Back
AI tooling for domain corpora

_ + attia store

1 min readZi Wang
Intent

This post explores wiring up domain-specific AI file stores—medical podcast transcripts and live sports commentary—as practical experiments in retrieval and multimodal generation.

Signal

A file store gated behind IP whitelisting, retrieval over a curated transcript corpus, and model selection (MedGemma vs MedLM vs Gemini audio/Flash) chosen per task—medical citation extraction vs live game commentary.

Question

How do you make a domain-specific corpus (Attia transcripts, NBA box scores) accessible and usefully queryable through AI models, and what access and modality choices does that require?

Topics

AI file stores / retrieval · Peter Attia podcast transcripts · model selection (MedGemma/MedLM/Gemini) · IP whitelisting / access control · NBA box-score-to-commentary generation · multimodal / live audio commentary

Z / Runner 🏃🏻‍♂️

  • 🏃🏻‍♂️ not seeing the file, assume you whitelisted ipv4 "70.134.61.57" ipv6 "2600:1700:3ec2:33f:f00a:fbf6:3c62:a672"

  • 🏃🏻‍♂️ drive link.

Stephen / Basketball 🏀

🏀 /attia = file store of all his podcast transcripts.

  • where's the google drive link for all attia's episode notes? vs your manus page.
  • medgemma vs medlm vs pdf understanding vs url context: "quote the paragraph on homocysteine and tmg from peterattiamd.com"

🏀 "describe in 5 paragraphs of this fantastic finish."

  • "The final minute features a duel between the teams' superstars. Edwards briefly reclaims the lead for Minnesota with a driving layup, reaching 42 points on the night. However, Jamal Murray responds with a massive three-pointer off a Jokić screen to put Denver ahead 131-130. On the ensuing possession, the Nuggets' defense swarms Edwards with a double team, forcing a crucial turnover. Jokić capitalizes on the other end with free throws, securing a 50-point triple-double and extending Denver's lead."
  • + gemini audio or flash native for (live) commentary.

Select any passage to copy a prompt · esc to return