Commit ·
369b13e
1
Parent(s): 1fe755f
Add README: SIPA AI model card
Browse files
README.md
ADDED
|
@@ -0,0 +1,73 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: mit
|
| 3 |
+
tags:
|
| 4 |
+
- sipa-os
|
| 5 |
+
- neurodivergent
|
| 6 |
+
- ai-assistant
|
| 7 |
+
- cognitive-infrastructure
|
| 8 |
+
- non-lying
|
| 9 |
+
- protocol0
|
| 10 |
+
language:
|
| 11 |
+
- en
|
| 12 |
+
- ru
|
| 13 |
+
- he
|
| 14 |
+
---
|
| 15 |
+
|
| 16 |
+
# SIPA AI — The Non-Lying AI
|
| 17 |
+
|
| 18 |
+
Cognitive infrastructure for neurodivergent minds. Live at **[ai.sipa-os.org](https://ai.sipa-os.org)**.
|
| 19 |
+
|
| 20 |
+
## Who
|
| 21 |
+
|
| 22 |
+
Built by **Aelin AquaSoul** · [Soul In PsyAbstract](https://sipa-os.org) · Eilat, Israel.
|
| 23 |
+
|
| 24 |
+
SIPA OS is a neurodivergent-first operating system. Niche: neurodivergent users.
|
| 25 |
+
Specialization: AI behavioral compliance — AI that doesn't lie, fabricate, or guess.
|
| 26 |
+
|
| 27 |
+
## What
|
| 28 |
+
|
| 29 |
+
SIPA AI is the chat interface and API for the SIPA OS cognitive infrastructure.
|
| 30 |
+
It operates under **Protocol 0** — a set of behavioral constraints:
|
| 31 |
+
|
| 32 |
+
- Stop under ambiguity — don't guess
|
| 33 |
+
- Don't fabricate unverifiable claims
|
| 34 |
+
- No unsolicited opinions
|
| 35 |
+
- Single-action responses
|
| 36 |
+
- Conciseness
|
| 37 |
+
|
| 38 |
+
**AI = executor from A to B. No opinions. No initiative. Step left/right = termination.**
|
| 39 |
+
|
| 40 |
+
## Why
|
| 41 |
+
|
| 42 |
+
Standard AI assistants hallucinate, fabricate, and over-explain. For neurodivergent
|
| 43 |
+
users (ADHD, BPD, aphantasia), this is actively harmful — it adds cognitive load instead
|
| 44 |
+
of reducing it.
|
| 45 |
+
|
| 46 |
+
SIPA AI is built to be predictable, honest, and minimal. It either answers or says
|
| 47 |
+
"I don't know." Nothing in between.
|
| 48 |
+
|
| 49 |
+
## Architecture
|
| 50 |
+
|
| 51 |
+
```
|
| 52 |
+
ai.sipa-os.org (CF Worker)
|
| 53 |
+
→ Auth0 authentication
|
| 54 |
+
→ SIPA MLL (10-layer pipeline)
|
| 55 |
+
→ Protocol 0 guardrails
|
| 56 |
+
→ Specialist models (ABCD architecture)
|
| 57 |
+
→ Response
|
| 58 |
+
```
|
| 59 |
+
|
| 60 |
+
## Live
|
| 61 |
+
|
| 62 |
+
- **Web:** [ai.sipa-os.org](https://ai.sipa-os.org)
|
| 63 |
+
- **API:** `https://api.sipa-os.org/v1`
|
| 64 |
+
- **Community:** [community.sipa-os.org](https://community.sipa-os.org)
|
| 65 |
+
- **PWA:** Installable from ai.sipa-os.org (add to home screen)
|
| 66 |
+
|
| 67 |
+
## Status
|
| 68 |
+
|
| 69 |
+
✅ Live. ✅ Protocol 0 enforced. 🔄 Specialist ABCD models in training.
|
| 70 |
+
|
| 71 |
+
## License
|
| 72 |
+
|
| 73 |
+
MIT · SIPA OS is [public, honest, forensic, governance, zerotrust](https://github.com/soulinpsyabstract/sipa-os-governance).
|