Your documents, indexed for grounded retrieval.
Store your documents, index and embed them on-premise, and let your agents retrieve grounded context with citations back to the source file. Nothing is sent to a third-party API, not even for inference.
Everything your documents need to become answers.
Storage, OCR, on-prem embeddings and grounded retrieval, built into one governed store rather than stitched together from separate products.
On-prem embeddings
Index and embed every document inside your network, using the model you choose. The vectors live on your disks.
Source citations
Every grounded answer links back to the exact file behind it.
Governed access
Per-file visibility, and an audit trail that records every view, change and deletion. RBAC throughout.
Built-in OCR
Turn scans, prints and images into searchable, indexable text.
Version history
A new upload supersedes, it does not overwrite. The wording you signed off last quarter is still there and still numbered.
Live indexing status
Watch each file move from upload to indexed, OCR and embedding in between.
The formats an office actually holds
PDFs, Word, Excel and PowerPoint, scans and images, audio and video. One store, one pipeline, one index.
Every file has a life, not just a location.
Filestore holds the whole of it: the day it arrived, every version since, who opened it, and how it leaves.
Ingest
It arrives by upload or from a connected source. The store records when.
Read
OCR turns scans, faxes and photographs into machine-readable text.
Index
Embedded on your hardware, with the model you choose, then indexed.
Retrieval
Agents read the passages that answer, and name the file they came from.
Versions
Every version it has ever had, in order, with the current one marked.
Access and audit
Per-file visibility decides who may open it. The audit trail records who did.
Disposition
It leaves when someone with the right to remove it does. Nothing ages out on our schedule.
The same life in every scope: your own store, a shared workspace store, or a store scoped to one agent and nothing else.
Storage is the easy half.
What a store has to do after the upload: answer with a source, read what was never typed, and know who is allowed to see it.
Refunds are issued within 30 days of purchase, provided the item is unused and in its original packaging.
Answers you can trace to the source.
Agents retrieve grounded context from your indexed documents and cite the exact file behind every answer. No guessing, no black box, and nothing retrieved from a store the agent was not given.
- The exact file, not the folder it sits in
- Retrieval scoped per agent
- Grounded on your documents, not the open web
Every document becomes searchable text.
Filestore's built-in OCR reads scans, prints and images, including the ones nobody expects to work: faxed forms, stamped pages, a photocopy of a photocopy. It runs on your hardware like everything else here, so even paper-era documents are indexable.
- Faxes, stamps and photocopies
- 120+ languages
- 99.2% accuracy on the document types it is set up for
Control who sees what, down to the file.
Set per-file visibility and keep the full version history. Every upload, change and deletion lands in the audit trail, so who saw what and when is a query rather than an investigation.
- Per-file visibility under RBAC
- Superseded versions pruned to a limit you set
- Every view, change and deletion logged
The backfile stops being where documents go to be forgotten.
Most of what your organisation knows is already written down and filed. Once every page is read, embedded and indexed, the contract from 1998 answers as readily as last week's memo, and with the file it came from. Questions never worth asking, because answering meant a trip to the archive, become questions anyone can ask.
A records request stops being a project and becomes a question.
Scanned paper sits in the same index as everything born digital, so the whole record answers, not just the recent half.
Every answer arrives with the file behind it, so the person who has to sign it can check it.
Filestore, by the spec.
- Deployment
- On-prem · private cloud · air-gapped
- Embeddings
- On-premise · your choice of model
- OCR
- Built-in · 120+ languages
- Retrieval
- Grounded · cited to the source file
- Scopes
- Personal · shared · per-agent
- Formats
- PDF · Office · scans · images · media
- Storage
- Your hardware · per-store quotas
- Version retention
- Full history by default · capped per store
- Governance
- Per-file visibility · full audit trail
- Access
- RBAC · full audit logging
- Compliance
- SOC 2 Type II · HIPAA · GDPR · ISO 27001
100% on-premise. Zero data leakage.
Your data never leaves your network, not even for model inference.
The store, the index and the audit trail all sit inside your estate. These are the audits that cover them.
See Exemplary AI in action
Book a demo and we'll show you purpose-built agents grounded in your own knowledge — deployed on your infrastructure.