Ask your documents. Get answers with sources. Keep your data.
A private AI assistant that answers staff questions only from a department's own documents — running entirely on the department's own server.
The Problem
Government departments hold years of valuable knowledge in documents, but staff spend too much time finding it, and public AI tools are not a safe way to search it.
Knowledge Is Buried
Policies, circulars, reports and manuals sit across shared drives, email and paper archives.
Search Is Slow
Keyword search returns files, not answers. Staff still have to open and read each one.
Public AI Is a Risk
Pasting official documents into public chatbots sends data to foreign servers, and answers can be invented.
Answers Are Inconsistent
Different officers give different answers to the same question. Expertise leaves when people do.
How It Works
Staff use it in three steps, like asking a well-read colleague who has every department document at hand.
Upload Documents
Staff add policies, reports and manuals as PDF or Word files.
Ask a Question
Typed in plain English, in any web browser. Nothing to install.
Answer With Sources
Written only from the department's documents, with the source documents and pages named. If the answer isn't there, it says so instead of guessing.
All three steps run on the department's own server — nothing is sent to public AI services.
What Makes It Different
Private by design, grounded in official documents, and affordable to run long term.
| Public AI chatbots | Vitisoft RAG AI | |
|---|---|---|
| Where data goes | Sent to the vendor's overseas servers | Stays on the department's own server |
| Source of answers | General internet training data | Only the department's uploaded documents |
| Can you check the answer? | Rarely shows sources | Every answer names its source documents |
| When it doesn't know | Often invents a plausible answer | Says the answer is not in the documents |
| Who controls access | The vendor | The department, by user and document collection |
| Cost model | Per user or per question | No AI usage fees; hardware, setup and support |
| Internet dependency | Always required | Works on the office network without internet |
Built on Open Technology
A single compiled Go service with PostgreSQL (pgvector) for search and Ollama to run open-source AI models locally. Models are swappable as better ones are released, so departments are not locked in to any single AI vendor.
Docker · Linux or Windows Server · on-premise or government private cloud
Who It's For
Any department or public institution that relies on a large body of documents.
| Organisation type | Documents it would hold | Example staff question |
|---|---|---|
| Regulators | Acts, regulations, licence conditions, guidelines, past decisions | “What are the renewal requirements for a class licence?” |
| Agricultural research institutes | Research reports, trial results, crop variety guides, extension manuals | “Which rice varieties were tested for drought tolerance, and what were the results?” |
| Education departments | Curriculum frameworks, policy circulars, school guidelines, examination rules | “What does current policy say about continuous assessment in lower secondary?” |
| Ministries and agencies | Standing orders, HR manuals, procurement guidelines, strategic plans | “What approvals are needed for a procurement above the micro threshold?” |
| Front-line service offices | FAQs, service charters, procedures | “What documents does a citizen need to apply for this service?” |
How We Engage
Departments move from a live demonstration to production in four steps, committing only to a fixed-period pilot before deciding on full rollout.
Demonstration
A live session showing RAG AI answering questions from sample documents. No cost, no commitment.
Discovery
The department completes a questionnaire covering its documents, users and IT setup. We return a recommended setup and pilot proposal.
Pilot
A fixed-period pilot with one unit, its documents and a defined group of users, including staff training.
Production
Rollout to more units on the department's own infrastructure, with ongoing support and updates.
Delivered with a local implementation partner in Uganda for on-site support and training.
Where We Are Today
Being clear about status: document question-and-answer with cited sources is working in our lab, and we are now looking for departments to run pilots. Production features such as sign-in, document library management and a JSON API are being built as part of the pilot phase. The current release works with English documents and questions.
RAG AI is also the engine behind the reporting capability on our CleanTechAI platform.
See It on Your Documents
Book a no-cost demonstration with sample documents from your department.
Get in Touch