Company knowledge.
Ready to use.
Smart search and knowledge management built on your information, for your team. We’ve delivered these tools for several companies; some of that work can’t be shown publicly.
Try the search illustration ↓What we can show
Private work.
A public illustration.
Some client work is under NDA. The demonstration below runs on sample data — not client documents, not screenshots of private systems.
The implementation described here is one recruitment-company search project, known as AskKate. Its architecture and delivery history describe that project, not every knowledge system we build.
The wider offer is knowledge management. This example demonstrates private document discovery, not generated chatbot answers.
People / Policies
Leave requests, entitlement and approvalPeople / Getting started
The starting point
The files hadn’t gone.
The way in had.
The company’s old Windows search tool had grown fragile and stopped finding things reliably. The documents were all still there — Office files and PDFs across shared drives.
Without it, finding anything meant remembering where it lived or walking the folders by hand.
The brief: restore access without asking everyone to reorganise their files first.
Try it with sample documents
Different words.
The right document.
Choose a preset query and compare meaning-based search against keyword matching. Sample data throughout — not client documents, not a live service.
Click a preset query above to see it in action
What we built
A familiar place.
A better way through.
AskKate brings keyword and semantic search together. Staff can find an exact phrase, describe a concept, browse folders and preview supported documents before opening the original.
Search
Exact terms when the wording matters. Related concepts when you know what you mean but not what the file is called.
Browse
The folder structure stays familiar. Search is another way in, not a replacement for how the team organises its work.
Review
Filters and previews help staff decide which document they need before opening it.
AskKate returns documents and excerpts. It doesn’t generate answers or pretend to be a chatbot.
Under the surface
Two ways to search.
One useful list.
Python and Flask extract text and metadata. BM25 through Lucene catches exact words and names; ChromaDB catches related concepts. Reciprocal Rank Fusion merges the lists, with optional cross-encoder reranking.
The React interface first shipped as a Windows desktop app through Electron, so files could open in their native applications. Browser access followed later.
The system is designed for client-controlled documents and private deployment. AskKate is owned and operated by the client; Spectrix designed and built the replacement.
Have a shared drive like this? ↗Your next useful conversation
Tell us what’s
getting in the way.↗
A few lines about the process, the tools and the part that’s frustrating. That’s plenty to start with.
Chamber of CommerceBacked by a University of Chester
Venture award