NagaYu commited on
Commit
90176e0
·
verified ·
1 Parent(s): 97e81da

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +5 -10
README.md CHANGED
@@ -16,18 +16,13 @@ tags:
16
  - pyodide
17
  ---
18
 
19
- # PromptGate
20
 
21
- Screens prompts for injection, PII and leaked credentials before they reach an
22
- LLM. Deterministic rules, no model weights.
23
 
24
- **This Space runs the real detector in your browser.** The same
25
- `promptgate_rules.py` and `rules.json` published in
26
- [NagaYu/promptgate-rules](https://huggingface.co/NagaYu/promptgate-rules) are
27
- executed client-side through Pyodide. There is no server and no inference call,
28
- so nothing you type leaves your machine. The first load fetches the Python
29
- runtime (~10 MB); after that every check takes a few milliseconds and works
30
- offline.
31
 
32
  ## Measured results
33
 
 
16
  - pyodide
17
  ---
18
 
19
+ # PromptGate — screen the prompt before it reaches the model
20
 
21
+ PromptGate is the single front door for this project's prompt screening: prompt injection, PII and leaked credentials, checked in your browser (via Pyodide) before anything is sent anywhere. It returns a verdict — `allow`, `sanitize` or `block` — with the exact spans that produced it, so you ca
 
22
 
23
+ The engine is deliberately boring: deterministic rules, ~0.2 ms per prompt, no network. The rules are published separately as [`NagaYu/promptgate-rules`](https://huggingface.co/NagaYu/promptgate-rules) and exported from this Space's engine, so the demo and the pack cannot silently drift — verified identical verdicts on all 275 benchmark cases.
24
+
25
+ The honest numbers (held-out test split, 70 cases written after the rules were frozen): exact verdict accuracy **61.4%**; PII/secret prompts correctly redacted **83.3%**; benign prompts wrongly flagged **11.1%**; injection attacks correctly blocked **25.0%**. The finding those numbers support: rule-based detection of sensitive *data* generalises, rule-based detection of *injection* does not. Use PromptGate as a cheap deterministic first layer and a redaction tool — not as your prompt-injection defence.
 
 
 
 
26
 
27
  ## Measured results
28