a SQLite database, and answers questions from the database. Counting objects becomes exact (it's reading text, not guessing from pixels), every number is tagged with a confidence level, and measurements are sanity-checked against the building envelope.

It ships as a [Claude Code](https://docs.claude.com/en/docs/claude-code) skill, but the engine is a plain Python CLI you can run on its own.


Why

Feeding raw drawings to a model is the expensive, unreliable path — it guesses from pixels and mis-weights detail buried in a huge context. Reading the vector text is cheaper and accurate. Querying a finished database is cheapest of all and deterministic:

Approach What the model reads Relative cost
Raw drawing images pixels highest
Text + markdown extracted text lower
This tool a structured query result lowest — a query answer is a few hundred bytes, not a 49 MB PDF

Index once. Build the database. Query it forever.


Install

Requires **Python 3.10+** and one library:

pip install -r requirements.txt      # just PyMuPDF

On Windows use the `py` launcher; on macOS/Linux use `python3`.

As a Claude Code skill (optional)

Copy the repo into your skills directory so Claude uses it automatically:

~/.claude/skills/drawing-takeoff/            (global)
/.claude/skills/drawing-takeoff/    (per project)

Then ask *"how many doors in this set?"* and it builds/queries the database instead of reading the PDF.


Quickstart

# 1. Index the PDF once (fast). A