fix: wire UnifiedClassifier to ExtProc router for consistent LoRA-based classification

[yossiovadia]

Summary

This PR fixes inconsistent model selection behavior by ensuring the ExtProc router uses the same UnifiedClassifier (LoRA-based) instance as the Classification API.

Problem

• Classification API (port 8080): Used UnifiedClassifier (LoRA models) ✅
• ExtProc router (port 8801): Used legacy Classifier (traditional BERT) ❌
• This caused different classification results for the same query, leading to incorrect model selection in category-based routing

Example Issue

Math queries were classified correctly by the API but routed to the wrong model through Envoy:

Query: "What is 234 + 567?"

Classification API (8080):
- Category: math, Confidence: 0.886 ✅
- Expected model: Model-B (score 1.0)

Envoy Routing (8801):
- Actual model: Model-A ❌
- Reason: Using uninitialized legacy classifier → fallback to default

Root Cause

In router.go, the UnifiedClassifier was never wired to the legacy Classifier:

// Before fix - UnifiedClassifier not connected ❌
autoSvc, err := services.NewClassificationServiceWithAutoDiscovery(cfg)
if err != nil {
    services.NewClassificationService(classifier, cfg)  // Not stored
} else {
    _ = autoSvc  // UnifiedClassifier never wired to Classifier
}

Solution

Wire the UnifiedClassifier to enable delegation:

// After fix - UnifiedClassifier properly wired ✅
var classificationSvc *services.ClassificationService
autoSvc, err := services.NewClassificationServiceWithAutoDiscovery(cfg)
if err != nil {
    classificationSvc = services.NewClassificationService(classifier, cfg)
} else {
    classificationSvc = autoSvc
    if classificationSvc.HasUnifiedClassifier() {
        unifiedClassifier := classificationSvc.GetUnifiedClassifier()
        if unifiedClassifier != nil {
            classifier.UnifiedClassifier = unifiedClassifier  // Now wired!
        }
    }
}

Changes

• router.go: Wire UnifiedClassifier to Classifier during initialization, store ClassificationService reference
• classifier.go: Add UnifiedClassifier field, delegate to it before trying in-tree classifier, add classifyWithUnifiedClassifier() helper method
• classification.go: Add GetUnifiedClassifier() getter method

Testing

Before fix:

Query: What is 789 + 123?
Category: math, Confidence: 0.886
Expected: Model-B
Actual: Model-A ❌ FAIL

After fix:

Query: What is 789 + 123?
Category: math, Confidence: 0.896
Expected: Model-B
Actual: Model-B ✅ PASS

Related Issues

Closes #640

[netlify]

✅ Deploy Preview for vllm-semantic-router ready!

Name	Link
🔨 Latest commit	62e4158
🔍 Latest deploy log	https://app.netlify.com/projects/vllm-semantic-router/deploys/691a915fd43289000847fd1a
😎 Deploy Preview	https://deploy-preview-641--vllm-semantic-router.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.

[github-actions]

👥 vLLM Semantic Team Notification

The following members have been identified for the changed files in this PR and have been automatically assigned:

📁 src

Owners: @rootfs, @Xunzhuo, @wangchen615
Files changed:

• src/semantic-router/pkg/classification/classifier.go
• src/semantic-router/pkg/extproc/router.go
• src/semantic-router/pkg/services/classification.go

---

🎉 Thanks for your contributions!

This comment was automatically generated based on the OWNER files in the repository.