refactor(http,llm): Using new Logger module from core

[szanata]

Summary

• Refactored LLM and HTTP to use new Logger from Core module in place of console.*.

Test plan

• Unit tests refactored
• Manually tested

[bnchrch]

Main thing: the refactor dropped the log assertions instead of pointing them at Logger, so a few "we warn on X" behaviors are now untested (inline). Couple of nits too.

[bnchrch]

🧪 The swap dropped the log checks instead of moving them to Logger, so we no longer test that these warnings fire:

• Failed to fetch models pricing and Missing cost reference for model (this file)
• the error log in the catch (this file)
• the no-cache / stale-cache warnings in fetch_models_pricing.spec.js
• the snake_case warning in validations.spec.js
• the warn / no-warn checks in http/cost.spec.ts

In tests Logger.* falls back to console.log, so a console.log spy works, or mock @outputai/core's Logger. If we've decided log lines aren't worth testing that's fine, I'd just want it on purpose, not a side effect of the swap.

[szanata]

I changed on purpose, as I dont think those are worth checking...

[bnchrch]

💅 @outputai/core gets imported twice now (line 1 and here). Merge them:

import { ValidationError, z, Logger } from '@outputai/core';

[szanata]

My bad

[bnchrch]

❓ Logger has no namespace, so these land under the generic Activity logs with nothing saying they came from llm cost. validations.js keeps its [output-llm] prefix but these don't, so it's a bit inconsistent. Worth a short prefix here too so they stay traceable? Small.

[szanata]

I agree that this isnt ideal. I am adding namespace overwriting on the Logger.

[mintlify]

Preview deployment for your docs. Learn more about Mintlify Previews.

Project	Status	Preview	Updated (UTC)
outputai	🟢 Ready	View Preview	Jun 25, 2026, 5:24 PM

💡 Tip: Enable Workflows to automatically generate PRs for you.

[bnchrch]

❓ wdyt about being able to apply name spaces across all instances? e.g.

HttpLogger = new Logger("http")
HttpLogger.warn("foo")

or

HttpLogger = Logger.createLogger("http")
HttpLogger.warn("foo")

Seems both more familiar, prevent us from having to remember in each package at each log line what name space to add, and would prevent having the collission issue with the key "namespace:"

[szanata]

I would steer away from both. I think the is very convenient to be able to just Logger.*() at any point, without dealing with instances. We could have a .bindNamespace() module that returns a logger "instance" with the namespace already set.

[szanata]

Done. Make de the default Logger an instance with a .createLogger() that creates another instance with bound namespace.