chore: cleanups

ee/query-service/app/server.go

@@ -106,7 +106,7 @@ func NewServer(config signoz.Config, signoz *signoz.SigNoz) (*Server, error) {
 	// initiate agent config handler
 	agentConfMgr, err := agentConf.Initiate(&agentConf.ManagerOptions{
 		Store:         signoz.SQLStore,
-		AgentFeatures: []agentConf.AgentFeature{logParsingPipelineController},
+		AgentFeatures: []agentConf.AgentFeature{logParsingPipelineController, signoz.Modules.LLMPricingRule},
 	})
 	if err != nil {
 		return nil, err

pkg/query-service/app/server.go

@@ -128,6 +128,7 @@ func NewServer(config signoz.Config, signoz *signoz.SigNoz) (*Server, error) {
 			Store: signoz.SQLStore,
 			AgentFeatures: []agentConf.AgentFeature{
 				logParsingPipelineController,
+				signoz.Modules.LLMPricingRule,
 			},
 		},
 	)

pkg/types/llmpricingruletypes/pricing.go

@@ -53,7 +53,8 @@ var (
 	LLMPricingRuleCacheModeSubtract = LLMPricingRuleCacheMode{valuer.NewString("subtract")}
 	// LLMPricingRuleCacheModeAdditive: cached tokens are reported separately (Anthropic-style).
 	LLMPricingRuleCacheModeAdditive = LLMPricingRuleCacheMode{valuer.NewString("additive")}
-	// LLMPricingRuleCacheModeUnknown: provider behaviour is unknown; falls back to subtract.
+	// LLMPricingRuleCacheModeUnknown: provider behaviour is unknown. buildProcessorConfig
+	// normalizes this to an empty mode in the collector config.
 	LLMPricingRuleCacheModeUnknown = LLMPricingRuleCacheMode{valuer.NewString("unknown")}
 )
 

pkg/types/llmpricingruletypes/processorconfig.go

@@ -26,24 +26,23 @@ type LLMPricingRuleProcessorAttrs struct {
 	CacheWrite string `yaml:"cache_write" json:"cache_write"`
 }
 
-// LLMPricingRuleProcessorDefaultPricing holds the pricing unit and the list of model-specific rules.
+// LLMPricingRuleProcessorDefaultPricing holds the list of model-specific rules.
 type LLMPricingRuleProcessorDefaultPricing struct {
-	Unit  string                    `yaml:"unit" json:"unit"`
 	Rules []LLMPricingRuleProcessor `yaml:"rules" json:"rules"`
 }
 
 // LLMPricingRuleProcessor is a single pricing rule inside the processor config.
 type LLMPricingRuleProcessor struct {
-	Name    string                       `yaml:"name" json:"name"`
-	Pattern []string                     `yaml:"pattern" json:"pattern"`
-	Cache   LLMPricingRuleProcessorCache `yaml:"cache" json:"cache"`
-	In      float64                      `yaml:"in" json:"in"`
-	Out     float64                      `yaml:"out" json:"out"`
+	Name    string                        `yaml:"name" json:"name"`
+	Pattern []string                      `yaml:"pattern" json:"pattern"`
+	Cache   *LLMPricingRuleProcessorCache `yaml:"cache,omitempty" json:"cache,omitempty"`
+	In      float64                       `yaml:"in" json:"in"`
+	Out     float64                       `yaml:"out" json:"out"`
 }
 
 // LLMPricingRuleProcessorCache describes how cached tokens are accounted for.
 type LLMPricingRuleProcessorCache struct {
-	Mode  string  `yaml:"mode" json:"mode"`
+	Mode  string  `yaml:"mode,omitempty" json:"mode,omitempty"`
 	Read  float64 `yaml:"read" json:"read"`
 	Write float64 `yaml:"write" json:"write"`
 }
@@ -61,10 +60,14 @@ type LLMPricingRuleProcessorOutputAttrs struct {
 func buildProcessorConfig(rules []*LLMPricingRule) *LLMPricingRuleProcessorConfig {
 	pricingRules := make([]LLMPricingRuleProcessor, 0, len(rules))
 	for _, r := range rules {
-		var cache LLMPricingRuleProcessorCache
+		var cache *LLMPricingRuleProcessorCache
 		if r.Pricing.Cache != nil {
-			cache = LLMPricingRuleProcessorCache{
-				Mode:  r.Pricing.Cache.Mode.StringValue(),
+			mode := r.Pricing.Cache.Mode.StringValue()
+			if mode != LLMPricingRuleCacheModeSubtract.StringValue() && mode != LLMPricingRuleCacheModeAdditive.StringValue() {
+				mode = ""
+			}
+			cache = &LLMPricingRuleProcessorCache{
+				Mode:  mode,
 				Read:  r.Pricing.Cache.Read,
 				Write: r.Pricing.Cache.Write,
 			}
@@ -87,7 +90,6 @@ func buildProcessorConfig(rules []*LLMPricingRule) *LLMPricingRuleProcessorConfi
 			CacheWrite: GenAIUsageCacheCreationInputTokens,
 		},
 		DefaultPricing: LLMPricingRuleProcessorDefaultPricing{
-			Unit:  UnitPerMillionTokens.StringValue(),
 			Rules: pricingRules,
 		},
 		OutputAttrs: LLMPricingRuleProcessorOutputAttrs{

pkg/types/llmpricingruletypes/processorconfig_test.go

@@ -41,6 +41,16 @@ func makePricingRule(model string, patterns []string, cacheMode LLMPricingRuleCa
 	}
 }
 
+func makePricingRuleNoCache(model string, patterns []string, costIn, costOut float64) *LLMPricingRule {
+	return &LLMPricingRule{
+		Model:        model,
+		ModelPattern: StringSlice(patterns),
+		Unit:         UnitPerMillionTokens,
+		Pricing:      LLMRulePricing{Input: costIn, Output: costOut},
+		Enabled:      true,
+	}
+}
+
 func TestGenerateCollectorConfigWithLLMPricingProcessor(t *testing.T) {
 	tests := []struct {
 		name         string
@@ -62,6 +72,24 @@ func TestGenerateCollectorConfigWithLLMPricingProcessor(t *testing.T) {
 			rules:        nil,
 			expectedFile: "collector_no_rules.yaml",
 		},
+		// A rule without cache pricing omits the cache block entirely so the
+		// collector does not apply any cache cost.
+		{
+			name: "rule_without_cache",
+			rules: []*LLMPricingRule{
+				makePricingRuleNoCache("gpt-4o", []string{"gpt-4o*"}, 5.0, 15.0),
+			},
+			expectedFile: "collector_rule_without_cache.yaml",
+		},
+		// An unknown cache mode still emits the cache block (with prices) but omits
+		// the mode field; the collector handles a missing mode in its default branch.
+		{
+			name: "rule_cache_mode_unknown",
+			rules: []*LLMPricingRule{
+				makePricingRule("gpt-4o", []string{"gpt-4o*"}, LLMPricingRuleCacheModeUnknown, 5.0, 15.0, 2.5, 0),
+			},
+			expectedFile: "collector_rule_cache_mode_unknown.yaml",
+		},
 	}
 
 	input, err := os.ReadFile(filepath.Join("testdata", "collector_baseline.yaml"))

pkg/types/llmpricingruletypes/testdata/collector_baseline.yaml

@@ -11,7 +11,6 @@ processors:
       cache_read: gen_ai.usage.cache_read.input_tokens
       cache_write: gen_ai.usage.cache_creation.input_tokens
     default_pricing:
-      unit: per_million_tokens
       rules: []
     output_attrs:
       in: _signoz.gen_ai.cost_input

pkg/types/llmpricingruletypes/testdata/collector_no_rules.yaml

@@ -11,7 +11,6 @@ processors:
             cache_read: gen_ai.usage.cache_read.input_tokens
             cache_write: gen_ai.usage.cache_creation.input_tokens
         default_pricing:
-            unit: per_million_tokens
             rules: []
         output_attrs:
             in: _signoz.gen_ai.cost_input

pkg/types/llmpricingruletypes/testdata/collector_rule_cache_mode_unknown.yaml

@@ -0,0 +1,42 @@
+exporters:
+    otlp:
+        endpoint: localhost:4317
+processors:
+    batch: {}
+    signozllmpricing:
+        attrs:
+            model: gen_ai.request.model
+            in: gen_ai.usage.input_tokens
+            out: gen_ai.usage.output_tokens
+            cache_read: gen_ai.usage.cache_read.input_tokens
+            cache_write: gen_ai.usage.cache_creation.input_tokens
+        default_pricing:
+            rules:
+                - name: gpt-4o
+                  pattern:
+                    - gpt-4o*
+                  cache:
+                    read: 2.5
+                    write: 0
+                  in: 5
+                  out: 15
+        output_attrs:
+            in: _signoz.gen_ai.cost_input
+            out: _signoz.gen_ai.cost_output
+            cache_read: _signoz.gen_ai.cost_cache_read
+            cache_write: _signoz.gen_ai.cost_cache_write
+            total: _signoz.gen_ai.total_cost
+receivers:
+    otlp:
+        protocols:
+            grpc: null
+service:
+    pipelines:
+        traces:
+            exporters:
+                - otlp
+            processors:
+                - batch
+                - signozllmpricing
+            receivers:
+                - otlp

pkg/types/llmpricingruletypes/testdata/collector_rule_without_cache.yaml

@@ -0,0 +1,39 @@
+exporters:
+    otlp:
+        endpoint: localhost:4317
+processors:
+    batch: {}
+    signozllmpricing:
+        attrs:
+            model: gen_ai.request.model
+            in: gen_ai.usage.input_tokens
+            out: gen_ai.usage.output_tokens
+            cache_read: gen_ai.usage.cache_read.input_tokens
+            cache_write: gen_ai.usage.cache_creation.input_tokens
+        default_pricing:
+            rules:
+                - name: gpt-4o
+                  pattern:
+                    - gpt-4o*
+                  in: 5
+                  out: 15
+        output_attrs:
+            in: _signoz.gen_ai.cost_input
+            out: _signoz.gen_ai.cost_output
+            cache_read: _signoz.gen_ai.cost_cache_read
+            cache_write: _signoz.gen_ai.cost_cache_write
+            total: _signoz.gen_ai.total_cost
+receivers:
+    otlp:
+        protocols:
+            grpc: null
+service:
+    pipelines:
+        traces:
+            exporters:
+                - otlp
+            processors:
+                - batch
+                - signozllmpricing
+            receivers:
+                - otlp

pkg/types/llmpricingruletypes/testdata/collector_with_rule.yaml

@@ -11,7 +11,6 @@ processors:
             cache_read: gen_ai.usage.cache_read.input_tokens
             cache_write: gen_ai.usage.cache_creation.input_tokens
         default_pricing:
-            unit: per_million_tokens
             rules:
                 - name: gpt-4o
                   pattern: