Redact sensitive data from LLM prompts using a local model before transmitting to external LLMs.
Sponsored
Safeguard sensitive information when interacting with external Large Language Models by intercepting and redacting data locally before it leaves your machine. This middleware automatically replaces employee names, project codenames, infrastructure details, and other PII with typed placeholders using an on-premise model like Ollama or LM Studio. The redacted data is then restored in the LLM's response, ensuring that external services never see the original sensitive values while maintaining a natural conversation flow for the user, mitigating the risk of data leakage and intellectual property exposure.
Key Features
01Local data redaction with Ollama or LM Studio
02Secure restoration of original data in LLM responses
03Automatic sensitive data replacement with typed placeholders
041 GitHub stars
05Comprehensive redaction categories including PII, credentials, and internal info
06Configurable detection rules for enhanced accuracy
Use Cases
01Ensuring compliance and privacy when leveraging cloud-based AI
02Maintaining confidentiality of PII, credentials, and proprietary information in AI interactions
03Protecting sensitive internal data from exposure to third-party LLMs
