[router] complete router oai spec (#8828)

sgl-router/benches/request_processing.rs

@@ -8,12 +8,116 @@ use sglang_router_rs::openai_api_types::{
 };
 use sglang_router_rs::routers::request_adapter::{RouteableRequest, ToPdRequest};
 
+/// Create a default GenerateRequest for benchmarks with minimal fields set
+fn default_generate_request() -> GenerateRequest {
+    GenerateRequest {
+        text: None,
+        prompt: None,
+        input_ids: None,
+        stream: false,
+        parameters: None,
+        sampling_params: None,
+        return_logprob: false,
+        // SGLang Extensions
+        lora_path: None,
+        session_params: None,
+        return_hidden_states: false,
+        rid: None,
+    }
+}
+
+/// Create a default ChatCompletionRequest for benchmarks with minimal fields set
+fn default_chat_completion_request() -> ChatCompletionRequest {
+    ChatCompletionRequest {
+        model: String::new(),
+        messages: vec![],
+        max_tokens: None,
+        max_completion_tokens: None,
+        temperature: None,
+        top_p: None,
+        n: None,
+        stream: false,
+        stream_options: None,
+        stop: None,
+        presence_penalty: None,
+        frequency_penalty: None,
+        logit_bias: None,
+        logprobs: false,
+        top_logprobs: None,
+        user: None,
+        response_format: None,
+        seed: None,
+        tools: None,
+        tool_choice: None,
+        parallel_tool_calls: None,
+        function_call: None,
+        functions: None,
+        // SGLang Extensions
+        top_k: None,
+        min_p: None,
+        min_tokens: None,
+        repetition_penalty: None,
+        regex: None,
+        ebnf: None,
+        stop_token_ids: None,
+        no_stop_trim: false,
+        ignore_eos: false,
+        continue_final_message: false,
+        skip_special_tokens: true,
+        // SGLang Extensions
+        lora_path: None,
+        session_params: None,
+        separate_reasoning: true,
+        stream_reasoning: true,
+        return_hidden_states: false,
+    }
+}
+
+/// Create a default CompletionRequest for benchmarks with minimal fields set
+fn default_completion_request() -> CompletionRequest {
+    CompletionRequest {
+        model: String::new(),
+        prompt: StringOrArray::String(String::new()),
+        suffix: None,
+        max_tokens: None,
+        temperature: None,
+        top_p: None,
+        n: None,
+        stream: false,
+        stream_options: None,
+        logprobs: None,
+        echo: false,
+        stop: None,
+        presence_penalty: None,
+        frequency_penalty: None,
+        best_of: None,
+        logit_bias: None,
+        user: None,
+        seed: None,
+        // SGLang Extensions
+        top_k: None,
+        min_p: None,
+        min_tokens: None,
+        repetition_penalty: None,
+        regex: None,
+        ebnf: None,
+        json_schema: None,
+        stop_token_ids: None,
+        no_stop_trim: false,
+        ignore_eos: false,
+        skip_special_tokens: true,
+        // SGLang Extensions
+        lora_path: None,
+        session_params: None,
+        return_hidden_states: false,
+        other: serde_json::Map::new(),
+    }
+}
+
 // Sample request data for benchmarks
 fn create_sample_generate_request() -> GenerateRequest {
     GenerateRequest {
         text: Some("Write a story about artificial intelligence".to_string()),
-        input_ids: None,
-        prompt: None,
         parameters: Some(GenerateParameters {
             max_new_tokens: Some(100),
             temperature: Some(0.8),
@@ -31,8 +135,7 @@ fn create_sample_generate_request() -> GenerateRequest {
             repetition_penalty: Some(1.0),
             ..Default::default()
         }),
-        stream: false,
-        return_logprob: false,
+        ..default_generate_request()
     }
 }
 
@@ -58,22 +161,10 @@ fn create_sample_chat_completion_request() -> ChatCompletionRequest {
         temperature: Some(0.7),
         top_p: Some(1.0),
         n: Some(1),
-        stream: false,
-        stream_options: None,
-        stop: None,
         presence_penalty: Some(0.0),
         frequency_penalty: Some(0.0),
-        logit_bias: None,
-        logprobs: false,
-        top_logprobs: None,
-        user: None,
-        response_format: None,
-        seed: None,
-        tools: None,
-        tool_choice: None,
         parallel_tool_calls: Some(true),
-        function_call: None,
-        functions: None,
+        ..default_chat_completion_request()
     }
 }
 
@@ -81,23 +172,14 @@ fn create_sample_completion_request() -> CompletionRequest {
     CompletionRequest {
         model: "text-davinci-003".to_string(),
         prompt: StringOrArray::String("Complete this sentence: The future of AI is".to_string()),
-        suffix: None,
         max_tokens: Some(50),
         temperature: Some(0.8),
         top_p: Some(1.0),
         n: Some(1),
-        stream: false,
-        stream_options: None,
-        logprobs: None,
-        echo: false,
-        stop: None,
         presence_penalty: Some(0.0),
         frequency_penalty: Some(0.0),
         best_of: Some(1),
-        logit_bias: None,
-        user: None,
-        seed: None,
-        other: serde_json::Map::new(),
+        ..default_completion_request()
     }
 }
 
@@ -121,6 +203,7 @@ fn create_large_chat_completion_request() -> ChatCompletionRequest {
             name: None,
             tool_calls: None,
             function_call: None,
+            reasoning_content: None,
         });
     }
 
@@ -132,22 +215,13 @@ fn create_large_chat_completion_request() -> ChatCompletionRequest {
         temperature: Some(0.7),
         top_p: Some(0.95),
         n: Some(1),
-        stream: false,
-        stream_options: None,
-        stop: None,
         presence_penalty: Some(0.1),
         frequency_penalty: Some(0.1),
-        logit_bias: None,
-        logprobs: false,
         top_logprobs: Some(5),
         user: Some("benchmark_user".to_string()),
-        response_format: None,
         seed: Some(42),
-        tools: None,
-        tool_choice: None,
         parallel_tool_calls: Some(true),
-        function_call: None,
-        functions: None,
+        ..default_chat_completion_request()
     }
 }
 
@@ -331,32 +405,17 @@ fn bench_throughput_by_size(c: &mut Criterion) {
     // Create requests of different sizes
     let small_generate = GenerateRequest {
         text: Some("Hi".to_string()),
-        input_ids: None,
-        prompt: None,
-        parameters: None,
-        sampling_params: None,
-        stream: false,
-        return_logprob: false,
+        ..default_generate_request()
     };
 
     let medium_generate = GenerateRequest {
         text: Some("Write a medium length story about AI".repeat(10)),
-        input_ids: None,
-        prompt: None,
-        parameters: None,
-        sampling_params: None,
-        stream: false,
-        return_logprob: false,
+        ..default_generate_request()
     };
 
     let large_generate = GenerateRequest {
         text: Some("Write a very long and detailed story about artificial intelligence and its impact on society".repeat(100)),
-        input_ids: None,
-        prompt: None,
-        parameters: None,
-        sampling_params: None,
-        stream: false,
-        return_logprob: false,
+        ..default_generate_request()
     };
 
     for (name, req) in [

sgl-router/src/openai_api_types.rs

@@ -6,6 +6,21 @@ use serde::{Deserialize, Serialize};
 use serde_json::Value;
 use std::collections::HashMap;
 
+/// Helper function for serde default value
+fn default_true() -> bool {
+    true
+}
+
+// ============= SGLang-Specific Types =============
+
+/// LoRA adapter path - can be single path or batch of paths
+#[derive(Debug, Clone, Deserialize, Serialize)]
+#[serde(untagged)]
+pub enum LoRAPath {
+    Single(Option<String>),
+    Batch(Vec<Option<String>>),
+}
+
 /// Common trait for all generation requests
 pub trait GenerationRequest: Send + Sync {
     /// Check if the request is for streaming
@@ -92,6 +107,64 @@ pub struct CompletionRequest {
     #[serde(skip_serializing_if = "Option::is_none")]
     pub seed: Option<i64>,
 
+    // ============= SGLang Extensions =============
+    /// Top-k sampling parameter (-1 to disable)
+    #[serde(skip_serializing_if = "Option::is_none")]
+    pub top_k: Option<i32>,
+
+    /// Min-p nucleus sampling parameter
+    #[serde(skip_serializing_if = "Option::is_none")]
+    pub min_p: Option<f32>,
+
+    /// Minimum number of tokens to generate
+    #[serde(skip_serializing_if = "Option::is_none")]
+    pub min_tokens: Option<u32>,
+
+    /// Repetition penalty for reducing repetitive text
+    #[serde(skip_serializing_if = "Option::is_none")]
+    pub repetition_penalty: Option<f32>,
+
+    /// Regex constraint for output generation
+    #[serde(skip_serializing_if = "Option::is_none")]
+    pub regex: Option<String>,
+
+    /// EBNF grammar constraint for structured output
+    #[serde(skip_serializing_if = "Option::is_none")]
+    pub ebnf: Option<String>,
+
+    /// JSON schema constraint for structured output
+    #[serde(skip_serializing_if = "Option::is_none")]
+    pub json_schema: Option<String>,
+
+    /// Specific token IDs to use as stop conditions
+    #[serde(skip_serializing_if = "Option::is_none")]
+    pub stop_token_ids: Option<Vec<i32>>,
+
+    /// Skip trimming stop tokens from output
+    #[serde(default)]
+    pub no_stop_trim: bool,
+
+    /// Ignore end-of-sequence tokens during generation
+    #[serde(default)]
+    pub ignore_eos: bool,
+
+    /// Skip special tokens during detokenization
+    #[serde(default = "default_true")]
+    pub skip_special_tokens: bool,
+
+    // ============= SGLang Extensions =============
+    /// Path to LoRA adapter(s) for model customization
+    #[serde(skip_serializing_if = "Option::is_none")]
+    pub lora_path: Option<LoRAPath>,
+
+    /// Session parameters for continual prompting
+    #[serde(skip_serializing_if = "Option::is_none")]
+    pub session_params: Option<HashMap<String, serde_json::Value>>,
+
+    /// Return model hidden states
+    #[serde(default)]
+    pub return_hidden_states: bool,
+
     /// Additional fields including bootstrap info for PD routing
     #[serde(flatten)]
     pub other: serde_json::Map<String, serde_json::Value>,
@@ -166,7 +239,7 @@ pub struct ChatCompletionRequest {
 
     /// Modify the likelihood of specified tokens appearing in the completion
     #[serde(skip_serializing_if = "Option::is_none")]
-    pub logit_bias: Option<HashMap<String, i32>>,
+    pub logit_bias: Option<HashMap<String, f32>>,
 
     /// A unique identifier representing your end-user
     #[serde(skip_serializing_if = "Option::is_none")]
@@ -207,6 +280,72 @@ pub struct ChatCompletionRequest {
     /// Deprecated: use tool_choice instead
     #[serde(skip_serializing_if = "Option::is_none")]
     pub function_call: Option<FunctionCall>,
+
+    // ============= SGLang Extensions =============
+    /// Top-k sampling parameter (-1 to disable)
+    #[serde(skip_serializing_if = "Option::is_none")]
+    pub top_k: Option<i32>,
+
+    /// Min-p nucleus sampling parameter
+    #[serde(skip_serializing_if = "Option::is_none")]
+    pub min_p: Option<f32>,
+
+    /// Minimum number of tokens to generate
+    #[serde(skip_serializing_if = "Option::is_none")]
+    pub min_tokens: Option<u32>,
+
+    /// Repetition penalty for reducing repetitive text
+    #[serde(skip_serializing_if = "Option::is_none")]
+    pub repetition_penalty: Option<f32>,
+
+    /// Regex constraint for output generation
+    #[serde(skip_serializing_if = "Option::is_none")]
+    pub regex: Option<String>,
+
+    /// EBNF grammar constraint for structured output
+    #[serde(skip_serializing_if = "Option::is_none")]
+    pub ebnf: Option<String>,
+
+    /// Specific token IDs to use as stop conditions
+    #[serde(skip_serializing_if = "Option::is_none")]
+    pub stop_token_ids: Option<Vec<i32>>,
+
+    /// Skip trimming stop tokens from output
+    #[serde(default)]
+    pub no_stop_trim: bool,
+
+    /// Ignore end-of-sequence tokens during generation
+    #[serde(default)]
+    pub ignore_eos: bool,
+
+    /// Continue generating from final assistant message
+    #[serde(default)]
+    pub continue_final_message: bool,
+
+    /// Skip special tokens during detokenization
+    #[serde(default = "default_true")]
+    pub skip_special_tokens: bool,
+
+    // ============= SGLang Extensions =============
+    /// Path to LoRA adapter(s) for model customization
+    #[serde(skip_serializing_if = "Option::is_none")]
+    pub lora_path: Option<LoRAPath>,
+
+    /// Session parameters for continual prompting
+    #[serde(skip_serializing_if = "Option::is_none")]
+    pub session_params: Option<HashMap<String, serde_json::Value>>,
+
+    /// Separate reasoning content from final answer (O1-style models)
+    #[serde(default = "default_true")]
+    pub separate_reasoning: bool,
+
+    /// Stream reasoning tokens during generation
+    #[serde(default = "default_true")]
+    pub stream_reasoning: bool,
+
+    /// Return model hidden states
+    #[serde(default)]
+    pub return_hidden_states: bool,
 }
 
 #[derive(Debug, Clone, Deserialize, Serialize)]
@@ -234,6 +373,9 @@ pub enum ChatMessage {
         tool_calls: Option<Vec<ToolCall>>,
         #[serde(skip_serializing_if = "Option::is_none")]
         function_call: Option<FunctionCallResponse>,
+        /// Reasoning content for O1-style models (SGLang extension)
+        #[serde(skip_serializing_if = "Option::is_none")]
+        reasoning_content: Option<String>,
     },
     Tool {
         role: String, // "tool"
@@ -378,7 +520,20 @@ impl GenerationRequest for ChatCompletionRequest {
                         Some(texts.join(" "))
                     }
                 },
-                ChatMessage::Assistant { content, .. } => content.clone(),
+                ChatMessage::Assistant {
+                    content,
+                    reasoning_content,
+                    ..
+                } => {
+                    // Combine content and reasoning content for routing decisions
+                    let main_content = content.clone().unwrap_or_default();
+                    let reasoning = reasoning_content.clone().unwrap_or_default();
+                    if main_content.is_empty() && reasoning.is_empty() {
+                        None
+                    } else {
+                        Some(format!("{} {}", main_content, reasoning).trim().to_string())
+                    }
+                }
                 ChatMessage::Tool { content, .. } => Some(content.clone()),
                 ChatMessage::Function { content, .. } => Some(content.clone()),
             })
@@ -418,6 +573,23 @@ pub struct GenerateRequest {
     /// Whether to return logprobs
     #[serde(default)]
     pub return_logprob: bool,
+
+    // ============= SGLang Extensions =============
+    /// Path to LoRA adapter(s) for model customization
+    #[serde(skip_serializing_if = "Option::is_none")]
+    pub lora_path: Option<LoRAPath>,
+
+    /// Session parameters for continual prompting
+    #[serde(skip_serializing_if = "Option::is_none")]
+    pub session_params: Option<HashMap<String, serde_json::Value>>,
+
+    /// Return model hidden states
+    #[serde(default)]
+    pub return_hidden_states: bool,
+
+    /// Request ID for tracking
+    #[serde(skip_serializing_if = "Option::is_none")]
+    pub rid: Option<String>,
 }
 
 #[derive(Debug, Clone, Deserialize, Serialize)]
@@ -485,6 +657,18 @@ pub struct SamplingParams {
     pub skip_special_tokens: Option<bool>,
     #[serde(skip_serializing_if = "Option::is_none")]
     pub json_schema: Option<String>,
+    #[serde(skip_serializing_if = "Option::is_none")]
+    pub regex: Option<String>,
+    #[serde(skip_serializing_if = "Option::is_none")]
+    pub ebnf: Option<String>,
+    #[serde(skip_serializing_if = "Option::is_none")]
+    pub min_p: Option<f32>,
+    #[serde(skip_serializing_if = "Option::is_none")]
+    pub min_tokens: Option<u32>,
+    #[serde(skip_serializing_if = "Option::is_none")]
+    pub stop_token_ids: Option<Vec<i32>>,
+    #[serde(skip_serializing_if = "Option::is_none")]
+    pub no_stop_trim: Option<bool>,
 }
 
 impl GenerationRequest for GenerateRequest {
@@ -561,6 +745,12 @@ pub struct CompletionChoice {
     #[serde(skip_serializing_if = "Option::is_none")]
     pub logprobs: Option<LogProbs>,
     pub finish_reason: Option<String>, // "stop", "length", "content_filter", etc.
+    /// Information about which stop condition was matched
+    #[serde(skip_serializing_if = "Option::is_none")]
+    pub matched_stop: Option<serde_json::Value>, // Can be string or integer
+    /// Hidden states from the model (SGLang extension)
+    #[serde(skip_serializing_if = "Option::is_none")]
+    pub hidden_states: Option<Vec<f32>>,
 }
 
 #[derive(Debug, Clone, Deserialize, Serialize)]
@@ -591,6 +781,12 @@ pub struct ChatChoice {
     #[serde(skip_serializing_if = "Option::is_none")]
     pub logprobs: Option<ChatLogProbs>,
     pub finish_reason: Option<String>, // "stop", "length", "tool_calls", "content_filter", "function_call"
+    /// Information about which stop condition was matched
+    #[serde(skip_serializing_if = "Option::is_none")]
+    pub matched_stop: Option<serde_json::Value>, // Can be string or integer
+    /// Hidden states from the model (SGLang extension)
+    #[serde(skip_serializing_if = "Option::is_none")]
+    pub hidden_states: Option<Vec<f32>>,
 }
 
 #[derive(Debug, Clone, Deserialize, Serialize)]
@@ -681,6 +877,9 @@ pub struct ChatMessageDelta {
     pub tool_calls: Option<Vec<ToolCallDelta>>,
     #[serde(skip_serializing_if = "Option::is_none")]
     pub function_call: Option<FunctionCallDelta>,
+    /// Reasoning content delta for O1-style models (SGLang extension)
+    #[serde(skip_serializing_if = "Option::is_none")]
+    pub reasoning_content: Option<String>,
 }
 
 #[derive(Debug, Clone, Deserialize, Serialize)]

sgl-router/src/routers/pd_types.rs

@@ -278,11 +278,11 @@ mod bootstrap_tests {
     use crate::core::BasicWorker;
     use crate::openai_api_types::StringOrArray;
 
-    #[test]
-    fn test_completion_batch_size_with_array_prompt() {
-        let req = CompletionRequest {
-            model: "test".to_string(),
-            prompt: StringOrArray::Array(vec!["prompt1".to_string(), "prompt2".to_string()]),
+    /// Create a default CompletionRequest for testing with minimal fields set
+    fn default_completion_request() -> CompletionRequest {
+        CompletionRequest {
+            model: String::new(),
+            prompt: StringOrArray::String(String::new()),
             n: None,
             other: serde_json::Map::new(),
             suffix: None,
@@ -300,6 +300,31 @@ mod bootstrap_tests {
             logit_bias: None,
             user: None,
             seed: None,
+            // SGLang Extensions
+            top_k: None,
+            min_p: None,
+            min_tokens: None,
+            repetition_penalty: None,
+            regex: None,
+            ebnf: None,
+            json_schema: None,
+            stop_token_ids: None,
+            no_stop_trim: false,
+            ignore_eos: false,
+            skip_special_tokens: true,
+            // SGLang Extensions
+            lora_path: None,
+            session_params: None,
+            return_hidden_states: false,
+        }
+    }
+
+    #[test]
+    fn test_completion_batch_size_with_array_prompt() {
+        let req = CompletionRequest {
+            model: "test".to_string(),
+            prompt: StringOrArray::Array(vec!["prompt1".to_string(), "prompt2".to_string()]),
+            ..default_completion_request()
         };
 
         // Should return batch size for array prompt
@@ -311,23 +336,7 @@ mod bootstrap_tests {
         let req = CompletionRequest {
             model: "test".to_string(),
             prompt: StringOrArray::String("single prompt".to_string()),
-            n: None,
-            other: serde_json::Map::new(),
-            suffix: None,
-            max_tokens: None,
-            temperature: None,
-            top_p: None,
-            stream: false,
-            stream_options: None,
-            logprobs: None,
-            echo: false,
-            stop: None,
-            presence_penalty: None,
-            frequency_penalty: None,
-            best_of: None,
-            logit_bias: None,
-            user: None,
-            seed: None,
+            ..default_completion_request()
         };
 
         // Should return None for single prompt
@@ -340,22 +349,7 @@ mod bootstrap_tests {
             model: "test".to_string(),
             prompt: StringOrArray::String("single prompt".to_string()),
             n: Some(3),
-            other: serde_json::Map::new(),
-            suffix: None,
-            max_tokens: None,
-            temperature: None,
-            top_p: None,
-            stream: false,
-            stream_options: None,
-            logprobs: None,
-            echo: false,
-            stop: None,
-            presence_penalty: None,
-            frequency_penalty: None,
-            best_of: None,
-            logit_bias: None,
-            user: None,
-            seed: None,
+            ..default_completion_request()
         };
 
         // Should return None for single string prompt, even with n > 1
@@ -368,23 +362,7 @@ mod bootstrap_tests {
         let mut req = CompletionRequest {
             model: "test".to_string(),
             prompt: StringOrArray::Array(vec!["prompt1".to_string(), "prompt2".to_string()]),
-            n: None,
-            other: serde_json::Map::new(),
-            suffix: None,
-            max_tokens: None,
-            temperature: None,
-            top_p: None,
-            stream: false,
-            stream_options: None,
-            logprobs: None,
-            echo: false,
-            stop: None,
-            presence_penalty: None,
-            frequency_penalty: None,
-            best_of: None,
-            logit_bias: None,
-            user: None,
-            seed: None,
+            ..default_completion_request()
         };
 
         // Set bootstrap info - should always use single values
@@ -418,23 +396,7 @@ mod bootstrap_tests {
         let mut req = CompletionRequest {
             model: "test".to_string(),
             prompt: StringOrArray::Array(vec!["prompt1".to_string(), "prompt2".to_string()]),
-            n: None,
-            other: serde_json::Map::new(),
-            suffix: None,
-            max_tokens: None,
-            temperature: None,
-            top_p: None,
-            stream: false,
-            stream_options: None,
-            logprobs: None,
-            echo: false,
-            stop: None,
-            presence_penalty: None,
-            frequency_penalty: None,
-            best_of: None,
-            logit_bias: None,
-            user: None,
-            seed: None,
+            ..default_completion_request()
         };
 
         // Set bootstrap info with arrays

sgl-router/tests/benchmark_integration.rs

@@ -8,14 +8,118 @@ use sglang_router_rs::openai_api_types::{
 };
 use sglang_router_rs::routers::request_adapter::{RouteableRequest, ToPdRequest};
 
+/// Create a default GenerateRequest for benchmarks with minimal fields set
+fn default_generate_request() -> GenerateRequest {
+    GenerateRequest {
+        text: None,
+        prompt: None,
+        input_ids: None,
+        stream: false,
+        parameters: None,
+        sampling_params: None,
+        return_logprob: false,
+        // SGLang Extensions
+        lora_path: None,
+        session_params: None,
+        return_hidden_states: false,
+        rid: None,
+    }
+}
+
+/// Create a default ChatCompletionRequest for benchmarks with minimal fields set
+fn default_chat_completion_request() -> ChatCompletionRequest {
+    ChatCompletionRequest {
+        model: String::new(),
+        messages: vec![],
+        max_tokens: None,
+        max_completion_tokens: None,
+        temperature: None,
+        top_p: None,
+        n: None,
+        stream: false,
+        stream_options: None,
+        stop: None,
+        presence_penalty: None,
+        frequency_penalty: None,
+        logit_bias: None,
+        logprobs: false,
+        top_logprobs: None,
+        user: None,
+        response_format: None,
+        seed: None,
+        tools: None,
+        tool_choice: None,
+        parallel_tool_calls: None,
+        function_call: None,
+        functions: None,
+        // SGLang Extensions
+        top_k: None,
+        min_p: None,
+        min_tokens: None,
+        repetition_penalty: None,
+        regex: None,
+        ebnf: None,
+        stop_token_ids: None,
+        no_stop_trim: false,
+        ignore_eos: false,
+        continue_final_message: false,
+        skip_special_tokens: true,
+        // SGLang Extensions
+        lora_path: None,
+        session_params: None,
+        separate_reasoning: true,
+        stream_reasoning: true,
+        return_hidden_states: false,
+    }
+}
+
+/// Create a default CompletionRequest for benchmarks with minimal fields set
+fn default_completion_request() -> CompletionRequest {
+    CompletionRequest {
+        model: String::new(),
+        prompt: StringOrArray::String(String::new()),
+        suffix: None,
+        max_tokens: None,
+        temperature: None,
+        top_p: None,
+        n: None,
+        stream: false,
+        stream_options: None,
+        logprobs: None,
+        echo: false,
+        stop: None,
+        presence_penalty: None,
+        frequency_penalty: None,
+        best_of: None,
+        logit_bias: None,
+        user: None,
+        seed: None,
+        // SGLang Extensions
+        top_k: None,
+        min_p: None,
+        min_tokens: None,
+        repetition_penalty: None,
+        regex: None,
+        ebnf: None,
+        json_schema: None,
+        stop_token_ids: None,
+        no_stop_trim: false,
+        ignore_eos: false,
+        skip_special_tokens: true,
+        // SGLang Extensions
+        lora_path: None,
+        session_params: None,
+        return_hidden_states: false,
+        other: serde_json::Map::new(),
+    }
+}
+
 #[test]
 fn test_benchmark_request_creation() {
     // Ensure all benchmark request types can be created without panicking
 
     let generate_req = GenerateRequest {
         text: Some("Test prompt".to_string()),
-        input_ids: None,
-        prompt: None,
         parameters: Some(GenerateParameters {
             max_new_tokens: Some(100),
             temperature: Some(0.8),
@@ -33,8 +137,7 @@ fn test_benchmark_request_creation() {
             repetition_penalty: Some(1.0),
             ..Default::default()
         }),
-        stream: false,
-        return_logprob: false,
+        ..default_generate_request()
     };
 
     let chat_req = ChatCompletionRequest {
@@ -49,44 +152,23 @@ fn test_benchmark_request_creation() {
         temperature: Some(0.7),
         top_p: Some(1.0),
         n: Some(1),
-        stream: false,
-        stream_options: None,
-        stop: None,
         presence_penalty: Some(0.0),
         frequency_penalty: Some(0.0),
-        logit_bias: None,
-        logprobs: false,
-        top_logprobs: None,
-        user: None,
-        response_format: None,
-        seed: None,
-        tools: None,
-        tool_choice: None,
         parallel_tool_calls: Some(true),
-        function_call: None,
-        functions: None,
+        ..default_chat_completion_request()
     };
 
     let completion_req = CompletionRequest {
         model: "test-model".to_string(),
         prompt: StringOrArray::String("Test prompt".to_string()),
-        suffix: None,
         max_tokens: Some(50),
         temperature: Some(0.8),
         top_p: Some(1.0),
         n: Some(1),
-        stream: false,
-        stream_options: None,
-        logprobs: None,
-        echo: false,
-        stop: None,
         presence_penalty: Some(0.0),
         frequency_penalty: Some(0.0),
         best_of: Some(1),
-        logit_bias: None,
-        user: None,
-        seed: None,
-        other: serde_json::Map::new(),
+        ..default_completion_request()
     };
 
     // Test serialization works
@@ -101,12 +183,7 @@ fn test_benchmark_serialization_roundtrip() {
 
     let generate_req = GenerateRequest {
         text: Some("Test prompt".to_string()),
-        input_ids: None,
-        prompt: None,
-        parameters: None,
-        sampling_params: None,
-        stream: false,
-        return_logprob: false,
+        ..default_generate_request()
     };
 
     // Serialize and deserialize
@@ -125,12 +202,7 @@ fn test_benchmark_request_adaptation() {
 
     let generate_req = GenerateRequest {
         text: Some("Test prompt".to_string()),
-        input_ids: None,
-        prompt: None,
-        parameters: None,
-        sampling_params: None,
-        stream: false,
-        return_logprob: false,
+        ..default_generate_request()
     };
 
     let chat_req = ChatCompletionRequest {
@@ -145,44 +217,23 @@ fn test_benchmark_request_adaptation() {
         temperature: Some(0.7),
         top_p: Some(1.0),
         n: Some(1),
-        stream: false,
-        stream_options: None,
-        stop: None,
         presence_penalty: Some(0.0),
         frequency_penalty: Some(0.0),
-        logit_bias: None,
-        logprobs: false,
-        top_logprobs: None,
-        user: None,
-        response_format: None,
-        seed: None,
-        tools: None,
-        tool_choice: None,
         parallel_tool_calls: Some(true),
-        function_call: None,
-        functions: None,
+        ..default_chat_completion_request()
     };
 
     let completion_req = CompletionRequest {
         model: "test-model".to_string(),
         prompt: StringOrArray::String("Test prompt".to_string()),
-        suffix: None,
         max_tokens: Some(50),
         temperature: Some(0.8),
         top_p: Some(1.0),
         n: Some(1),
-        stream: false,
-        stream_options: None,
-        logprobs: None,
-        echo: false,
-        stop: None,
         presence_penalty: Some(0.0),
         frequency_penalty: Some(0.0),
         best_of: Some(1),
-        logit_bias: None,
-        user: None,
-        seed: None,
-        other: serde_json::Map::new(),
+        ..default_completion_request()
     };
 
     // Test PD adaptation (should not panic)
@@ -197,12 +248,7 @@ fn test_benchmark_regular_routing() {
 
     let generate_req = GenerateRequest {
         text: Some("Test prompt".to_string()),
-        input_ids: None,
-        prompt: None,
-        parameters: None,
-        sampling_params: None,
-        stream: false,
-        return_logprob: false,
+        ..default_generate_request()
     };
 
     // Test regular routing methods (should not panic)
@@ -217,12 +263,7 @@ fn test_benchmark_performance_baseline() {
 
     let generate_req = GenerateRequest {
         text: Some("Short test prompt".to_string()),
-        input_ids: None,
-        prompt: None,
-        parameters: None,
-        sampling_params: None,
-        stream: false,
-        return_logprob: false,
+        ..default_generate_request()
     };
 
     // Serialization should be fast (< 1ms for simple requests)