v1.0.0 · Free & Open Source

Private AI that runs
100% on your machine

Dunju Local Custom AI is a self-hosted AI assistant. Chat, analyse documents of any length, generate files, and automate work — with no cloud, no subscription, and no data ever leaving your computer.

Build. Automate. Scale.

100%
Local & Private
6
AI Experts
Document Length
0
Cloud / API Keys
What's Inside

Everything runs offline, on your computer

A complete private AI workspace — retrieval, documents, automation and privacy, with nothing sent to the cloud.

🔍

Smart Retrieval (RAG)

Reads whole documents of any length. Your file is chunked, embedded with a local model and stored on your machine — answers stay grounded in what you uploaded.

🎯

Hybrid Search + Citations

Keyword and semantic search combined, so exact names, IDs, SKUs and figures are matched reliably. Every answer shows the exact passages it came from.

🧠

6 AI Experts

General, Business, Coding, Medical, Kids and Fashion — each with its own focus. Create your own, pin a model per expert, and switch on reasoning mode for hard problems.

📄

Document Intelligence

Drop in PDF, Word, Excel, PowerPoint, CSV, text or images. Automatic type detection (invoice, resume, contract, report…) with tailored smart actions.

⬇️

Generate & Export

Turn any reply into a Word, PDF, PowerPoint or Excel file — or export a whole conversation as PDF, Markdown or text.

📁

Batch / Folder Automation

Point Dunju at a folder and run one action — summarize, classify, extract or custom — across every document. Read-only: it never changes your files.

🔒

Encrypted Backups

Password-protected AES backups. Your live data is never encrypted, so a lost password only affects that one backup — never your working data.

🔄

Conversation Controls

Regenerate, edit-and-resend, or branch any message into a new chat. Live response stats and a local activity log keep you in control.

🔌

Offline by Design

After a one-time setup, no internet connection is required. No subscription, no API keys, no telemetry — ever.

Quick Start

Up and running in four steps

Windows, with Python 3.10+ and Ollama. Full instructions are in the README and Customer Guide.

1

Install Ollama

Download and install the local AI engine, then confirm it works.

ollama --version
2

Install Python & dependencies

Python 3.10+ (tick "Add Python to PATH"), then install everything — including reportlab (exports) and cryptography (encrypted backups).

pip install -r requirements.txt
3

Create the model & enable retrieval

Build the Dunju AI model, then pull the local embedding model that powers Smart Retrieval.

cd Configurations ollama create dunju-ai -f Modelfile ollama pull nomic-embed-text
4

Start & open

Launch the backend (double-click Local_Custom_AI_AutoStart.bat), then open DunjuAI_Chat.html in Chrome or Edge. Look for Backend ✓ and Engine ✓.

Get Dunju Local Custom AI

Free. Open source. Yours.

Download the latest release, or clone the repository and follow the README. Runs entirely on your own hardware — your data stays with you.

No cloudNo subscriptionNo API keysNo data collection