Point at gliner2-rs, which now holds this engine as a crate

The repository this card linked to has been folded into gliner2-rs as crates/gliner2-guardrails and is being archived; the link would have become a dead end. Updates the snippet for the workspace layout and adds the moderation helpers, including the decoding rule that never returns an empty list.

README.md

@@ -49,9 +49,11 @@ datasets:
 ONNX build of [`fastino/GLiNER2-Guardrails-PII-Multi`](https://huggingface.co/fastino/GLiNER2-Guardrails-PII-Multi),
 for running the model **without Python** at inference time.
 
-> **Rust engine: [github.com/dariofinardi/gliner2-guardrails-PII-Multi-onnx](https://github.com/dariofinardi/gliner2-guardrails-PII-Multi-onnx)**
-> The crate that consumes these files, along with the exporter that produced them
-> and the script that verifies them against PyTorch.
+> **Rust engine: [github.com/dariofinardi/gliner2-rs](https://github.com/dariofinardi/gliner2-rs)**
+> A Cargo workspace holding the engine, the exporter that produced these files and
+> the suite that verifies them against PyTorch. Use the `gliner2-guardrails`
+> crate for this checkpoint: it carries the moderation label sets with the
+> per-task thresholds the model expects.
 
 Converted and published by [Jugaad s.r.l.](https://edito-pdf.com), which uses it in
 production inside **Edito** and **Omissis** for GDPR-native document
@@ -133,9 +135,10 @@ tokenizer.json                                     15.3 MB
 ## Usage
 
 ```rust
-use gliner2_guardrails::{SchemaTask, SpanConfig, SpanEngine};
+use gliner2_core::{SchemaTask, SpanConfig, SpanEngine};
+use gliner2_guardrails::{Task, prompt_moderation_schema, verdict};
 
-gliner2_guardrails::init("my-app");
+gliner2_core::init("my-app");
 
 let mut engine = SpanEngine::new(SpanConfig::new("GLiNER2-Guardrails-PII-Multi-onnx"))?;
 let tasks = vec![SchemaTask::Entities(vec![
@@ -148,6 +151,18 @@ for e in engine.extract("Mario Rossi, mario.rossi@example.com, Cupertino.", &tas
 }
 ```
 
+For the guardrails side, `prompt_moderation_schema()` builds the three
+prompt-side tasks with the thresholds and single/multi-label settings this
+checkpoint was trained with, and `verdict()` applies gliner2's decoding rule —
+which never returns an empty list, falling back to the top-scoring label when
+nothing clears the threshold:
+
+```rust
+let out = engine.extract(prompt, &prompt_moderation_schema())?;
+println!("{:?}", verdict(&out, Task::PromptSafety));
+println!("{:?}", verdict(&out, Task::JailbreakDetection));
+```
+
 The engine picks the architecture and the best precision for the platform on its
 own. Byte offsets index the original text, so extracted spans keep their
 original casing — which matters when you are redacting a document rather than