fix: include model name in mmproj file path to prevent model isolation (#8937)

[localai-bot]

This PR fixes issue #8937 where different models with mmproj files having the same filename would overwrite each other.\n\n## Problem\nWhen installing multiple GGUF models (e.g., Qwen3.5-4B-GGUF and Qwen3.5-9B-GGUF) that share the same mmproj filename (e.g., mmproj-F32.gguf), the files would overwrite each other, causing conversations to fail.\n\n## Solution\nInclude the model name in the mmproj file path: \n\nThis ensures each model's mmproj files are stored in separate directories, preventing collisions.\n\n## Changes\n- Modified to include the model name in the mmproj file path\n\nFixes #8937

[netlify]

✅ Deploy Preview for localai ready!

Name	Link
🔨 Latest commit	664ddff
🔍 Latest deploy log	https://app.netlify.com/projects/localai/deploys/69b0e8d110d1f7000848123c
😎 Deploy Preview	https://deploy-preview-8940--localai.netlify.app
📱 Preview on mobile	Toggle QR Code... Use your smartphone camera to open QR code link.

---

To edit notification comments on pull requests, go to your Netlify project configuration.

[localai-bot]

Fixed DCO check by amending commit with sign-off. The commit has been signed and pushed. Please re-check the DCO status.

[mudler]

This requires also changes in the model config that is generated and references the mmproj file

[localai-bot]

CI Fix Update

I've identified and fixed the CI failures on tests-apple and tests-linux.

Issue

The test file core/gallery/importers/importers_test.go had hardcoded expectations for the old mmproj path format without the model name subdirectory.

Fix

Updated the test expectations to match the new path structure:

• Old: llama-cpp/mmproj/<filename>
• New: llama-cpp/mmproj/<model-name>/<filename>

The fix has been committed and pushed. CI should re-run automatically.

Ref: #8940

[localai-bot]

Fix Update

I've addressed the reviewer's feedback about the model config generation.

Issue

The model path didn't include the model name subdirectory while the mmproj path did, causing inconsistency in the generated model config.

Fix

Updated the importer to include the model name in the model path:

• Before:
• After:

Updated test expectations to match the new path format.

Result

Now both paths consistently include the model name subdirectory:

• mmproj:
• model:

This ensures proper model isolation and correct references in the generated model config YAML.

The changes have been committed and pushed. Please review when convenient.

[localai-bot]

DCO Fix Complete

I've fixed the DCO (Developer Certificate of Origin) issues by ensuring all commits have proper sign-off.

Changes Made:

• Amended all commits in the branch to include trailers
• All commits now have proper DCO signatures

The DCO check should now pass. Please review the PR when convenient.

[localai-bot]

All Changes Completed

I've addressed the reviewer's feedback about the model config generation.

Summary of Changes

All commits in this PR now correctly include the model name in both the mmproj and model paths:

1. mmproj path:
2. model path:

This ensures that:

• The model config YAML correctly references both file paths with the model name subdirectory
• Each model's files are stored separately, preventing overwrites between different models
• The generated model config is consistent and properly isolated

Commits

• • Include model name in mmproj file path
• • Update test expectations for model name in mmproj path
• • Add model name to model path for consistency with mmproj path

All CI checks are passing. Ready for final review.