John

🙌 hell yeah, much obliged

🙏 thanks Roger, I'll keep an eye on the issue - let me know if I can be of any help

https://github.com/Arize-ai/phoenix/issues/10257

Ah, I see - thanks for explaining

The frequent query is:

select
  phoenix.generative_models.name,
  phoenix.generative_models.provider,
  phoenix.generative_models.start_time,
  phoenix.generative_models.name_pattern,
  phoenix.generative_models.is_built_in,
  phoenix.generative_models.created_at,
  phoenix.generative_models.updated_at,
  phoenix.generative_models.deleted_at,
  phoenix.generative_models.id,
  token_prices_1.model_id,
  token_prices_1.token_type,
  token_prices_1.is_prompt,
  token_prices_1.base_rate,
  token_prices_1.customization,
  token_prices_1.id as id_1
from
  phoenix.generative_models
  left outer join phoenix.token_prices as token_prices_1 on phoenix.generative_models.id = token_prices_1.model_id
where
  phoenix.generative_models.deleted_at is null
order by
  phoenix.generative_models.name

which seems to be generated by SQLAlchemy from this function in the source code here 🔗. Via pg_audit, I've observed over 100,000 queries in the last 7 days, indicating a query about every ~5 seconds, with 62,000,000 rows processed. To be clear, there are no instances of the UI open anywhere. Can anyone help me understand why phoenix is querying the model info and prices every 5 seconds on the server, and if there's a way to stop this?

I was able to successfully use phoenix --no-ui serve to prevent serving the UI. However, I'm still seeing a ton of background queries (using a ton of DB egress) even in this "collector only" configuration. Reviving this thread with more details, if one of you folks would kindly take a look 🧵

Oh great, I'll give that a try! Thanks Roger 🙏

And if not possible as suggested by the bot, I would appreciate any creative workaround that come to mind 🙂 thanks