Delete README.md

README.md

@@ -1,128 +0,0 @@
----
-title: Laya Decision Engine
-emoji: 🧠
-colorFrom: indigo
-colorTo: purple
-sdk: docker
-app_port: 7860
-pinned: false
----
-
-# 🧠 Laya Decision Engine — Self-Hosted Server (Jev-Compatible)
-
-Self-hosted, non-autoregressive **System 1 decision engine** running 24/7 on Hugging Face Spaces CPU. Evaluates typed decisions (`choice`, `score`, `noul`) over any text or state in a single forward pass (~200–400 ms on CPU) across 100+ languages with calibrated probabilities.
-
-Compatible with TypeSafe Jev's wire protocol (`POST /v1/systemone`).
-
----
-
-## ⚡ API Endpoints
-
-| Endpoint | Method | Description |
-|---|---|---|
-| `/` | `GET` | Plain text status (`"Laya is live."`) |
-| `/v1/systemone` | `POST` | Primary Jev / System-1 Decision API |
-| `/health` | `GET` | Health check & loaded model status |
-
----
-
-## 🚀 Quick Usage Example
-
-### 1. Evaluate a Decision (cURL)
-
-```bash
-curl -X POST "https://<YOUR-SPACE-NAME>.hf.space/v1/systemone" \
-  -H "Content-Type: application/json" \
-  -d '{
-    "state": "Hi, we were billed twice for March on invoice #4411. Refund please or we cancel our plan.",
-    "questions": {
-      "department": {
-        "type": "choice",
-        "instructions": "Which department should handle this request?",
-        "criteria": {
-          "billing": "invoices, payments, refund requests",
-          "technical": "system bugs, outages, crashes",
-          "sales": "upgrades, contracts, enterprise plans"
-        }
-      },
-      "urgency": {
-        "type": "score",
-        "instructions": "How urgent is this request on a 0 to 2 scale?",
-        "criteria": ["not urgent", "soon", "critical deadline or blocking issue"]
-      },
-      "churn_risk": {
-        "type": "noul",
-        "instructions": "Does the customer threaten to leave or cancel?"
-      }
-    }
-  }'
-```
-
-### 2. Python (Requests)
-
-```python
-import requests
-
-HF_SPACE_URL = "https://<YOUR-SPACE-NAME>.hf.space/v1/systemone"
-
-payload = {
-    "state": "The checkout page is throwing 500 errors and users cannot place orders!",
-    "questions": {
-        "severity": {
-            "type": "choice",
-            "instructions": "What is the severity of this ticket?",
-            "criteria": {
-                "P0_critical": "full outage, payment broken, revenue blocking",
-                "P1_high": "major feature failure",
-                "P2_normal": "minor inconvenience or general inquiry"
-            }
-        },
-        "page_oncall": {
-            "type": "noul",
-            "instructions": "Should we immediately alert the on-call engineering team?"
-        }
-    }
-}
-
-response = requests.post(HF_SPACE_URL, json=payload)
-data = response.json()
-
-print("Severity   :", data["answers"]["severity"]["choice"], f"({data['answers']['severity']['confidence']:.2f} conf)")
-print("Page OnCall:", "YES" if data["answers"]["page_oncall"]["noul"] > 0.5 else "NO")
-print("Router     :", data["routing"]["model"])
-```
-
----
-
-## 🔄 Automatic Updates & Factory Reboot
-
-1. **Automatic Boot Update (No action required)**:
-   - `start.sh` automatically checks PyPI for the newest version of `laya[serve]` every time the container boots, restarts, or wakes up from sleep.
-2. **Factory Reboot (Full Rebuild from Scratch)**:
-   - When a major new release of Laya drops, simply go to your Space's **Settings** $\rightarrow$ Click **"Factory reboot"**.
-   - Hugging Face will wipe the container and rebuild the image with zero cache, downloading the latest packages directly.
-
----
-
-## ⚡ Maximum Resource Utilization Profile (16 GB RAM / 2 vCPU)
-
-The container is configured to fully utilize Hugging Face Spaces' hardware allocation:
-* **Memory (16 GB RAM)**:
-  - Preloads all three checkpoints into RAM at startup: `english` (ModernBERT-large), `multilingual` (mmBERT-base), and `typed-decisions` (ModernBERT-large).
-  - Keeps all 3 hot with `LAYA_MAX_LOADED=3` (Zero eviction, ~3.5–4.0 GB resident RAM).
-  - No cold-boot or checkpoint swap delays when alternating between languages or schemas.
-* **Compute (2 vCPUs)**:
-  - Tuned with `torch.set_num_threads(2)` and OpenMP/MKL flags (`OMP_NUM_THREADS=2`, `MKL_NUM_THREADS=2`).
-  - Utilizes 100% of both CPU cores for vector matrix multiplication during single forward passes (~30–40ms).
-
----
-
-## 🛡️ Optional API Key Authentication
-
-To restrict access with a Bearer API token:
-1. Go to your Hugging Face Space **Settings** $\rightarrow$ **Variables and Secrets**.
-2. Add a Secret named `LAYA_API_KEY` with your secret key (e.g. `sk-laya-secret-key-12345`).
-3. Clients must then include:
-   ```http
-   Authorization: Bearer sk-laya-secret-key-12345
-   ```