- Who it helps
- Developers evaluating an additional redaction layer for model requests
- Availability
- Public implementation uses Python 3.11, local Ollama entity recognition and Tesseract OCR. Model downloads, memory and upstream API access are required.
Inspect what would leave the machine
The README describes a dry-run preview and recommends failing closed when processing fails. Its mapping mechanism supports text, images and PDFs.
Detection can miss information
A local detector may fail to identify sensitive content. Restoring substitutions also depends on response structure. The proxy does not establish that a cloud request is anonymous or safe for every document.
What you can do
- Read the primary description and its stated limits before planning an evaluation. Open reference ↗
Where the evidence stops
Documentation checked November 11. No proxy, detector, OCR or leakage test executed.
Sources read. Public source text checked November 11, 2026 UTC. No third-party project or full workflow executed.
Sources & further reading
GitHub · implementation
aiproxy redaction proxy ↗
README explains local entity recognition and reversible mappings; no privacy or leakage tests run. Three stars visible.
cucooOnGithub · page checked 2026-11-11
3 stars · observed 2026-11-11 16:05 UTC
Coverage note: Documentation checked November 11. No proxy, detector, OCR or leakage test executed.
The record over time (1)
Dated source checks and published explanations. An observation date can differ from the original release date.
- · Added to Current
aiproxy is a local proxy that replaces detected sensitive content before forwarding a request to a model service, then restores mapped values in the response.