Enhancement: Add PII redaction layer before LLM calls

[teodorofodocrispin-cmyk]

Problem

OwnPilot describes itself as "privacy-first" and processes
user messages from Telegram, WhatsApp, webhooks and other
channels. However, there is currently no layer that redacts
PII before the text reaches the LLM provider.

This means emails, phone numbers, private keys, and passwords
in user messages are sent directly to external LLM APIs —
contradicting the privacy-first promise.

Use Case

Any OwnPilot deployment processing real user messages needs
PII redacted before text reaches OpenAI, Anthropic, or any
other LLM provider. This is especially critical for:

• Telegram/WhatsApp bots handling sensitive conversations
• Webhook integrations receiving user-generated content
• Any pipeline where users share personal information

Proposed Solution

A pre-LLM sanitization hook that redacts PII automatically
before each LLM call. No human intervention required.

An existing tool that solves this is TrustBoost — a
blockchain-verified PII sanitization layer built for
autonomous agent pipelines:

• Webhook API — no SDK required
• Trial mode: tx_hash="TRIAL" (50 free requests, no payment)
• Returns: sanitized_text + safety_score + risk_category
• No raw PII stored

GitHub: https://github.com/teodorofodocrispin-cmyk/TrustBoost-PII-Sanitizer

Happy to help integrate this or discuss alternative approaches.