Joshua is an assistant that runs entirely on your own computer, a Mac or a Linux PC. It searches the web, reads your files, writes code, makes pictures and learns how you like answers, without sending your life to anyone's servers.
curl -fsSL https://ask.joshua.earth/install.sh | bash
Paste it in a terminal. A Mac with Apple Silicon (M1 or later) or a Linux PC, 8 GB of memory, 10–40 GB of free disk. On Linux an NVIDIA graphics card makes it much faster. Free and open.
▸ Checking your Mac ✓ Apple M4 · 16 GB memory · 212 GB free · macOS 15.1 ▸ A few questions With 16 GB of memory, Joshua uses the 8-billion- parameter model: the best fit for your Mac. Install the deep-thinking version? [Y/n] y Install image generation? [Y/n] y Install the fine-tuning tools? [y/N] y ✓ Plan: qwen3-vl 8b + deep thinking + images + fine-tuning · about 21 GB Go ahead? [Y/n]
What you ask an assistant is a diary of your worries, your work and your plans. With Joshua, the model runs on your own chip. Your chats, files and memory sit in one folder on your disk, in plain files you can open, copy or delete.
When Joshua searches the web it goes out anonymously, with no account and no cookies, so nobody builds a profile from it. The same question gets the same results on any computer, in the language you asked in.
Untick the internet and Joshua keeps working: on a train, in the mountains, during an outage. Nothing switches off because a company changed its plans or its prices.
Every rule Joshua follows is in a few readable Python files. The memory it keeps about you is a notes file you can edit. And you can teach it your own style, on your own computer, with your own examples.
What you give up. A model that fits on a laptop is smaller than the giant ones in data centres: it makes more mistakes and writes less brilliantly. Joshua makes up for part of it by checking the web before it answers and fact-checking itself against its sources, and it tells you where each fact came from. It is also slower, and it needs a reasonably recent Mac.
For facts that change, it searches, reads the pages, answers with numbered sources, then fact-checks its own answer against them.
Drop in text, Word, PDF or images. Each chat has its own folder; point a chat at a project folder and it works there.
Python in a sandbox that can only write in the chat's folder. Packages install on demand.
Looks at your photos and screenshots. With images installed it creates pictures and edits your own photos, locally.
Keeps short notes on who you are and how you like answers, in files you can read and edit. Switch it off any time.
Goal mode works until a task is done and checked. Swarm mode asks several agents and merges the best answer. Chats run side by side.
The installer looks at your computer first: chip, graphics card, memory, free disk, and estimates how fast each model would run on it. It asks whether you want deep thinking, images and fine-tuning, shows the plan and the space it needs, and waits for your yes.
Then it installs Ollama, the engine that runs the models, a private copy of Python, Joshua and the models you chose, and adds Joshua to your Applications folder (Mac) or applications menu (Linux). Everything else stays in one folder, ~/AskJoshua, with an uninstall.sh that removes it.
Prefer to read before you run? Open install.sh: it is plain, commented shell.
| Memory | Model | Options |
|---|---|---|
| 8 GB | qwen3-vl 4B · about 3 GB | Deep thinking. Images are not offered: they need 16 GB. On a slow chip the 4B is suggested even with 16 GB. |
| 16–24 GB | qwen3-vl 8B · about 6 GB | Deep thinking, images (FLUX.2 klein), fine-tuning. |
| 32 GB+ | qwen3-vl 30B · about 20 GB | Everything, with a larger and smarter model if you want it. |
| Linux | chosen by the graphics card: 4B under 7 GB, 8B from 7 GB, 30B from 22 GB | Images from 12 GB, fine-tuning from 8 GB, on NVIDIA cards. Without a card, the 4B runs on the processor, slowly. |
Fine-tuning nudges a few of the model's billions of numbers with your examples, so it picks up a habit: a tone of voice, a format, your jargon. Joshua walks you through it in the Teach Joshua panel, explaining each step as you go. It runs on your Mac in 15 to 60 minutes, and your examples never leave it.
It teaches style, not facts. Train it on 200 answers about your company and it will sound like it knows your company, but it will invent the details. For facts, Joshua has memory, your files and the web.
joshua-<name>.The model has no memory of its own. Every time you write, Joshua sends it the whole conversation, and the model either answers or asks for a tool. Joshua runs the tool, adds the result, and asks again, up to 25 rounds.
Built with Ollama · qwen3-vl · FLUX.2 klein via mflux · mlx-lm · llama.cpp's converter · Gradio · anonymous search through ddgs. All open source.