LLM Middleware Demo

See how RedactiPHI protects PHI when using LLMs. Clinical notes are de-identified before sending to the AI, then re-identified in the response.

Step 1
Original Note
→
Step 2
De-identify
→
Step 3
Send to LLM
→
Step 4
Re-identify
How it works: PHI (names, dates, etc.) is replaced with tokens before the LLM sees it. The LLM response is then re-identified, restoring original values. This keeps PHI safe while still getting useful AI responses.
🎭
DEMO MODE
Using simulated LLM responses. Perfect for demonstrations - no API keys required!
📝 Original Clinical Note
Contains PHI
🔒 De-identified (Safe for LLM)
PHI Removed
Click "De-identify" to process the note...