Take 2 - Working on local-ai-search project
Summary
On this Wednesday AI stream, I kicked off the local-ai-search project — a local image search app powered by Reka Edge, a vision model running on my home server. Before writing any code, I used my new grill-me skill in OpenCode to stress-test my architecture decisions (Avalonia over Uno, Turso for DB, single folder, no thumbnails), then let the PRD-to-issues skill generate all the GitHub issues automatically. We wrapped up by validating the full pipeline with a quick .NET demo — sending images to the local Reka model and getting descriptions back, including one that called me “elderly” (it was the lighting!). 📺 - Twitch archive - stream no. 404
Replay
Project
All the code for this project is available on GitHub: local-ai-search - https://github.com/FBoucher/local-ai-search
Game Results
- @codebymistakes: 92.7
- @surlydev: 90.31
- @fboucheros: 75.78
- @lurkydev: 74.71
- @theunoriginaljerk: 73.81
- @therealsurlybot: 71.97
- @groversaurus: 71.09
- @garysgilet: 68.25
- @my_friend_asks: 65.31
- @undefined_process: 52.01
- @marcipopsis: 42.36
- @jeffs_hat_stand: 23.47
- @a_friend_asks: 23.24
- @fredda_the_cat: 22.83
- @jtsom: 21.33
- @phrakberg: 14.97
- @mcnets: 0
- @tbdgamer: 0
- @coppersbeard: 0
- @procrastoholik: 0
- @cmgbeanie: 0
- @thecliptographer: 0
- @dj_surly: 0
Statistics
- 🏆Best score: @codebymistakes with 92.7
- 😭Biggest loser: @coppersbeard with 2 drops and no high score
- 🍀Luckiest: @codebymistakes with best score 92.7 and only 2 drops
- 🎖️Super participant: @fredda_the_cat with 10 drops
Notes/ References / Snippets
- Model: https://huggingface.co/RekaAI/reka-edge-2603
- Grill-Me Skill https://www.aihero.dev/my-grill-me-skill-has-gone-viral