feat: implement decision-based routing with plugin architecture

This commit introduces a comprehensive decision-based routing system with a flexible plugin architecture, replacing the previous category-based approach.

## Core Changes

### 1. Decision-Based Architecture
- Replaced Category-based routing with Decision-based routing
- Decisions combine multiple rules (keyword, embedding, domain) using AND/OR operators
- Added DecisionEngine for evaluating rule combinations and selecting optimal decisions
- Support for priority and confidence-based decision selection strategies

### 2. Plugin System
- Introduced flexible plugin architecture for Decision-level configurations
- Supported plugin types: semantic-cache, jailbreak, pii, system_prompt
- Each plugin has type-specific configuration stored as raw JSON
- Helper methods for type-safe plugin configuration access

### 3. Model References
- Renamed ModelScores to ModelRefs, removed score field
- Currently supports single model per decision (maxItems: 1)
- Simplified model selection logic based on decision priority

### 4.
## Core Changes

### 1. Decision-Based Architecture
- Replaced Category-based routing with Decision-based routing
- Decisions combine multiple rules (kete
### 1. Decisiatc- Replaced Category-based routinged- Decisions combine multiple rules (keyword, embedding, domapo- Added DecisionEngine for evaluating rule combinations and selecting optimal decisioh - Support for priority and confidence-based decision selection strategies

### 2. Plug a
### 2. Plugin System
- Introduced flexible plugin architecture for Deci.go- Introduced flexibur- Supported plugin types: semantic-cache, jailbreak, pii, system_prompt
- ch- Each plugin has type-specific configuration stored as raw JSON
- Hel C- Helper methods for type-safe plugin configuration access

###s
### 3. Model References
- Renamed ModelScores to ModelReute- Renamed ModelScores si- Currently supports single model per decision (maxItee - Simplified model selection logic based on decision priorira
### 4.
## Core Changes

### 1. Decision-Based Architecture: w## Cor C
### 1. Decisis i- Replaced Category-based routingrt- Decisions combine multiple rules (kete
### 1. Decisiatc- Rge### 1. Decisiatc- Replaced Category-bass

### 2. Plug a
### 2. Plugin System
- Introduced flexible plugin architecture for Deci.go- Introduced flexibur- Supported plugin types: semantic-cache, jailbreak, pii, system_prompt
- ch- Each plugin has type-specific configuration stored as raw JSON
- Hel C- Helper methorra### 2. Plugiio- Introduced flexib:
- ch- Each plugin has type-specific configuration stored as raw JSON
- Hel C- Helper methods for type-safe plugin configuration access

###s
##

## Corst
### 1. Decisie p### 1. Decisis i- Replaced Category-based ross### 1. Decisiatc- Rge### 1. Decisiatc- Replaced Category-bass

### 2. Plug a
### 2. PluginEx
### 2. Plug a
### 2. Plugin System
- Introduced flexible plXun### 2. Plugirs- Introduced
git reset HEAD~1

git status --short

cd /Users/bitliu/vs
cd /Users/bitliu/vs/src/semantic-router && git status

git checkout -b feat/decision-based-routing-with-plugins 2>&1

cd /Users/bitliu/vs
bash -c cd

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

Author: Xunzhuo

src/semantic-router/pkg/apis/vllm.ai/v1alpha1/types_route.go

@@ -22,11 +22,11 @@ import (
 
 // IntelligentRouteSpec defines the desired state of IntelligentRoute
 type IntelligentRouteSpec struct {
-	// Strategy defines how to select among multiple matching decisions
+	// DecisionStrategy defines how to select among multiple matching decisions
 	// +kubebuilder:validation:Required
 	// +kubebuilder:validation:Enum=priority;confidence
 	// +kubebuilder:default=priority
-	Strategy DecisionStrategy `json:"strategy" yaml:"strategy"`
+	DecisionStrategy DecisionStrategy `json:"decisionStrategy" yaml:"decisionStrategy"`
 
 	// KeywordRules defines keyword-based classification rules
 	// +optional
@@ -38,10 +38,10 @@ type IntelligentRouteSpec struct {
 	// +kubebuilder:validation:MaxItems=100
 	EmbeddingRules []EmbeddingRule `json:"embeddingRules,omitempty" yaml:"embeddingRules,omitempty"`
 
-	// DomainRules defines domain-based classification rules
+	// DomainRules defines MMLU domain categories for classification
 	// +optional
-	// +kubebuilder:validation:MaxItems=100
-	DomainRules []DomainRule `json:"domainRules,omitempty" yaml:"domainRules,omitempty"`
+	// +kubebuilder:validation:MaxItems=14
+	DomainRules []string `json:"domainRules,omitempty" yaml:"domainRules,omitempty"`
 
 	// Decisions defines the routing decisions based on rule combinations
 	// +kubebuilder:validation:Required
@@ -62,17 +62,12 @@ const (
 
 // KeywordRule defines a keyword-based classification rule
 type KeywordRule struct {
-	// Name is the unique identifier for this rule
+	// Name is the unique identifier for this rule (also used as category name)
 	// +kubebuilder:validation:Required
 	// +kubebuilder:validation:MinLength=1
 	// +kubebuilder:validation:MaxLength=100
 	Name string `json:"name" yaml:"name"`
 
-	// Category is the category name generated by this rule
-	// +optional
-	// +kubebuilder:validation:MaxLength=100
-	Category string `json:"category,omitempty" yaml:"category,omitempty"`
-
 	// Operator defines the logical operator for keywords (AND/OR)
 	// +kubebuilder:validation:Required
 	// +kubebuilder:validation:Enum=AND;OR
@@ -92,60 +87,29 @@ type KeywordRule struct {
 
 // EmbeddingRule defines an embedding-based classification rule
 type EmbeddingRule struct {
-	// Name is the unique identifier for this rule
+	// Name is the unique identifier for this rule (also used as category name)
 	// +kubebuilder:validation:Required
 	// +kubebuilder:validation:MinLength=1
 	// +kubebuilder:validation:MaxLength=100
 	Name string `json:"name" yaml:"name"`
 
-	// Category is the category name generated by this rule
-	// +optional
-	// +kubebuilder:validation:MaxLength=100
-	Category string `json:"category,omitempty" yaml:"category,omitempty"`
-
 	// Threshold is the similarity threshold for matching (0.0-1.0)
 	// +kubebuilder:validation:Required
 	// +kubebuilder:validation:Minimum=0
 	// +kubebuilder:validation:Maximum=1
 	Threshold float32 `json:"threshold" yaml:"threshold"`
 
-	// Keywords is the list of keywords to compute embeddings for
+	// Candidates is the list of candidate phrases for semantic matching
 	// +kubebuilder:validation:Required
 	// +kubebuilder:validation:MinItems=1
 	// +kubebuilder:validation:MaxItems=100
-	Keywords []string `json:"keywords" yaml:"keywords"`
+	Candidates []string `json:"candidates" yaml:"candidates"`
 
-	// AggregationMethod defines how to aggregate multiple keyword similarities
+	// AggregationMethod defines how to aggregate multiple candidate similarities
 	// +optional
 	// +kubebuilder:validation:Enum=mean;max;any
 	// +kubebuilder:default=max
 	AggregationMethod string `json:"aggregationMethod,omitempty" yaml:"aggregationMethod,omitempty"`
-
-	// Model specifies the embedding model to use
-	// +optional
-	// +kubebuilder:validation:Enum=auto;qwen3;gemma
-	// +kubebuilder:default=auto
-	Model string `json:"model,omitempty" yaml:"model,omitempty"`
-
-	// Dimension specifies the target embedding dimension
-	// +optional
-	// +kubebuilder:validation:Enum=128;256;512;768;1024
-	Dimension int `json:"dimension,omitempty" yaml:"dimension,omitempty"`
-}
-
-// DomainRule defines a domain-based classification rule
-type DomainRule struct {
-	// Name is the unique identifier for this rule
-	// +kubebuilder:validation:Required
-	// +kubebuilder:validation:MinLength=1
-	// +kubebuilder:validation:MaxLength=100
-	Name string `json:"name" yaml:"name"`
-
-	// Domains is the list of MMLU domain categories
-	// +kubebuilder:validation:Required
-	// +kubebuilder:validation:MinItems=1
-	// +kubebuilder:validation:MaxItems=14
-	Domains []string `json:"domains" yaml:"domains"`
 }
 
 // Decision defines a routing decision based on rule combinations