See It In Action

What Does Owning Your AI Actually Mean?

Two minutes that explain why local AI changes everything — and what SIP gives you that cloud AI never can.

Alpha Release — v0.1.4.2

Download
SIP.

SIP Alpha is real, running software. It has conversations, remembers what you tell it, and learns over time. This is not a demo. Start here.

Loading…
No data centers. Just people.

🔒 This does not break your privacy. SIP sends only an anonymous count — no name, no location, no IP address, no identifying information. A one-way mathematical fingerprint lets us count unique instances without being able to identify any of them. You can turn this off in SIP’s settings at any time.

info

Alpha Software

SIP Alpha is actively developed software shared with early testers. It is stable enough for daily use and improves with each release. You may encounter rough edges — feedback is welcomed and genuinely acted on. Read the Model Differences Guide included in the download before selecting a model.

What's New — v0.1.4.2
June 12, 2026 · Global presence counter + privacy-safe fingerprinting
✓Live SIP counter — SIP now sends an anonymous ping on startup. See how many instances are active worldwide with /peers global. The counter on this page updates live.
✓Privacy-safe unique tracking — a one-way mathematical fingerprint counts unique instances without identifying anyone. No IP, no name, no location ever stored.
✓mDNS LAN discovery — SIP instances on the same WiFi network now find each other automatically (requires optional zeroconf package).
✓Skin selector crash fixed — switching skins no longer causes errors. The Cockpit skin right rail is now scrollable so Known Peers and Plugins sections are always visible.
✓SIP Network — connect SIP instances peer-to-peer. Use /peers connect <ip:port> to pair with another SIP on your network. Encrypted, sovereign, no central server.
✓Skin system — 5 visual themes: Refined Dracula, Calm Professional, Modern Dashboard, Cockpit, and Notebook. Click 🎨 Skin in the top bar to switch live.
✓Mobile access — click 📱 Mobile for a QR code. Scan with any phone on the same WiFi. No app install needed.
terminal Alpha — Windows

SIP Alpha

The complete SIP system — cognitive engine, Tkinter UI, plugin loader, memory systems, and local model server. Bring your own GGUF model or download one separately.

Requirements — Windows Build

Windows 10/11 · Python 3.11+ · 8 GB RAM minimum · No GPU required

GGUF model file required (not included)

macOS & Linux builds coming — SIP is platform-agnostic by design

download Download SIP Alpha v0.1.4.2 Release Notes (Repo Coming Public)
menu_book Required Reading

Model Differences Guide

SIP works with any GGUF-format model, but different models have different personalities and quirks. This guide covers every model tested with SIP Alpha — what works well, what doesn't, and what to expect.

Covers: Mistral 7B · LLaMA 3.1 8B · LLaMA 2 13B · Phi-4 14B · Phi-4 Mini · TinyLlama 1.1B

description Included in Download Recommended Models (Included in Download)
settings Documentation

Setup & Configuration

Step-by-step instructions for getting SIP running — from downloading a model to having your first conversation. Includes troubleshooting for common issues.

Covers Python setup, llama.cpp, model placement, first run, and plugin configuration.

description Included in Download Setup Guide (Included in Download)
AI Models

Download a Model

SIP requires a local GGUF-format model to think. Place your downloaded model in the models/ folder and SIP detects it automatically. All models below are available free from Hugging Face — the leading open-source AI model repository.

Hugging Face Models hosted on huggingface.co — open, free, community-maintained
Recommended

Mistral 7B Instruct v0.2

Q4_K_M

The best all-around model for SIP. Warm personality, follows the SIP persona reliably, handles both casual chat and factual questions well. Developed by Mistral AI (France).

RAM required6 GB min, 8 GB recommended
File size~4.4 GB
SIP LabelBalanced
downloadGet on Hugging Face dnsMirror: sip-ai.info
Smarter

Meta LLaMA 3.1 8B Instruct

Q4_K_M

Stronger logical reasoning than Mistral. Good for complex questions and analysis. Best on systems with 10+ GB RAM. Developed by Meta.

RAM required8-10 GB recommended
File size~4.9 GB
SIP LabelSmarter
downloadGet on Hugging Face dnsMirror: sip-ai.info
Fastest

TinyLlama 1.1B Chat

Q4_K_M

Tiny, fast, and light. SIP automatically falls back to this model if your system RAM gets critical. Also great for low-RAM systems or quick responses.

RAM required2-3 GB
File size~0.7 GB
SIP LabelFaster
downloadGet on Hugging Face dnsMirror: sip-ai.info
menu_book

Read the Model Differences Guide

Not all models behave the same as SIP. Some have strong independent personalities that can conflict with SIP's identity. The included Model Differences Guide covers all tested models so you know what to expect before you download.

open_in_newRead Online
extension

Plugin Suggestions Welcome

Have an idea for a SIP plugin? We'd love to hear it. Community suggestions directly shape what gets built next. Reach out at pachilles@sip-ai.info — no idea is too simple or too ambitious.

Up and Running in Minutes.

1

Download SIP Alpha

Unzip the download. Run setup_alpha.bat once to install the required components. After that, SIP launches directly from SIP_Alpha.exe.

2

Add a Model

Download a GGUF model (we recommend Mistral 7B for most systems) and place it in the models/ folder. SIP detects it automatically.

3

Start Talking

Launch SIP by double-clicking SIP_Alpha.exe. Tell SIP your name. It will remember — and everything that follows stays on your machine.

Download Now
psychology

System Requirements

check_circle

Minimum

  • Windows 10/11, macOS, or Linux
  • Python 3.11 or newer
  • 8 GB RAM minimum
  • No GPU required
  • 4 GB free disk space (+ model size)
  • CPU with 4+ cores
stars

Recommended

  • Windows 11, macOS, or Linux
  • 8 GB RAM or more
  • 10+ GB free disk space
  • CPU with 8+ cores
  • NVIDIA GPU (optional, improves speed)
rocket_launch

Optimal

  • Windows 11 (64-bit)
  • 16+ GB RAM
  • NVIDIA RTX GPU (8GB+ VRAM)
  • NVMe SSD storage
  • Enables larger, smarter models

Open Source Mission

SIP's code will be fully open source one year after the v1 release.

Until then, the Alpha is available on GitHub for review. Audit it. Break it. Tell us what you find.

lockGitHub — Public at v1 Release