AI Business OS
An assistant that reads your documents and admits what it does not know.
Upload your policies, contracts and handbooks and ask questions in plain language. Every answer cites its source — and when the documents do not cover it, it says so.
- Runs on
- Your own hardware
- Data leaves?
- Never
- Every answer
- Cites its source
- Monthly AI bill
- None
Billing questions
Hybrid retrieval · 4 documents in scope
Annual plans can be refunded in full within 30 days of the charge. After day 30 the remaining months are credited rather than refunded, and the credit does not expire.
Grounded in
I could not find that in the indexed documents, so I am not going to guess. Nothing in the knowledge base covers a 2027 enterprise discount.
No matching passages · answer withheld rather than invented
The challenge
Teams want AI over their own documents, and two things stop them. They do not want contracts and HR files sitting on someone else's servers. And a tool that answers confidently when it has no idea is worse than no tool at all — one invented refund policy quoted to a customer undoes any time saved.
Our approach
Answers are built only from passages actually retrieved from your documents, and each one shows the file and page it came from, so anything important can be checked in seconds. When retrieval finds nothing relevant, the assistant says so rather than filling the gap — that behaviour is deliberately tested, not hoped for. The model and the search index both run on your own machine, so no document content leaves the building and there is no per-question bill.
What shipped
A self-contained system — document upload, indexing, cited answers, streaming chat and automation hooks — that runs on a single machine and is reachable securely from anywhere without renting a server.
What it does
- Ask questions across every document you have uploaded
- See the exact file and page each answer came from
- Get an honest “not in your documents” instead of a guess
- Upload PDFs, Word files and spreadsheets — indexed automatically
- Connect it to your existing tools and automations
- Run the whole thing in-house, with no per-question cost
Stack
- FastAPI
- PostgreSQL
- Qdrant
- Ollama
- Celery
- Docker
Our role
System architecture, retrieval pipeline, evaluation, self-host deployment.
Building something like this?
Tell us what you are building and we will come back with technical direction, scope and a timeline.
Start a similar build