Geometric Design Search

Search CAD models by shape using 3D shape embeddings.

Industry	Level	Difficulty	Recommended mode
Mechanical engineering	4 of 5	Expert	hnsw

Coming soon

A complete, runnable Geometric Design Search reference implementation is in progress. The architecture and core code below already work with SrvDB 0.3.

The problem

• Similar shapes hide behind different names.
• CAD search is metadata-only.
• Reuse needs geometry.

How SrvDB powers Geometric Design Search

1. A 3D shape embedding model turns models into vectors stored in SrvDB.
2. Query by an example model.
3. hnsw scales to large libraries.
4. Per-programme databases.

What changes

Today	With Geometric Design Search on SrvDB
Metadata search	Shape search
Duplicate geometry	Found by similarity
Programme silos	Searchable per programme

How it works

1. Export meshes.
2. Embed with a shape model.
3. Store.
4. Query by example.
5. Review.

SrvDB features it uses

• HNSW graph (hnsw): approximate search that visits a small part of large collections; tune with set_ef_search().
• Several databases: each database is a directory, so different data types or tenants stay separate.
• Crash-safe persist(): after it returns, data survives a crash or power loss and reopens exactly as committed.
• search_batch: runs many queries at once, in parallel inside Rust.

Use cases

• Casting reuse
• Tooling reuse
• Supplier quoting
• Duplicate detection
• Design inspiration
• Cost estimation

Core code

embed() is your embedding model (for example sentence-transformers, or a function that turns your records into feature vectors).

import json
import srvdb

db = srvdb.Database("./geometric-design-search-db", dimension=512, mode="hnsw")

def ingest(items):
    """items: list of (id, text, metadata dict)"""
    db.upsert(
        ids=[i for i, _, _ in items],
        embeddings=[embed(t) for _, t, _ in items],
        metadatas=[json.dumps(m) for _, _, m in items],
    )
    db.persist()                      # crash-safe checkpoint

def similar_shapes(model_description, k=5):
    hits = db.search(embed(model_description), k)
    return [(json.loads(db.get(i)), round(score, 3)) for i, score in hits]

Limits and honest notes

• Needs a suitable 3D shape embedding model (research-grade).

Build roadmap

1. Prototype: load a sample of real data with flat mode and check that searches return what an expert expects.
2. Measure: if you move to an approximate mode, compare its results with flat on the same queries (recall).
3. Pipeline: add continuous upsert and decide when to call persist().
4. Harden: test restarts and crashes, add logging, review every output path with domain experts.
5. Deploy: ship the database directory with the application or service.

All Mechanical engineering projects

Level	Project	What it does
1 · Beginner	Part Finder	Find existing parts by describing what they do.
2 · Intermediate	FMEA Assistant	Drafts and reviews FMEAs grounded in past FMEAs and failures.
3 · Advanced	Vibration Signature Matcher	Matches vibration signatures to known fault patterns.
4 · Expert	Geometric Design Search (this page)	Search CAD models by shape using 3D shape embeddings.
5 · Frontier	Simulation Lookup	Find the nearest previously simulated cases before running a new simulation.

See also: Use cases · Live examples (coming soon)