YourAI

yourai CLI v0.0.4 · yourai.bdtren.com

yourai init [folder]

A step-by-step wizard: goal (chat or code), capabilities, model, quantization, context length, a model for each capability you turned on, interface, network access, and port. Options that don't fit your computer are listed separately with the reason. You see a review with download size and memory estimates before anything is created. Nothing is downloaded yet.

Examples below follow Duy on a MacBook Pro (Apple M5 Pro, 32 GB, about 20.8 GB of usable GPU memory). His computer can run every capability; he starts small and adds more later.

Goal — pick one

This only chooses the main chat model family. Capabilities in the next question work with either goal.

Example — Duy's first question

$ yourai init
? Project name: study-buddy

What do you want to do?
Your goal decides which models and settings we suggest.

┌───────────────────────────────┬──────────────────────────────────────────────────────────────────────────────────┐
│ Option                        ┆ What it means                                                                    │
╞═══════════════════════════════╪══════════════════════════════════════════════════════════════════════════════════╡
│ Chat assistant  (Recommended) ┆ A general assistant you talk to in the terminal, a browser, or via an API.       │
│                               ┆ Simplest setup and lowest hardware needs.                                        │
│ Coding assistant              ┆ Uses a model tuned for code with a longer context. Better for programming help.  │
└───────────────────────────────┴──────────────────────────────────────────────────────────────────────────────────┘

Recommended for first-time users: it works on almost any computer.
? Choose one:
> Chat assistant (Recommended)
  Coding assistant

Duy presses Enter on Chat assistant. Coding would suggest a code-tuned model and a longer context later.

Capabilities — pick any

Multi-select: Space toggles a row, Enter confirms. Options your computer runs comfortably are pre-ticked and marked (Recommended). See Images, video, and speech and the chat agent.

Example — Duy's M5 Pro (about 20.8 GB usable)

What else should it be able to do?
Pick any. Each one adds a download; image, video, and speech engines only load while you use
them. Change them later in yourai.toml.

┌──────────────────────────────────────────┬────────────────────────────────────────────────────────────────────────┐
│ Option                                   ┆ What it means                                                          │
╞══════════════════════════════════════════╪════════════════════════════════════════════════════════════════════════╡
│ Read files (agent)  (Recommended)        ┆ `yourai chat` can list and read files in the project (text, code,      │
│                                          ┆ PDF) and use the tools you turn on below. Writing files is off until   │
│                                          ┆ you allow it.                                                          │
│ See images and hear audio  (Recommended) ┆ Picks a chat model that understands pictures (and, with omni models,   │
│                                          ┆ audio). Attach files with /image and /audio in `yourai chat`.          │
│ Create images  (Recommended)             ┆ Make pictures from a text description. Runs only while you use it.     │
│ Speech to text  (Recommended)            ┆ Turn recordings into text with Whisper (many languages).               │
│ Text to speech  (Recommended)            ┆ Read answers aloud and save them as WAV files.                         │
│ Create short videos  (Recommended)       ┆ A few seconds of video from a text description. Slow: minutes per      │
│                                          ┆ clip even on a good GPU.                                               │
└──────────────────────────────────────────┴────────────────────────────────────────────────────────────────────────┘

Recommended for 20.8 GB of GPU memory: everything that runs comfortably on it.
? Choose any (space to toggle, enter to confirm):
> [x] Read files (agent) (Recommended)
  [x] See images and hear audio (Recommended)
  [x] Create images (Recommended)
  [x] Speech to text (Recommended)
  [x] Text to speech (Recommended)
  [x] Create short videos (Recommended)

What each option means in practice:

  • Read files (agent) — yourai chat can list and read project files (text, code, PDF). Writing stays off until you set allow_write = true.
  • See images and hear audio — swaps in a vision (or omni) chat model. Attach files with /image and /audio.
  • Create images — yourai image "…" and the /v1/images/generations API. Engine loads only while you use it.
  • Speech to text — yourai transcribe recording.wav and /v1/audio/transcriptions.
  • Text to speech — yourai speak "…" and /v1/audio/speech.
  • Create short videos — yourai video "…" and /v1/videos/generations. Slow even on a good GPU.

Duy unticks everything except Read files (agent) so the first download stays small. Each capability he leaves on gets its own model question next. He can turn the others on later in yourai.toml.

Flags and --yes

For scripts, --yes accepts every recommendation, except that only the agent is turned on unless you pass --capabilities (media models are large). Flags such as --name, --model, --quant, --ctx, --capabilities agent,image,stt, --image-model, --video-model, --stt-model, --tts-model, --access, and --port set answers directly.

Example — Duy starts with the smallest model in one line

$ yourai init --yes --name study-buddy --model qwen2.5-0.5b
Review
┌────────────────┬───────────────────────────────────────┐
│ Setting        ┆ Your choice                           │
╞════════════════╪═══════════════════════════════════════╡
│ Project folder ┆ study-buddy                           │
│ Goal           ┆ Chat assistant                        │
│ Capabilities   ┆ Read files (agent)                    │
│ Model          ┆ Qwen2.5 0.5B Instruct (Small, 0.4 GB) │
│ Quantization   ┆ Q5_K_M                                │
│ Runtime        ┆ llama.cpp                             │
│ Context length ┆ 4096 tokens                           │
│ Interface      ┆ API + browser chat page               │
│ Network access ┆ Only this computer (127.0.0.1)        │
│ Port           ┆ 8080                                  │
└────────────────┴───────────────────────────────────────┘
┌──────────────────────┬────────────────────────────────────────────────────┐
│ Estimate             ┆ Amount                                             │
╞══════════════════════╪════════════════════════════════════════════════════╡
│ Download             ┆ 420 MB model + 16 MB engine                        │
│ Memory while running ┆ about 869 MB (budget 20.8 GB)                      │
│ Disk use             ┆ about 461 MB in ~/.yourai (shared by all projects) │
└──────────────────────┴────────────────────────────────────────────────────┘
✔ Created study-buddy

Next: cd study-buddy && yourai build && yourai start

The project folder:

study-buddy/
├── yourai.toml          # all settings, with comments
├── yourai.lock          # written by yourai build (commit this)
├── yourai_module/       # models and engines (gitignored)
├── prompts/system.md    # the instructions the AI follows
├── data/docs/           # reserved for document search (later release)
├── outputs/             # images, videos, and audio you create
├── scripts/             # start/stop helpers
└── README.md

Duy turns it into a tutor by editing prompts/system.md: "You are Duy's study buddy. Explain topics simply, with one short example. When he says 'quiz me', ask one question at a time."

Questions or problems? support@bdtren.com · Contact page.