Fix NameError: '_gpu_pool' is not defined in get_model()

[nexlabz]

Summary

• Fix NameError: name '_gpu_pool' is not defined when calling get_model()
• Use _get_gpu_pool() accessor instead of direct _gpu_pool reference

Problem

The /generate endpoint fails with:

NameError: name '_gpu_pool' is not defined
  File "backend/services/model_manager.py", line 326, in get_model
    model = await loop.run_in_executor(_gpu_pool, _load_model_sync)

Root Cause

_gpu_pool is a lazy module attribute only accessible via __getattr__ (lines 102-108) or the _get_gpu_pool() accessor (lines 90-98). Line 326 was accessing it directly, which bypasses the lazy initialization.

Line 357 already correctly uses _get_gpu_pool().

Fix

- model = await loop.run_in_executor(_gpu_pool, _load_model_sync)
+ model = await loop.run_in_executor(_get_gpu_pool(), _load_model_sync)

Test Plan

• Verified /generate endpoint now works after fix
• Generated audio successfully with voice design instructions

🤖 Generated with Claude Code

Summary by CodeRabbit

• Refactor
  • Improved internal consistency in background model loading task management to ensure reliable and uniform behavior across different model initialization scenarios.

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Pro Plus

Run ID: 53bcb4cf-71c2-459c-9efe-70d4a356bae2

📥 Commits

Reviewing files that changed from the base of the PR and between 651e63b and 01ff5c2.

📒 Files selected for processing (1)

• backend/services/model_manager.py

---

📝 Walkthrough

Walkthrough

The get_model() function in the model manager now obtains the GPU thread pool executor through the _get_gpu_pool() lazy accessor rather than direct _gpu_pool reference, aligning executor initialization with the existing preload_model() pattern.

Changes

GPU Executor Lazy Initialization

Layer / File(s)	Summary
Lazy pool accessor in get_model backend/services/model_manager.py	The get_model() function now calls loop.run_in_executor(_get_gpu_pool(), _load_model_sync) instead of passing _gpu_pool directly, ensuring the GPU thread pool is created and resolved through the module's lazy accessor only when the model load is triggered.

Estimated code review effort

🎯 1 (Trivial) | ⏱️ ~3 minutes

Poem

A rabbit hops through lazy code so bright,
Where pools now spawn just when they're needed right,
No eager threads await the dawn to break,
Both functions share one path, for goodness sake! 🐰

🚥 Pre-merge checks | ✅ 4 | ❌ 1

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Docstring Coverage	⚠️ Warning	Docstring coverage is 0.00% which is insufficient. The required threshold is 80.00%.	Write docstrings for the functions missing them to satisfy the coverage threshold.

✅ Passed checks (4 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title accurately and concisely describes the bug fix: replacing direct '_gpu_pool' reference with '_get_gpu_pool()' accessor to resolve the NameError.
Description check	✅ Passed	The description covers the summary, problem, root cause, fix with code diff, and test plan. However, it doesn't include the template sections: Type (Bug fix checkbox), Testing methodology details, and Checklist items.
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands and usage tips.}