@baturhantasdelen/nexus-shield
v0.1.0
Published
Sub-10ms in-RAM PII redaction proxy for Vercel AI SDK, LangChain, and Node.js LLM apps.
Maintainers
Readme
⚡ Nexus Shield: Sub-10ms PII Guardrail for LLMs
Nexus Shield is an in-RAM pattern buffer proxy designed to strip PII (SSN, Credit Cards, API Keys, Emails) before forwarding payloads to OpenAI, LiteLLM, or LangChain—without killing stream TTL.
📊 Performance Comparison (300 Payloads Benchmark)
| Engine | Avg Latency (P50) | P99 Latency | Memory Overhead | | :--- | :---: | :---: | :---: | | Nexus Shield (In-RAM) | < 2.4 ms | < 6.1 ms | ~12 MB | | Standard Python Regex | 18.2 ms | 45.1 ms | ~45 MB | | MS Presidio (spaCy NER) | 120.5 ms | 245.0 ms | ~450 MB |
🚀 Quick Start
from nexus_shield import NexusClient
nexus = NexusClient(base_url="https://api.nexusshield.ai/v1", api_key="nx_live_...")
config = nexus.get_proxy_config()