Ourahou O.

Commented on Managing trace retention in Phoenix and trade-offs...·Posted inDiscussions

Thanks a lot, When you say cold storage, what’s a go to solution here? classic db (Mongodb like) with indexation for easy retrieval?

Posted in Discussions·

Ourahou O.

Managing trace retention in Phoenix and trade-offs of moving from self-hosted Phoenix to Arize

Hey everyone, two quick questions:

1.
Retention & traces: How are you managing trace retention in Phoenix? Is the standard practice to stick with the 15-day retention window? If you need to debug something older, how do you handle that? Do you export traces elsewhere, increase retention, or rely on another storage layer?
2.
Moving from self-hosted Phoenix to Arize Could someone give me a clear TL;DR on the trade-offs? Context:

~100k traces/month
Multiple AI agents & Multimodal inputs
Evaluations running & Prompt versioning in place

What do we concretely gain by moving to Arize vs staying self-hosted?

3Comments

Posted in Introductions·

Ourahou O.

Seeking Guidance on Building Eval Pipelines in Phoenix

Hey everyone! 👋 Ouss here, founder of Palete (We deploy AI agent use cases on top of whatsapp for B2B). I’ve been using Phoenix pretty heavily for the last ~12 months, mostly for tracing, and it’s been super valuable. I’m still struggling a bit with building solid eval pipelines on top of it though. I’d love guidance on either:

clear resources / examples for eval workflows, or
chatting with someone who’s already doing this well

If anyone has pointers, repos, or is open to a quick discussion, I’d really appreciate it 🙏

1Comment