Your documents never leave your servers
DevelMoGPT is a private AI chat assistant that runs entirely on your own hardware. Upload your company documents and your team can ask questions in plain language and get answers drawn from those documents, with the source files listed on every reply.
- The language model
- The document index
- Every conversation
- Your uploaded files
The answer is in a folder nobody can search
Contracts, policies, handbooks and reports pile up faster than anyone can read them, and the assistants that could answer questions about them want the files uploaded somewhere else. For teams with data residency rules, or anything commercially sensitive, that is the end of the conversation.
From a folder of files to an answer with sources
Upload
Admins add documents in the admin console or paste text directly. PDF, Word, Excel, CSV, PowerPoint and plain text are supported.
Index
Each document is converted to text, turned into an embedding by a local embedding model and stored in a vector index on your own server.
Ask
A user types a question in the chat window.
Retrieve
DevelMoGPT finds the three most relevant documents by meaning, not just keywords.
Answer
A language model running locally writes the answer from those documents and the recent conversation, then lists the sources it used with a confidence score.
What it does
- ✓Runs fully on premises. The language model, the search index and the chat history all stay on your hardware.
- ✓Answers grounded in your own documents, with the source documents listed on every reply.
- ✓Semantic search that matches on meaning, so people can ask in their own words.
- ✓Reads PDF, Word, Excel, CSV, PowerPoint and text files.
- ✓Admin console to upload, browse and search the knowledge base.
- ✓Secure sign in, passwords stored as secure hashes, and separate user and admin roles.
- ✓Saved, searchable conversations for every user, so work picks up where it left off.
- ✓No per message API fees. Once installed, it runs on your own GPU.
What it is built on, and what it needs
| Component | What it uses |
|---|---|
| Chat app | Next.js and React |
| Sign in | Token based sessions, hashed passwords, separate user and admin roles |
| App database | MongoDB, holding users and chat history |
| AI service | A Python REST API |
| Language model | Served locally, on your own GPU |
| Embeddings | A local embedding model, 768 dimensions |
| Vector search | FAISS, cosine similarity |
| File types | PDF, DOCX, XLSX, CSV, PPTX, TXT |
| Hardware | What you need |
|---|---|
| Minimum | 16 GB RAM, an NVIDIA GPU with 8 GB of video memory, 50 GB of free disk |
| Recommended | 32 GB RAM, an NVIDIA GPU with 16 GB of video memory, an NVMe SSD |
| Operating system | Ubuntu, Windows or macOS |
Where it is today
DevelMoGPT is a working prototype. We are taking it into pilots now, so the honest next step is a demo on a sample of your own documents.
Common questions
Does any of our data leave our servers?
No. The language model, the document index and the chat history all run on your own hardware. Nothing is sent to an outside AI provider.
What files can it read?
PDF, Word, Excel, CSV, PowerPoint and plain text. Admins can also paste text straight in.
What hardware do we need?
A server or workstation with an NVIDIA GPU, 8 GB of video memory minimum and 16 GB recommended, and at least 16 GB of RAM. DevelMo can advise on sizing for your team.
Can we try it?
Yes. DevelMoGPT is a working prototype today, so the next step is a demo on a sample of your own documents rather than a signup page.
Ask your own documents, on your own hardware
Book a demo and we will run DevelMoGPT against a sample of your files.