Skip Lakera check for educational content to avoid false positives

Implement two-layered security filter in backend

Add hate speech pre-screening (Approach A & B)

Add comprehensive injection detection patterns

Add phone, IBAN, password, and DB connection PII patterns

Add injection patterns for attack demonstration requests

Add comprehensive PII detection patterns

Increase max tokens to 16384 and disable rate limit for testing

Improve safety check resilience and logging

Increase default max_tokens to 2048

Initial commit: Enterprise-AI-Gateway - Secure LLM gateway