Threadlinq
From scattered paperwork to a live compliance picture on your terms.
Threadlinq reads your certificates, inspection reports, test records and manuals in any format, scans and phone photos included. It works out what each document is, which site, asset or area it covers and when it expires, then shows every checklist item as in date, due soon, expired or missing. It checks outcomes too, so an in-date report that says unsatisfactory is flagged, not waved through. Every answer links back to the exact page it came from.
| Evidence | Block A | Block B | Block C |
|---|---|---|---|
| Fire risk assessment | ?Check outcome | !Due soon | ✓In date |
| Fire door inspection | ✓In date | ✕Expired | ⚠Action needed |
| Fire alarm service | ✓In date | ✓In date | !Due soon |
| Emergency lighting | ✓In date | –Missing | ✓In date |
| Dry riser test | ✓In date | ✓In date | ✓In date |
| Smoke control | ✓In date | ✓In date | ✓In date |
| Lift examination | ✓In date | ✕Expired | ✓In date |
| Electrical report | ✓In date | ✕Expired | ⚠Action needed |
The Problem
Proving compliance shouldn't mean reading every file.
Compliance evidence arrives from many suppliers and contractors, over many years, in whatever form each one uses. Regulators, auditors and customers expect it to be accurate, up to date and available on request. In practice, that usually means someone opening every file.
Evidence is scattered everywhere
Certificates and reports sit in inboxes, shared drives, supplier portals and paper files. Just finding them all takes time.
Scans can't be searched
A scanned inspection report or a phone photo of a certificate is just a picture. Someone has to read it by eye.
Every supplier, a different format
PDFs, Word files, web-page logs and email attachments, each laid out differently, with dates buried in tables.
In date isn't the same as passed
An electrical report can be in date and still say unsatisfactory; a door schedule can hide one failed door. Checking dates alone misses both.
Missing evidence is invisible
Search only finds documents that exist. A certificate that was never filed stays hidden until an auditor or inspector asks for it.
Sensitive data rules out most AI tools
Business, customer and personal information often can't be uploaded to general online AI services, so the most useful tools are off the table. Threadlinq can run entirely inside your own network for exactly this case.
Threadlinq reads the files, checks them against your checklist and shows you what needs doing.
See How It WorksHow It Works
Six steps. From a pile of files to a clear answer.
You only do the first step. Threadlinq does the reading, sorting and checking, and you review what it found.
Collect
Drop in your documents in one go: PDFs, scans, photos, Word, Excel, web pages and emails.
YouRead
OCR reads scans and photos. Tables such as door schedules keep their rows and columns.
AutomaticKeep
Every document is saved as a complete digital copy, with the position of every word and table on its page.
AutomaticSort
Each document's type, the site or asset it covers, and its dates are worked out. Anyone can correct a field, and corrections are kept.
Automatic, you reviewCheck
Renewal dates are checked against today, and results too: in-date documents that failed are marked for action.
AutomaticAsk
Ask questions in plain English and get answers that name their sources, down to the page.
You
Demo Results
One upload. Nine things to act on.
We built a realistic demo building, Riverside Court, with three blocks and a deliberately messy pile of evidence: clean PDFs, skewed and coffee-stained photocopies, a phone photo, a Word file, a web-page log, an email with an attachment, and a contractor's quote mixed in.
Riverside Court and its documents are fictional, created for demonstration. Reading ran on an ordinary PC processor, with no GPU needed. The same outcome checks work in our food factory example, flagging a metal detector that missed its stainless test piece and a hygiene swab that found Listeria.
In Production
A hospitality group in the North of England
A group running hotels, pubs and restaurants across five sites uses Threadlinq for its compliance evidence, from food safety to fire safety, and every other check its pubs, restaurants and 3-star hotel are required to keep.
Before: the documents existed, but they were scattered across folders, email inboxes and paper files at each pub, restaurant and hotel, and no one knew when anything was due to expire. Every time evidence was needed, the owner and managers spent parts of several days hunting through emails and PDFs and checking on site.
Now: nobody has to go looking. They upload everything to Threadlinq, log on, and see every site in one place: what is in date, what is due soon and what has expired. No chaos, and the answer is always at their fingertips.
GenAI Answers
Ask in plain English. Check every answer.
A local AI model answers questions about your records using only what Threadlinq has read. Each answer lists the documents it used, and a click opens the original page with the source highlighted.
Show every fire door certificate for Block B and flag anything expired or missing.
Block B has one fire door inspection, carried out in April. It is expired: communal doors need checking every three months, so it lapsed on 27 July. It also recorded two doors failing on their closers.
What's due in the next 30 days at Riverside Court?
Two items: the Block C fire alarm service is due on 6 October, and the Block B fire risk assessment review is due on 23 October.
Sources: RC_BlockC_fire_alarm_service.pdf, RC_BlockB_fire_risk_assessment_scan.pdf
What needs action at Riverside Court, even if it's in date?
Two in-date documents failed: the Block C electrical report is coded unsatisfactory (no RCD protection on the plant-room sockets), and the Block C fire door inspection failed door C-FD04 on its closer. Block A's fire risk assessment gives no risk rating, so it needs checking.
Sources: RC_BlockC_EICR.pdf, RC_BlockC_fire_door_email.eml, RC_BlockA_fire_risk_assessment.pdf
Example questions and answers from the Riverside Court building safety demo; the same works for any checklist. Threadlinq also works with AI assistants such as Claude through the open Model Context Protocol, with the same sources attached.
Data Sovereignty
Your data, where you choose to run it.
Compliance records, customer data and supplier reports are sensitive. Run Threadlinq on-premise or in your own cloud to keep every document inside your environment, or let Fluxor host it for you as a managed service.
On-premise, your cloud, or hosted
A web app with sign-in, used from any browser. Host it on your own server or in your own cloud account, or use the Fluxor-hosted service; reading, sorting, checking and search all run in that one deployment.
Local AI when you need it
Self-hosted on-premise or in your own cloud, questions are answered by a model running on your own machines, and documents are never sent to a third-party AI service. If your documents must never leave your network, self-host; if they can, let Fluxor run it for you.
Records in an open format
Every document is kept as a complete, open digital copy. You can take your data to any other system without re-processing the originals.
- On-premise, in your own cloud, or hosted by Fluxor
- Works fully offline on-premise
- No special hardware or GPU needed
- No per-page or per-question fees
Fluxor Platform
Part of the Fluxor platform.
Threadlinq is the document side of Fluxor's data and AI products. Together they connect what your cameras see and your sensors report with what your standards and certificates say.
- Sightlinq for visionThreadlinq pulls defect definitions and acceptance criteria out of quality manuals. Sightlinq's GenAI annotation-quality review can check labels against that written standard.
- PiEdge for sensorsAlert limits can be suggested from equipment manuals, and each alert can point to the procedure that says what to do.
- SchemaShifts for dataDetections, sensor events and document records land in one governed database, ready for audit evidence packs and AI questions.
Where Threadlinq Helps
Any industry. Any compliance paperwork.
The engine stays the same. Only the checklist and sorting rules change for each industry.
Higher-risk residential buildings
Keep the golden thread complete and current for each block, and answer regulator or resident requests with sources.
Fire risk assessments, fire door inspections, alarm and lighting certificates, O&M manuals
Social housing
See every home or block across the common landlord compliance areas: in date, due, expired or missing.
Gas, electrical, fire, water hygiene, asbestos and lift certificates
Estates and facilities
One board per site and a portfolio view of renewals due, for schools, hospitals and commercial landlords.
Statutory inspection certificates across many sites
Manufacturing suppliers
Answer customer compliance questionnaires from cited evidence, without handing over the underlying documents.
Safety data sheets, material declarations, test reports, certificates of conformity
Machine builders and integrators
Keep technical files complete ahead of the EU Machinery Regulation, which applies from 20 January 2027.
Risk assessments, test reports, manuals, declarations of conformity
Your use case
If your team proves compliance from scattered documents, Threadlinq can be set up with your checklist.
Why Threadlinq
Built for the job, not just the documents.
Shared drives store files, and general document AI reads them. Threadlinq checks them against what you're required to hold.
| Capability | Shared drives and spreadsheets | General cloud document AI | Threadlinq |
|---|---|---|---|
| Reads scans and photos | No | Yes | Yes, with OCR |
| Checks evidence against a checklist | By hand | No | Yes |
| Shows what is missing | No | No | Yes |
| Tracks expiry automatically | By hand | No | Yes |
| Flags in-date documents that failed | By hand | No | Yes |
| Answers link to the source page | No | Varies | Yes, highlighted |
| Runs on-premise or in your own cloud | Yes | No | Yes |
| No per-page fees | Yes | Per page | Yes |
Platform Capabilities
Everything in one place. Nothing you don't need.
For IT and technical teams: the building blocks behind the board.
PDF, scanned PDF, PNG, JPEG, TIFF, Word, Excel, PowerPoint, HTML, Markdown, CSV, email (.eml, .msg) and text. Duplicate files are detected and skipped.
Scans and photos are read with OCR. Digital PDFs with a text layer skip OCR automatically, which keeps conversion fast on a CPU.
Headings, reading order and table rows and columns are preserved, so door schedules and results tables stay intact.
Each document is stored as lossless JSON with the page and position of every element. Classification and search can be rebuilt without converting again.
Document type, site or asset, and dates come from transparent rules with a confidence score. Human corrections are never overwritten and become training data.
Reads each document's result, such as pass, fail, unsatisfactory or a risk rating, and failed rows in tables. In-date but failed shows as action needed; a missing result shows as check outcome. People can confirm the outcome after remedial work.
One app for every industry. Each pack sets the checklist, renewal periods, what counts as a failed result and how documents are recognised, edited in the app with a live test and versioned publishing.
Uses the expiry date stated in a document, or calculates it from the issue date and the category's renewal period. Periods are configurable.
Every finding links to the original page, with the exact region highlighted, so any answer can be checked in seconds.
Questions are answered by a local model via Ollama. An MCP server exposes the same tools to Claude and other MCP-capable assistants.
Used from any browser, with sign-in and an activity log. Host it on-premise or in your own cloud, or use the Fluxor-hosted service; it runs on a standard CPU, and models can be pre-installed for fully offline sites.
The same JSON records feed data warehouses, reporting and machine learning, and connect to SchemaShifts-governed databases.
Get Started
See your own documents on the board.
Bring a sample of your documents and we'll show you what they reveal, including the in-date documents that quietly failed.