GitHub user awdemos added a comment to the discussion: [DISCUSS] New Agentic 
lakehouse module & built-in AI agent gateway

Love the direction. Making every Hudi table queryable by agents and keeping it 
self-hostable (Ollama by default) is the right default for data platforms.

One thing to design in early: cost predictability for the agent loop. Local 
inference is free-ish, but once deployments switch to API providers or vLLM 
clusters, per-token metered costs can surprise operators. A "cost ceiling" or 
flat-rate backend option—where heavy agent loops run on a fixed monthly GPU 
environment—would make the gateway more palatable for production teams.

We run UltraWork on that flat monthly model 
(https://vibecodingagency.com/gpu-cloud/) for exactly this reason: teams want 
agentic capability without variable inference bills. Disclosure: I help run 
Vibe Coding Agency. Happy to share the budget-vs-telemetry patterns we've used 
if useful.

GitHub link: 
https://github.com/apache/hudi/discussions/19264#discussioncomment-17650783

----
This is an automatically sent email for [email protected].
To unsubscribe, please send an email to: [email protected]

Reply via email to