Documentation

Struktive Documentation

Field definitions, quality score formula, classification rules, and export format specifications.

Getting Started Field Definitions Quality Score Classification Rules Export Formats Privacy Policy Terms of Service

Getting Started

Struktive accepts any CSV file containing data centre asset records. You don't need to rename your columns or reformat your data before uploading.

1. Upload your CSV

Go to the Upload page and drag-and-drop or select your CSV file. Optionally set a site default (applied to records without an explicit site in the location field) and a job name.

2. Processing

The seven-stage pipeline runs automatically. Processing typically takes 20–60 seconds depending on file size and the proportion of records requiring LLM classification.

3. Review results

The Results page shows the normalised asset register with quality scores, category badges, and flagged issues. Use the filters to focus on exceptions.

4. Download

Click Download and choose your output format. For DCIM exports (NetBox, Device42, Sunbird dcTrack), a pre-flight summary panel appears after download showing what needs to be created in your target platform.

Output Field Definitions

All output formats include these normalised fields. The Struktive Native CSV includes all fields. DCIM export profiles map a subset to platform-specific field names.

FieldTypeDescription

normVendorstringCanonical manufacturer name after alias resolution. E.g. 'Dell Technologies', 'Cisco', 'Schneider Electric'.

normModelstringModel name as provided, with minor normalisation (leading/trailing whitespace, double spaces).

hostnamestringHostname as provided. Not normalised.

normSerialstringSerial number with whitespace and punctuation normalised. Original preserved in rawSerial.

normSitestringSite name extracted from location string, or the upload-time site default if no site was found in the location.

normRackstringRack identifier in 'RACK-{n}' format. E.g. 'RACK-03', 'RACK-12'.

normUnumberU position as an integer. E.g. 24. Null if not resolved.

normRowstringRow identifier. E.g. 'ROW-A', 'ROW-B'. Null if not resolved.

categoryenumOne of: Compute, Storage, Networking, Power, Cooling, Infrastructure, Monitoring, Security, Out of Scope.

classificationConfidencenumber0.0–1.0. Confidence of the classification. ≥0.85 = High, 0.65–0.84 = Medium, <0.65 = Low.

classificationMethodenumkeyword | llm | out_of_scope. How the classification was determined.

qualityScorenumber0–100. See Quality Score documentation for the scoring formula.

issuesstring[]Pipe-delimited list of quality issues. E.g. 'missing_serial|low_confidence|no_rack'.

normStatusstringNormalised status. One of: Active, Inactive, Decommissioned, Planned, Unknown.

ipAddressstringIP address as provided. Not validated.

isDuplicatebooleanTrue if this record was flagged as a likely duplicate by any detection method.

duplicateOfstringRow index of the record this is a duplicate of, if isDuplicate is true.

isOutOfScopebooleanTrue if the asset was classified as Out of Scope (VoIP phone, laptop, printer, etc.).

netboxSlugstringNetBox DeviceType Library slug if the model was matched. Null if not matched.

uHeightnumberU-height from NetBox DeviceType Library. Null if model not matched.

processedAttimestampUTC timestamp (ms since epoch) when this record was processed.

jobIdstringUUID of the normalisation job that produced this record.

Quality Score

Each record receives a quality score from 0 to 100. The score is calculated from seven factors and capped at 100. Records scoring 70 or above are considered import-ready.

FactorPoints

Hostname present+5

Serial number present+15

Model present+10

Vendor normalised (exact alias match)+10

Site resolved+10

Rack resolved+10

U position resolved+5

Classification confidence ≥0.85 (High)+15

Classification confidence 0.65–0.84 (Medium)+8

NetBox DeviceType matched+10

Duplicate flagged−15

Out of scope0 (excluded from scoring)

≥85: Excellent

70–84: Import-ready

50–69: Needs review

<50: Critical

Classification Rules

Assets are classified into one of eight categories (plus Out of Scope) using a two-tier engine. Keyword rules handle approximately 80% of assets deterministically. LLM inference resolves the remaining 20% with confidence scoring.

CategoryExample assets

ComputeServers, blade chassis, hyperconverged appliances, workstations

StorageSAN arrays, NAS devices, tape libraries, disk shelves

NetworkingSwitches, routers, firewalls, load balancers, Fibre Channel switches

PowerUPS units, PDUs, busways, ATS units, power meters

CoolingCRAC/CRAH units, in-row coolers, chillers, cooling controllers

InfrastructureRack enclosures, cable managers, patch panels, KVM switches

MonitoringEnvironmental sensors, DCIM probes, power monitoring units

SecurityFirewalls, IDS/IPS appliances, HSMs, physical security controllers

Out of ScopeVoIP phones, laptops, desktops, printers, label makers, IP cameras

Export Formats

See the Exports & Integrations page for detailed field mapping tables and pre-flight validation documentation for each DCIM profile.

NetBox CSV

Slug-based CSV for NetBox bulk import API

Device42 CSV

3-row-header CSV for Device42 bulk import

Sunbird dcTrack XLSX

Multi-sheet XLSX with Models Not Found worksheet

Struktive Native CSV

Flat 30-column CSV with all normalised fields

Asset Register XLSX

Formatted Excel workbook with conditional formatting

Compliance Audit Pack

6-tab XLSX: Cover, Register, Audit Trail, Duplicates, Exceptions, Methodology

Capacity Summary XLSX

6-tab XLSX: Cover, Site Summary, Rack Utilisation, Power Density, Category Breakdown, Heat Map

EU EED Pre-Report CSV

Asset-derived fields for EU EED Annex I & II reporting

Privacy Policy

Struktive is operated by GeekSpeak Commerce. Uploaded CSV files are processed in-memory and stored temporarily for the duration of the normalisation job. Processed results are stored in our database associated with the job ID.

If you are not logged in, your job is anonymous and accessible only via the direct job URL. If you are logged in, your jobs are associated with your account and accessible via the History page.

We do not share uploaded data with third parties. LLM classification calls are made to our internal API and do not include personally identifiable information from your asset records.

For questions about data retention or deletion, contact us at [email protected].

Terms of Service

By using Struktive, you agree that you have the right to upload the data you provide and that you will not upload data containing personally identifiable information about individuals without appropriate consent.

Struktive is provided as-is. GeekSpeak Commerce makes no warranty about the accuracy of normalisation, classification, or enrichment results. You are responsible for validating output before importing into production systems.

The free tier is provided without SLA. Paid tiers include SLA terms as specified in the relevant service agreement.

For questions, contact [email protected].