Redact is a real-time security platform powered by artificial intelligence dedicated to protecting LLM systems against prompt injection attacks. Designed and developed by Q2BSTUDIO, a custom software and application development company specializing in artificial intelligence, cybersecurity, AWS and Azure cloud services, and business intelligence services, Redact operates as a secure middleware between users and LLM applications, analyzing, classifying, and sanitizing prompts before they reach the target model.
Main features: real-time detection of prompt injection attempts, multi-layer defense against varied attack vectors, semantic analysis of suspicious patterns, instant feedback and attack visualization, seamless integration with AI workflows, and compatibility with custom solutions.
Detection and semantic analysis: Redact uses embeddings and sentence transformer models to detect malicious modifications in prompts based on meaning and context, not just literal matches. This enables identification of sophisticated manipulation attempts and protects the business logic of applications that use artificial intelligence and AI agents.
Semantic cache: implementation of a semantic cache that stores embeddings of processed prompts to avoid redundant calculations and speed up responses. Use of cosine similarity with configurable threshold for cache lookups, storage with namespaces for different cache types such as embeddings and responses.
Rate limiting and throttling: multidimensional traffic control with per-user, per-IP, and global limits, sliding windows for accurate request counting, and automatic cleanup of expired windows. Ideal for protecting critical endpoints and preventing targeted attacks on content generation services or internal APIs.
Streaming processing: prompt queue based on Redis Streams for asynchronous processing of security checks, enabling horizontal scaling of workers and event auditing for compliance and traceability.
Connection management: configurable connection pool, automatic reconnection, and health checks to ensure resilience and performance in production and cloud environments, both on AWS and Azure.
How we use Redis 8: Redis 8 is integrated at the heart of the architecture for semantic caching (embeddings), processing with Streams, Pub/Sub for real-time alerts, hashes and sorted sets for metadata and limit management, and structures optimized for high performance in enterprise AI environments.
Technical stack: Frontend with React, TypeScript, and Tailwind CSS; Backend with FastAPI and Python; AI and ML with Sentence Transformers model all-MiniLM-L6-v2 and acceleration with Cerebras Inference; Caching and messaging with Redis 8 Streams Pub/Sub Hashes Sorted Sets; deployment on Vercel; integrated monitoring and metrics for tracking limits and events.
Demo and code: demo available at https://redact-llm.vercel.app and repository at https://github.com/VaishakhVipin/Redact-LLM for reference and testing.
Benefits for businesses: Redact protects custom applications and custom software that integrate artificial intelligence, reduces risks of information leakage and model manipulation, and facilitates regulatory compliance. It is especially useful for enterprise AI projects, AI agents, and solutions that use Power BI for visualization and analysis where the integrity of input data is critical.
Future improvements: integration with Redis Search for advanced attack pattern queries, use of RedisTimeSeries for detailed metrics, adaptive similarity thresholds for semantic cache, deployment of RedisAI for edge inference, and event sourcing with Redis Streams for complete audits.
About Q2BSTUDIO: Q2BSTUDIO is a company dedicated to custom software and application development, specializing in artificial intelligence, cybersecurity, and migrations to AWS and Azure cloud services. We offer business intelligence services, custom AI agents, AI consulting for businesses, and integrated solutions with Power BI to maximize data value. If you are looking for assistance integrating Redact as a security solution in your custom applications or implementing custom software projects focused on cybersecurity and artificial intelligence, contact the Q2BSTUDIO team for specialized consulting.
Keywords for positioning: custom applications, custom software, artificial intelligence, cybersecurity, AWS and Azure cloud services, business intelligence services, AI for businesses, AI agents, Power BI.
