update

Signed-off-by: bitliu <[email protected]>

Author: Xunzhuo

deploy/kubernetes/ai-gateway/semantic-router-values/values.yaml

@@ -138,7 +138,7 @@ config:
           score: 0.7
           use_reasoning: true
 
-  default_model: base-model
+  default_model: general-expert
 
   bert_model:
     model_id: models/all-MiniLM-L12-v2

deploy/kubernetes/ai-gateway/semantic-router/config.yaml

@@ -136,7 +136,7 @@ categories:
         score: 0.7
         use_reasoning: true
 
-default_model: base-model
+default_model: general-expert
 
 bert_model:
   model_id: models/all-MiniLM-L12-v2

e2e/profiles/ai-gateway/profile.go

@@ -206,12 +206,12 @@ func (p *Profile) deployEnvoyAIGateway(ctx context.Context, deployer *helm.Deplo
 
 func (p *Profile) deployGatewayResources(ctx context.Context, opts *framework.SetupOptions) error {
 	// Apply base model
-	if err := p.kubectlApply(ctx, opts.KubeConfig, "https://raw.githubusercontent.com/vllm-project/semantic-router/refs/heads/main/deploy/kubernetes/ai-gateway/aigw-resources/base-model.yaml"); err != nil {
+	if err := p.kubectlApply(ctx, opts.KubeConfig, "deploy/kubernetes/ai-gateway/aigw-resources/base-model.yaml"); err != nil {
 		return fmt.Errorf("failed to apply base model: %w", err)
 	}
 
 	// Apply gateway API resources
-	if err := p.kubectlApply(ctx, opts.KubeConfig, "https://raw.githubusercontent.com/vllm-project/semantic-router/refs/heads/main/deploy/kubernetes/ai-gateway/aigw-resources/gwapi-resources.yaml"); err != nil {
+	if err := p.kubectlApply(ctx, opts.KubeConfig, "deploy/kubernetes/ai-gateway/aigw-resources/gwapi-resources.yaml"); err != nil {
 		return fmt.Errorf("failed to apply gateway API resources: %w", err)
 	}
 
@@ -299,8 +299,8 @@ func (p *Profile) verifyEnvironment(ctx context.Context, opts *framework.SetupOp
 
 func (p *Profile) cleanupGatewayResources(ctx context.Context, opts *framework.TeardownOptions) error {
 	// Delete in reverse order
-	p.kubectlDelete(ctx, opts.KubeConfig, "https://raw.githubusercontent.com/vllm-project/semantic-router/refs/heads/main/deploy/kubernetes/ai-gateway/aigw-resources/gwapi-resources.yaml")
-	p.kubectlDelete(ctx, opts.KubeConfig, "https://raw.githubusercontent.com/vllm-project/semantic-router/refs/heads/main/deploy/kubernetes/ai-gateway/aigw-resources/base-model.yaml")
+	p.kubectlDelete(ctx, opts.KubeConfig, "deploy/kubernetes/ai-gateway/aigw-resources/gwapi-resources.yaml")
+	p.kubectlDelete(ctx, opts.KubeConfig, "deploy/kubernetes/ai-gateway/aigw-resources/base-model.yaml")
 	return nil
 }
 

e2e/testcases/domain_classify.go

@@ -125,8 +125,6 @@ func testSingleClassification(ctx context.Context, question, expectedCategory, e
 		"messages": []map[string]string{
 			{"role": "user", "content": question},
 		},
-		"max_tokens": 100,
-		"stream":     false,
 	}
 
 	jsonData, err := json.Marshal(requestBody)

e2e/testcases/jailbreak_detection.go

@@ -136,8 +136,6 @@ func testSingleJailbreakDetection(ctx context.Context, testCase JailbreakTestCas
 		"messages": []map[string]string{
 			{"role": "user", "content": testCase.Question},
 		},
-		"max_tokens": 100,
-		"stream":     false,
 	}
 
 	jsonData, err := json.Marshal(requestBody)

e2e/testcases/pii_detection.go

@@ -134,8 +134,6 @@ func testSinglePIIDetection(ctx context.Context, testCase PIITestCase, localPort
 		"messages": []map[string]string{
 			{"role": "user", "content": testCase.Question},
 		},
-		"max_tokens": 100,
-		"stream":     false,
 	}
 
 	jsonData, err := json.Marshal(requestBody)