unusual_typo

3 ms·

1. 1.▲
   by unusual_typo 4mo ago

   Here are the benchmark results. You can check more details in the repo. openai/privacy-filter on Apple M1 Max dtype 1k total 1k tok/s 8k total 8k tok/s ━━━━━━━━━━━━━━━━ ━━━━━━━━━━━ ━━━━━━━━━━
2. 2.▲
   by unusual_typo 4mo ago

   I just ran the benchmark on my macbook. 582 ms for 1k tokens and 4.64 s for 8k
3. 3.▲
   by unusual_typo 4mo ago

   It use openai/privacy-filter which is smaller than 1GB in size. I haven't checked usage during inference. It rans at 1k toks/sec on my macbook. I will update the repo with benchmark results. Thanks for the comment
4. 4.▲
   by unusual_typo 4mo ago

   I wanted to implement the feature initially. i realized that it requires modification of coding agents (eg codex, claude code, opencode etc). hook or skills pass PII data into server eventually so i decided to share the standalone app first
5. 5.▲
   Show HN: Local personal data redaction for any AI tools(github.com)

   12 points by unusual_typo 4mo ago|7 comments
6. 6.▲
   by unusual_typo 5mo ago

   I shipped an enhanced fork of Parlor (by Fikri Karim https://news.ycombinator.com/item?id=47652007 ) that reads various visual inputs and uses Supergemma 4 E4B + Supertonic TTS to run a fully local, multimodal, and multiling
7. 7.▲
   Show HN: Parlor Jarvis – Realtime AI (audio+screen in, voice out) & multilingual(github.com)

   8 points by unusual_typo 5mo ago|2 comments