Defect Matcher

Matches new defects to past ones, from text and images.

Industry	Level	Difficulty	Recommended mode
Manufacturing & IoT	2 of 5	Intermediate	flat

Coming soon

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

The problem

• Same defects recur on different lines.
• Root causes are re-investigated.
• Images are not searchable.

How SrvDB powers Defect Matcher

1. Defect reports and photos are embedded (multimodal).
2. New defects retrieve the closest past ones with root causes.
3. Cosine thresholds flag recurring defects.
4. 8D reports added with upsert.

What changes

Today	With Defect Matcher on SrvDB
Repeated investigations	Past root causes in seconds
Images unused	Image similarity search
Line silos	Plant-wide matches

How it works

1. Embed reports and images.
2. Query with a new defect.
3. Review matches.
4. Confirm root cause.
5. Upsert the outcome.

SrvDB features it uses

• Exact search (flat): every query compares against every stored vector, so results are exact and explainable.
• True cosine scores: scores in [-1, 1] mean the same in every mode, so thresholds are stable.
• upsert: inserts new items and replaces changed ones, so the knowledge base stays current.
• Crash-safe persist(): after it returns, data survives a crash or power loss and reopens exactly as committed.

Use cases

• Surface defects
• Weld quality issues
• Assembly errors
• Supplier non-conformances
• Customer returns
• Warranty claims

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("./defect-matcher-db", dimension=512)

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

Limits and honest notes

• Image similarity needs a suitable vision embedding model.

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 Manufacturing & IoT projects

Level	Project	What it does
1 · Beginner	Maintenance Assistant	Technicians describe a symptom and get the right manual section and past fixes.
2 · Intermediate	Defect Matcher (this page)	Matches new defects to past ones, from text and images.
3 · Advanced	Sensor Anomaly Detector	Flags sensor windows far from normal operation.
4 · Expert	Plant Knowledge Agent	An agent with shift memory across lines, manuals and quality records.
5 · Frontier	Predictive Maintenance at the Edge	Finds machines whose behaviour resembles past pre-failure signatures.

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