fix: prevent duplicate LM Studio models by only showing loaded models

[roomote]

Summary

This PR fixes the issue where LM Studio models appeared twice in the Provider Configuration Profile when they were both downloaded and loaded.

Problem

The previous implementation fetched models from two different LM Studio APIs:

• listDownloadedModels() - showed all models on disk
• listLoaded() - showed models currently in memory

This caused duplicates when a model was both downloaded and loaded, with inconsistent keys (model.path vs model.modelKey).

Solution

As suggested by @daniel-lxs, this PR simplifies the implementation by:

• Removing the listDownloadedModels() call entirely (lines 74-83)
• Only showing loaded models - these are the only ones that can actually process requests
• Using model.path as the consistent key to prevent any potential duplicates
• Updating tests to reflect the simplified implementation

Benefits

• ✅ Eliminates duplicates entirely (no deduplication logic needed)
• ✅ Only shows models that can actually process requests
• ✅ Provides accurate runtime metadata (contextLength vs maxContextLength)
• ✅ Simpler, more maintainable code
• ✅ Better performance (one less API call)

UX Change

Users will need to load models in LM Studio before they appear in Roo Code. This is already required to use them anyway, so it aligns the UI with the actual functionality.

Testing

• All existing tests have been updated and pass
• Removed tests for listDownloadedModels functionality
• Updated test to verify that loaded models use model.path as the key

Fixes #6954

---

Important

Fixes duplicate LM Studio models by removing listDownloadedModels() and using model.path as the key in lmstudio.ts.

• Behavior:
  • Removes listDownloadedModels() call in lmstudio.ts, only using listLoaded() to fetch models.
  • Uses model.path as the key to prevent duplicates.
• Testing:
  • Updates tests in lmstudio.test.ts to reflect removal of listDownloadedModels().
  • Verifies that loaded models use model.path as the key.
• Misc:
  • Users must load models in LM Studio for them to appear in Roo Code.

^{This description was created by for 6f7c4bc. You can customize this summary. It will automatically update as commits are pushed.}

[roomote]

Reviewing my own code is like debugging in a mirror - everything looks backwards but the bugs are still mine.

[roomote]

Good simplification! Using model.path as the consistent key should effectively prevent duplicates. However, I notice that the forceFullModelDetailsLoad function (not in this diff) still uses modelId to add to modelsWithLoadedDetails. Should that be updated to use the model path for consistency?

[roomote]

Nice change to only fetch loaded models. This makes sense since only loaded models can actually process requests. The comment clearly explains the rationale.

[roomote]

Good test update! The assertion correctly verifies that we're now using model.path as the key instead of modelKey.

[roomote]

Excellent test name update that clearly describes the new behavior - fetching only loaded models and using model.path as the key.

[mechanicmuthu]

Not a good feature to have:

LM Studio, when configured with Just-in-Time (JIT) Model Loading, can automatically load an unloaded model upon receiving an API call that requests that specific model.

Since 2024 october this change is effective. https://lmstudio.ai/blog/lmstudio-v0.3.5#on-demand-model-loading

Also there is a auto model unload on 60 minutes timeout. So, it is better to show all models with simple deduplication.

[daniel-lxs]

Closing this PR based on feedback from @mechanicmuthu about LM Studio's JIT Model Loading feature (available since v0.3.5). The current approach of only showing loaded models doesn't align well with the JIT loading capability where models can be automatically loaded on-demand. Will open a new PR with a simpler deduplication approach that shows all models while preventing duplicates.