Industrial Parts Data Extractor

Turn messy industrial part names into clean, structured JSON: product type, seal profile, dimensions, material, hardness, thread specs, hose and fitting codes. Purpose-built for hydraulic and pneumatic components — seals, hoses, fittings, filters, cylinder parts — where generic text parsers fail.

Feed it product names from an ERP export, a supplier price list, or a webshop catalog, and get back machine-readable specifications you can filter, search, and match on.

Why this exists

B2B product data is broken. In a typical industrial distributor's catalog, the technical parameters live inside free-text product names like:

Rod seal - S06 - HPU - 63x78x12
Hose - 2SC10 x 2500 - DKOS12 - AGR3/8
Piston seal - K01 - NBR - 55x45x11/12

That data is invisible to faceted search, impossible to deduplicate, useless for cross-referencing — and unreadable for AI purchasing agents. Studies show 43% of B2B buyers abandon orders because of incomplete product information, and manually structuring a single SKU takes 20–46 minutes.

This Actor does it in milliseconds, at scale, for a fraction of a cent per record.

What it extracts

Field	Example input	Extracted
Product type	Rod seal - S06 - HPU - 63x78x12	rod_seal
Seal profile code	S06	S06
Labeled dimensions	63x78x12	id_mm: 63, od_mm: 78, h_mm: 12
Material + hardness	NBR70	NBR, shore 70
Hose type, DN, length	2SC10 x 2500	2SC, DN10, 2500 mm
Fitting codes	DKOS12 - AGR3/8	DKOS 12, AGR 3/8
Threads	G1/2, M22x1.5, NPT1/4	system + size
Steel grades & attributes	seamless, hot rolled, S235JRH	normalized attributes

Dimension labeling is type-aware: for a piston seal, the first dimension is the outer (bore) diameter; for a rod seal, it is the inner (rod) diameter; for an O-ring, it is ID × cross-section. This domain logic is what generic PDF/text extractors get wrong.

The parser understands English and Hungarian product names out of the box, with more languages on the roadmap. Compound materials (HPU+NBR, PTFE+FPM) and dual heights (11/12) are handled.

Use cases

1. E-commerce & PIM catalog enrichment. Fill empty size, material, and type attributes in your webshop or PIM from the names you already have. In a real-world benchmark on a hydraulics distributor's live catalog, this Actor structured 145,000 products with a 97% hit rate — turning a flat text list into a filterable, faceted catalog.

2. Procurement, BOM matching and deduplication. Normalize supplier price lists that arrive in a different format from every vendor. Once every line has structured dimensions and materials, matching offers against your BOM or finding duplicate SKUs becomes a simple query.

3. AI agents & automation pipelines. Every Apify Actor is callable from n8n, Make, Zapier, and any MCP-compatible AI agent (Claude, ChatGPT, and other agent frameworks) via the Apify MCP server. Give your purchasing or catalog agent the ability to understand industrial part descriptions instead of guessing. The output is deliberately flat and predictable so agents can use it without post-processing.

Input

Provide items directly:

{
  "items": [
    { "name": "Rod seal - S06 - HPU - 63x78x12", "brand": "Kastas", "partNumber": "IUN-6378" },
    { "name": "Hose - 2SC10 x 2500 - DKOS12 - AGR3/8" }
  ]
}

…or point the Actor at a CSV file with a csvUrl (columns: name, optionally nameSecondary, brand, partNumber). Only name is required. brand and partNumber are passed through to the output so you can join results back to your source data.

Output

One dataset record per input item:

{
  "product_type": "rod_seal",
  "profile_code": "S06",
  "material": "HPU",
  "id_mm": 63,
  "od_mm": 78,
  "h_mm": 12,
  "brand": "Kastas",
  "mfr_part_number": "IUN-6378",
  "confidence": 1.0,
  "input_name": "Rod seal - S06 - HPU - 63x78x12"
}

confidence tells you how much of the name was fully understood (1.0 = every segment parsed). Unrecognized segments are returned in unparsed_segments, so nothing is silently lost — you always see exactly what was and wasn't extracted.

Pricing

Pay-per-event: you are charged only for successfully extracted records — inputs that yield no data cost nothing. No subscription, no minimum. A 10,000-row price list costs less than a coffee; compare that with 20+ minutes of manual work per SKU.

Supported vocabulary (v0.1)

• Seals: O-rings, rod/piston seals, wipers, U-cups, back-up, guide and groove rings, rotary shaft seals, seal kits — with profile codes
• Hydraulics: hoses (EN 853/856/857 and SAE codes: 1SN/2SN/1SC/2SC/4SP/4SH, R-series), fittings (DKOL, DKOS, DKR, AGR, BEL/CEL, SFL, ORFS, JIC…), tubes, valves, filters
• Materials: NBR, FPM/FKM, PU/HPU, EPDM, MVQ, PTFE, POM, HNBR, compounds, steel grades
• Threads: G (BSP), metric fine, R/NPT/UNF/UNC
• Cylinder parts, fasteners, accessories

Roadmap

• PDF datasheet mode: extract full spec tables directly from manufacturer catalog PDFs
• LLM fallback for exotic name formats, with the same structured schema
• Cross-reference lookup: given a manufacturer part number, find equivalent parts across brands

FAQ

My names use a different convention — will it work? The parser is pattern-based, not template-based, so most conventions work. If your format isn't recognized, open an issue with a few example names — vocabulary updates ship fast.

Do you store my data? No. Results go to your own Apify dataset; nothing is retained by the Actor.

Found a wrong extraction? Open an issue in the Issues tab with the input name and expected output. Issues are typically answered within one business day.