[model-gateway] Tighten visibility across data_connector and grpc module (#16516)

sgl-model-gateway/src/data_connector/common.rs

@@ -2,28 +2,28 @@ use std::collections::HashMap;
 
 use serde_json::Value;
 
-pub fn parse_tool_calls(raw: Option<String>) -> Result<Vec<Value>, String> {
+pub(super) fn parse_tool_calls(raw: Option<String>) -> Result<Vec<Value>, String> {
     match raw {
         Some(s) if !s.is_empty() => serde_json::from_str(&s).map_err(|e| e.to_string()),
         _ => Ok(Vec::new()),
     }
 }
 
-pub fn parse_metadata(raw: Option<String>) -> Result<HashMap<String, Value>, String> {
+pub(super) fn parse_metadata(raw: Option<String>) -> Result<HashMap<String, Value>, String> {
     match raw {
         Some(s) if !s.is_empty() => serde_json::from_str(&s).map_err(|e| e.to_string()),
         _ => Ok(HashMap::new()),
     }
 }
 
-pub fn parse_raw_response(raw: Option<String>) -> Result<Value, String> {
+pub(super) fn parse_raw_response(raw: Option<String>) -> Result<Value, String> {
     match raw {
         Some(s) if !s.is_empty() => serde_json::from_str(&s).map_err(|e| e.to_string()),
         _ => Ok(Value::Null),
     }
 }
 
-pub fn parse_json_value(raw: Option<String>) -> Result<Value, String> {
+pub(super) fn parse_json_value(raw: Option<String>) -> Result<Value, String> {
     match raw {
         Some(s) if !s.is_empty() => serde_json::from_str(&s).map_err(|e| e.to_string()),
         _ => Ok(Value::Array(vec![])),

sgl-model-gateway/src/data_connector/memory.rs

@@ -288,7 +288,8 @@ impl MemoryResponseStorage {
     }
 
     /// Get statistics about the store
-    pub fn stats(&self) -> MemoryStoreStats {
+    #[allow(dead_code)]
+    pub(super) fn stats(&self) -> MemoryStoreStats {
         let store = self.store.read();
         MemoryStoreStats {
             response_count: store.responses.len(),
@@ -459,7 +460,8 @@ impl ResponseStorage for MemoryResponseStorage {
 
 /// Statistics for the memory store
 #[derive(Debug, Clone)]
-pub struct MemoryStoreStats {
+#[allow(dead_code)]
+pub(super) struct MemoryStoreStats {
     pub response_count: usize,
     pub identifier_count: usize,
 }

sgl-model-gateway/src/data_connector/noop.rs

@@ -18,7 +18,7 @@ use super::core::*;
 
 /// No-op implementation that synthesizes conversation responses without persistence
 #[derive(Default, Debug, Clone)]
-pub struct NoOpConversationStorage;
+pub(super) struct NoOpConversationStorage;
 
 impl NoOpConversationStorage {
     pub fn new() -> Self {
@@ -61,7 +61,7 @@ impl ConversationStorage for NoOpConversationStorage {
 
 /// No-op conversation item storage (does nothing)
 #[derive(Clone, Copy, Default)]
-pub struct NoOpConversationItemStorage;
+pub(super) struct NoOpConversationItemStorage;
 
 impl NoOpConversationItemStorage {
     pub fn new() -> Self {
@@ -136,7 +136,7 @@ impl ConversationItemStorage for NoOpConversationItemStorage {
 // ============================================================================
 
 /// No-op implementation of response storage (does nothing)
-pub struct NoOpResponseStorage;
+pub(super) struct NoOpResponseStorage;
 
 impl NoOpResponseStorage {
     pub fn new() -> Self {

sgl-model-gateway/src/data_connector/oracle.rs

@@ -232,7 +232,7 @@ impl Manager for OracleConnectionManager {
 // ============================================================================
 
 #[derive(Clone)]
-pub struct OracleConversationStorage {
+pub(super) struct OracleConversationStorage {
     store: OracleStore,
 }
 
@@ -420,7 +420,7 @@ impl ConversationStorage for OracleConversationStorage {
 // ============================================================================
 
 #[derive(Clone)]
-pub struct OracleConversationItemStorage {
+pub(super) struct OracleConversationItemStorage {
     store: OracleStore,
 }
 
@@ -775,7 +775,7 @@ const SELECT_BASE: &str = "SELECT id, previous_response_id, input, instructions,
     tool_calls, metadata, created_at, safety_identifier, model, conversation_id, raw_response FROM responses";
 
 #[derive(Clone)]
-pub struct OracleResponseStorage {
+pub(super) struct OracleResponseStorage {
     store: OracleStore,
 }
 

sgl-model-gateway/src/data_connector/postgres.rs

@@ -58,7 +58,7 @@ impl Clone for PostgresStore {
     }
 }
 
-pub struct PostgresConversationStorage {
+pub(super) struct PostgresConversationStorage {
     store: PostgresStore,
 }
 
@@ -198,7 +198,7 @@ impl ConversationStorage for PostgresConversationStorage {
     }
 }
 
-pub struct PostgresConversationItemStorage {
+pub(super) struct PostgresConversationItemStorage {
     store: PostgresStore,
 }
 
@@ -477,7 +477,7 @@ impl ConversationItemStorage for PostgresConversationItemStorage {
     }
 }
 
-pub struct PostgresResponseStorage {
+pub(super) struct PostgresResponseStorage {
     store: PostgresStore,
 }
 

sgl-model-gateway/src/routers/grpc/common/mod.rs

@@ -1,6 +1,6 @@
 //! Shared code for both regular and harmony routers
 
-pub mod response_collection;
-pub mod response_formatting;
-pub mod responses;
-pub mod stages;
+pub(crate) mod response_collection;
+pub(crate) mod response_formatting;
+pub(crate) mod responses;
+pub(crate) mod stages;

sgl-model-gateway/src/routers/grpc/common/response_collection.rs

@@ -21,7 +21,7 @@ use crate::routers::{
 ///
 /// # Returns
 /// Vector of GenerateComplete responses, one per index (n parameter)
-pub async fn collect_responses(
+pub(crate) async fn collect_responses(
     execution_result: ExecutionResult,
     merge_logprobs: bool,
 ) -> Result<Vec<ProtoGenerateComplete>, Response> {

sgl-model-gateway/src/routers/grpc/common/response_formatting.rs

@@ -16,7 +16,7 @@ use crate::{protocols::common::Usage, routers::grpc::proto_wrapper::ProtoGenerat
 ///
 /// # Returns
 /// Usage object with aggregated token counts
-pub fn build_usage(responses: &[ProtoGenerateComplete]) -> Usage {
+pub(crate) fn build_usage(responses: &[ProtoGenerateComplete]) -> Usage {
     let total_prompt_tokens: u32 = responses.iter().map(|r| r.prompt_tokens() as u32).sum();
     let total_completion_tokens: u32 = responses.iter().map(|r| r.completion_tokens() as u32).sum();
 

sgl-model-gateway/src/routers/grpc/common/responses/handlers.rs

@@ -18,7 +18,7 @@ use crate::{
 ///
 /// Retrieves a stored response from the database.
 /// Used by both regular and harmony implementations.
-pub async fn get_response_impl(ctx: &ResponsesContext, response_id: &str) -> Response {
+pub(crate) async fn get_response_impl(ctx: &ResponsesContext, response_id: &str) -> Response {
     let resp_id = ResponseId::from(response_id);
 
     // Retrieve response from storage
@@ -38,7 +38,7 @@ pub async fn get_response_impl(ctx: &ResponsesContext, response_id: &str) -> Res
 /// Implementation for POST /v1/responses/{response_id}/cancel
 ///
 /// Cancels a background response if it's still in progress.
-pub async fn cancel_response_impl(ctx: &ResponsesContext, response_id: &str) -> Response {
+pub(crate) async fn cancel_response_impl(ctx: &ResponsesContext, response_id: &str) -> Response {
     let resp_id = ResponseId::from(response_id);
 
     // Retrieve response from storage to check if it exists and get current status

sgl-model-gateway/src/routers/grpc/common/responses/mod.rs

@@ -1,9 +1,9 @@
 //! Shared response functionality used by both regular and harmony implementations
 
-pub mod handlers;
-pub mod streaming;
-pub mod utils;
+pub(crate) mod handlers;
+pub(crate) mod streaming;
+pub(crate) mod utils;
 
-pub use handlers::{cancel_response_impl, get_response_impl};
-pub use streaming::{build_sse_response, OutputItemType, ResponseStreamEventEmitter};
-pub use utils::{ensure_mcp_connection, persist_response_if_needed};
+// Re-export commonly used items
+pub(crate) use streaming::build_sse_response;
+pub(crate) use utils::{ensure_mcp_connection, persist_response_if_needed};

sgl-model-gateway/src/routers/grpc/common/responses/streaming.rs

@@ -25,7 +25,7 @@ use crate::{
     routers::grpc::harmony::responses::ToolResult,
 };
 
-pub enum OutputItemType {
+pub(crate) enum OutputItemType {
     Message,
     McpListTools,
     McpCall,
@@ -67,7 +67,7 @@ struct OutputItemState {
 /// - response.mcp_call_arguments.done
 /// - response.mcp_call.completed
 /// - response.mcp_call.failed
-pub struct ResponseStreamEventEmitter {
+pub(crate) struct ResponseStreamEventEmitter {
     sequence_number: u64,
     pub response_id: String,
     model: String,
@@ -828,7 +828,9 @@ impl ResponseStreamEventEmitter {
 /// Build a Server-Sent Events (SSE) response
 ///
 /// Creates a Response with proper SSE headers and streaming body.
-pub fn build_sse_response(rx: mpsc::UnboundedReceiver<Result<Bytes, std::io::Error>>) -> Response {
+pub(crate) fn build_sse_response(
+    rx: mpsc::UnboundedReceiver<Result<Bytes, std::io::Error>>,
+) -> Response {
     let stream = UnboundedReceiverStream::new(rx);
     Response::builder()
         .status(StatusCode::OK)

sgl-model-gateway/src/routers/grpc/common/responses/utils.rs

@@ -23,7 +23,7 @@ use crate::{
 ///
 /// Checks if request declares MCP tools, and if so, validates that
 /// the MCP client can be created and connected.
-pub async fn ensure_mcp_connection(
+pub(crate) async fn ensure_mcp_connection(
     mcp_manager: &Arc<McpManager>,
     tools: Option<&[ResponseTool]>,
 ) -> Result<bool, Response> {
@@ -56,7 +56,7 @@ pub async fn ensure_mcp_connection(
 }
 
 /// Validate that workers are available for the requested model
-pub fn validate_worker_availability(
+pub(crate) fn validate_worker_availability(
     worker_registry: &Arc<WorkerRegistry>,
     model: &str,
 ) -> Option<Response> {
@@ -90,7 +90,7 @@ pub fn validate_worker_availability(
 ///   the initial conversion from ResponsesRequest to ChatCompletionRequest. MCP tools
 ///   are merged later by the tool loop before being sent to the chat pipeline, where
 ///   tool_choice constraints are generated for ALL tools (function + MCP combined).
-pub fn extract_tools_from_response_tools(
+pub(crate) fn extract_tools_from_response_tools(
     response_tools: Option<&[ResponseTool]>,
     include_mcp: bool,
 ) -> Vec<Tool> {
@@ -124,7 +124,7 @@ pub fn extract_tools_from_response_tools(
 ///
 /// Common helper function to avoid duplication across sync and streaming paths
 /// in both harmony and regular responses implementations.
-pub async fn persist_response_if_needed(
+pub(crate) async fn persist_response_if_needed(
     conversation_storage: Arc<dyn ConversationStorage>,
     conversation_item_storage: Arc<dyn ConversationItemStorage>,
     response_storage: Arc<dyn ResponseStorage>,

sgl-model-gateway/src/routers/grpc/common/stages/client_acquisition.rs

@@ -14,7 +14,7 @@ use crate::routers::{
 };
 
 /// Client acquisition stage: Get gRPC clients from selected workers
-pub struct ClientAcquisitionStage;
+pub(crate) struct ClientAcquisitionStage;
 
 #[async_trait]
 impl PipelineStage for ClientAcquisitionStage {

sgl-model-gateway/src/routers/grpc/common/stages/dispatch_metadata.rs

@@ -16,7 +16,7 @@ use crate::{
 };
 
 /// Dispatch metadata stage: Prepare metadata for dispatch
-pub struct DispatchMetadataStage;
+pub(crate) struct DispatchMetadataStage;
 
 #[async_trait]
 impl PipelineStage for DispatchMetadataStage {

sgl-model-gateway/src/routers/grpc/common/stages/helpers.rs

@@ -14,7 +14,7 @@ use crate::{
 ///
 /// Used by both chat and generate request building stages when in PD mode.
 /// Only SGLang supports PD (prefill/decode) disaggregated mode.
-pub fn inject_bootstrap_metadata(
+pub(crate) fn inject_bootstrap_metadata(
     request: &mut ProtoGenerateRequest,
     prefill_worker: &Arc<dyn Worker>,
 ) {

sgl-model-gateway/src/routers/grpc/common/stages/mod.rs

@@ -28,12 +28,12 @@ pub trait PipelineStage: Send + Sync {
 
 mod client_acquisition;
 mod dispatch_metadata;
-pub mod helpers;
+pub(crate) mod helpers;
 mod request_execution;
 mod worker_selection;
 
 // Export stage implementations
-pub use client_acquisition::ClientAcquisitionStage;
-pub use dispatch_metadata::DispatchMetadataStage;
-pub use request_execution::{ExecutionMode, RequestExecutionStage};
-pub use worker_selection::{WorkerSelectionMode, WorkerSelectionStage};
+pub(crate) use client_acquisition::ClientAcquisitionStage;
+pub(crate) use dispatch_metadata::DispatchMetadataStage;
+pub(crate) use request_execution::{ExecutionMode, RequestExecutionStage};
+pub(crate) use worker_selection::{WorkerSelectionMode, WorkerSelectionStage};

sgl-model-gateway/src/routers/grpc/common/stages/request_execution.rs

@@ -19,12 +19,12 @@ use crate::routers::{
 type StreamResult = Result<ProtoStream, Box<dyn std::error::Error + Send + Sync>>;
 
 /// Request execution stage: Execute gRPC requests (single or dual dispatch)
-pub struct RequestExecutionStage {
+pub(crate) struct RequestExecutionStage {
     mode: ExecutionMode,
 }
 
 #[derive(Debug, Clone, Copy)]
-pub enum ExecutionMode {
+pub(crate) enum ExecutionMode {
     /// Regular mode: single worker execution
     Single,
     /// PD mode: dual dispatch to prefill + decode workers

sgl-model-gateway/src/routers/grpc/common/stages/worker_selection.rs

@@ -18,13 +18,13 @@ use crate::{
 };
 
 /// Worker selection stage: Select appropriate worker(s) based on routing mode
-pub struct WorkerSelectionStage {
+pub(crate) struct WorkerSelectionStage {
     worker_registry: Arc<WorkerRegistry>,
     policy_registry: Arc<PolicyRegistry>,
     mode: WorkerSelectionMode,
 }
 
-pub enum WorkerSelectionMode {
+pub(crate) enum WorkerSelectionMode {
     /// Regular mode: select single worker
     Regular,
     /// PD mode: select prefill + decode workers

sgl-model-gateway/src/routers/grpc/context.rs

@@ -4,14 +4,13 @@
 //! eliminating deep parameter passing chains and providing a single source of truth
 //! for request state.
 
-use std::{collections::HashMap, sync::Arc};
+use std::sync::Arc;
 
 use axum::http::HeaderMap;
-use serde_json::Value;
 
 use super::{
     client::GrpcClient,
-    proto_wrapper::{ProtoEmbedComplete, ProtoGenerateComplete, ProtoRequest, ProtoStream},
+    proto_wrapper::{ProtoEmbedComplete, ProtoRequest, ProtoStream},
 };
 use crate::{
     core::{attach_guards_to_response, Worker, WorkerLoadGuard},
@@ -32,14 +31,14 @@ use crate::{
 /// This is the single source of truth for all request state as it flows
 /// through the pipeline stages. Uses Rust's type system to enforce proper
 /// stage ordering at compile time.
-pub struct RequestContext {
+pub(crate) struct RequestContext {
     pub input: RequestInput,
     pub components: Arc<SharedComponents>,
     pub state: ProcessingState,
 }
 
 /// Immutable request input
-pub struct RequestInput {
+pub(crate) struct RequestInput {
     pub request_type: RequestType,
     pub headers: Option<HeaderMap>,
     pub model_id: Option<String>,
@@ -47,7 +46,7 @@ pub struct RequestInput {
 
 /// Request type variants
 /// Using Arc instead of Box to enable cheap cloning for background tasks
-pub enum RequestType {
+pub(crate) enum RequestType {
     Chat(Arc<ChatCompletionRequest>),
     Generate(Arc<GenerateRequest>),
     Responses(Arc<ResponsesRequest>),
@@ -56,15 +55,17 @@ pub enum RequestType {
 }
 
 /// Shared components (injected once at creation)
-pub struct SharedComponents {
+pub(crate) struct SharedComponents {
     pub tokenizer_registry: Arc<TokenizerRegistry>,
+    #[allow(dead_code)]
     pub tool_parser_factory: ToolParserFactory,
+    #[allow(dead_code)]
     pub reasoning_parser_factory: ReasoningParserFactory,
 }
 
 /// Mutable processing state (evolves through pipeline stages)
 #[derive(Default)]
-pub struct ProcessingState {
+pub(crate) struct ProcessingState {
     // Stage 1: Preparation outputs
     pub preparation: Option<PreparationOutput>,
 
@@ -92,7 +93,7 @@ pub struct ProcessingState {
 }
 
 /// Output from preparation stage (Step 1)
-pub struct PreparationOutput {
+pub(crate) struct PreparationOutput {
     /// Original text (for chat) or resolved text (for generate)
     pub original_text: Option<String>,
 
@@ -116,6 +117,7 @@ pub struct PreparationOutput {
     pub selection_text: Option<String>,
 
     /// Harmony messages for history tracking (Harmony only)
+    #[allow(dead_code)]
     pub harmony_messages: Option<Vec<super::harmony::HarmonyMessage>>,
 
     /// Stop token IDs for Harmony models
@@ -123,7 +125,7 @@ pub struct PreparationOutput {
 }
 
 /// Worker selection (Step 2)
-pub enum WorkerSelection {
+pub(crate) enum WorkerSelection {
     Single {
         worker: Arc<dyn Worker>,
     },
@@ -134,7 +136,7 @@ pub enum WorkerSelection {
 }
 
 /// Client selection (Step 3)
-pub enum ClientSelection {
+pub(crate) enum ClientSelection {
     Single {
         client: GrpcClient,
     },
@@ -146,17 +148,18 @@ pub enum ClientSelection {
 
 /// Dispatch metadata (Step 5)
 #[derive(Clone)]
-pub struct DispatchMetadata {
+pub(crate) struct DispatchMetadata {
     pub request_id: String,
     pub model: String,
     pub created: u64,
     pub weight_version: Option<String>,
+    #[allow(dead_code)]
     pub is_streaming: bool,
 }
 
 /// Load guards for worker load tracking
 /// Automatically decrements load when dropped
-pub enum LoadGuards {
+pub(crate) enum LoadGuards {
     Single(WorkerLoadGuard),
     Dual {
         prefill: WorkerLoadGuard,
@@ -200,19 +203,10 @@ impl LoadGuards {
 
 /// Response processing state (Step 6)
 #[derive(Default)]
-pub struct ResponseState {
+pub(crate) struct ResponseState {
     /// Stop sequence decoder
     pub stop_decoder: Option<StopSequenceDecoder>,
 
-    /// Per-index streaming state (for n>1 support)
-    pub streaming: StreamingState,
-
-    /// Collected responses (non-streaming)
-    pub collected: Option<Vec<ProtoGenerateComplete>>,
-
-    /// Collected embeddings (non-streaming)
-    pub collected_embeddings: Option<Vec<ProtoEmbedComplete>>,
-
     /// Execution result (streams from workers)
     pub execution_result: Option<ExecutionResult>,
 
@@ -221,32 +215,6 @@ pub struct ResponseState {
 
     /// Responses API iteration result (Harmony only, for tool loop orchestration)
     pub responses_iteration_result: Option<super::harmony::ResponsesIterationResult>,
-
-    // Harmony-specific parser state
-    /// Harmony parser for non-streaming (single parser for all indices)
-    pub harmony_parser: Option<super::harmony::HarmonyParserAdapter>,
-
-    /// Harmony parsers for streaming (one per index for n>1 support)
-    pub harmony_parser_per_index: Option<HashMap<usize, super::harmony::HarmonyParserAdapter>>,
-}
-
-/// Streaming state (per-choice tracking)
-#[derive(Default)]
-pub struct StreamingState {
-    pub is_firsts: HashMap<u32, bool>,
-    pub stream_buffers: HashMap<u32, String>,
-    pub finish_reasons: HashMap<u32, String>,
-    pub matched_stops: HashMap<u32, Option<Value>>,
-    pub prompt_tokens: HashMap<u32, u32>,
-    pub completion_tokens: HashMap<u32, u32>,
-    pub cached_tokens: HashMap<u32, u32>,
-
-    // Parser state (lazy initialization per index)
-    pub reasoning_parsers:
-        HashMap<u32, Arc<std::sync::Mutex<Box<dyn crate::reasoning_parser::ReasoningParser>>>>,
-    pub tool_parsers:
-        HashMap<u32, Arc<tokio::sync::Mutex<Box<dyn crate::tool_parser::ToolParser>>>>,
-    pub has_tool_calls: HashMap<u32, bool>,
 }
 
 impl RequestContext {
@@ -340,11 +308,6 @@ impl RequestContext {
         }
     }
 
-    /// Get reference to original request (type-safe)
-    pub fn request(&self) -> &RequestType {
-        &self.input.request_type
-    }
-
     /// Get chat request (panics if not chat)
     pub fn chat_request(&self) -> &ChatCompletionRequest {
         match &self.input.request_type {
@@ -377,14 +340,6 @@ impl RequestContext {
         }
     }
 
-    /// Get responses request (panics if not responses)
-    pub fn responses_request(&self) -> &ResponsesRequest {
-        match &self.input.request_type {
-            RequestType::Responses(req) => req.as_ref(),
-            _ => panic!("Expected responses request"),
-        }
-    }
-
     /// Get Arc clone of responses request (panics if not responses)
     pub fn responses_request_arc(&self) -> Arc<ResponsesRequest> {
         match &self.input.request_type {
@@ -393,38 +348,6 @@ impl RequestContext {
         }
     }
 
-    /// Get embedding request (panics if not embedding)
-    pub fn embedding_request(&self) -> &EmbeddingRequest {
-        match &self.input.request_type {
-            RequestType::Embedding(req) => req.as_ref(),
-            _ => panic!("Expected embedding request"),
-        }
-    }
-
-    /// Get Arc clone of embedding request (panics if not embedding)
-    pub fn embedding_request_arc(&self) -> Arc<EmbeddingRequest> {
-        match &self.input.request_type {
-            RequestType::Embedding(req) => Arc::clone(req),
-            _ => panic!("Expected embedding request"),
-        }
-    }
-
-    /// Get classify request (panics if not classify)
-    pub fn classify_request(&self) -> &ClassifyRequest {
-        match &self.input.request_type {
-            RequestType::Classify(req) => req.as_ref(),
-            _ => panic!("Expected classify request"),
-        }
-    }
-
-    /// Get Arc clone of classify request (panics if not classify)
-    pub fn classify_request_arc(&self) -> Arc<ClassifyRequest> {
-        match &self.input.request_type {
-            RequestType::Classify(req) => Arc::clone(req),
-            _ => panic!("Expected classify request"),
-        }
-    }
-
     /// Check if request is streaming
     pub fn is_streaming(&self) -> bool {
         match &self.input.request_type {
@@ -446,10 +369,12 @@ impl RequestContext {
 }
 
 impl WorkerSelection {
+    #[allow(dead_code)]
     pub fn is_dual(&self) -> bool {
         matches!(self, Self::Dual { .. })
     }
 
+    #[allow(dead_code)]
     pub fn single(&self) -> Option<&Arc<dyn Worker>> {
         match self {
             Self::Single { worker } => Some(worker),
@@ -476,6 +401,7 @@ impl WorkerSelection {
         }
     }
 
+    #[allow(dead_code)]
     #[allow(clippy::type_complexity)]
     pub fn dual(&self) -> Option<(&Arc<dyn Worker>, &Arc<dyn Worker>)> {
         match self {
@@ -484,6 +410,7 @@ impl WorkerSelection {
         }
     }
 
+    #[allow(dead_code)]
     pub fn prefill_worker(&self) -> Option<&Arc<dyn Worker>> {
         match self {
             Self::Dual { prefill, .. } => Some(prefill),
@@ -491,6 +418,7 @@ impl WorkerSelection {
         }
     }
 
+    #[allow(dead_code)]
     pub fn decode_worker(&self) -> Option<&Arc<dyn Worker>> {
         match self {
             Self::Dual { decode, .. } => Some(decode),
@@ -500,6 +428,7 @@ impl WorkerSelection {
 }
 
 impl ClientSelection {
+    #[allow(dead_code)]
     pub fn is_dual(&self) -> bool {
         matches!(self, Self::Dual { .. })
     }
@@ -518,6 +447,7 @@ impl ClientSelection {
         }
     }
 
+    #[allow(dead_code)]
     pub fn dual(&self) -> Option<(&GrpcClient, &GrpcClient)> {
         match self {
             Self::Dual { prefill, decode } => Some((prefill, decode)),
@@ -532,6 +462,7 @@ impl ClientSelection {
         }
     }
 
+    #[allow(dead_code)]
     pub fn prefill_client(&self) -> Option<&GrpcClient> {
         match self {
             Self::Dual { prefill, .. } => Some(prefill),
@@ -539,6 +470,7 @@ impl ClientSelection {
         }
     }
 
+    #[allow(dead_code)]
     pub fn prefill_client_mut(&mut self) -> Option<&mut GrpcClient> {
         match self {
             Self::Dual { prefill, .. } => Some(prefill),
@@ -546,6 +478,7 @@ impl ClientSelection {
         }
     }
 
+    #[allow(dead_code)]
     pub fn decode_client(&self) -> Option<&GrpcClient> {
         match self {
             Self::Dual { decode, .. } => Some(decode),
@@ -553,6 +486,7 @@ impl ClientSelection {
         }
     }
 
+    #[allow(dead_code)]
     pub fn decode_client_mut(&mut self) -> Option<&mut GrpcClient> {
         match self {
             Self::Dual { decode, .. } => Some(decode),
@@ -563,7 +497,7 @@ impl ClientSelection {
 
 /// Result of request execution (streams from workers)
 /// Uses ProtoStream to automatically abort on cancellation
-pub enum ExecutionResult {
+pub(crate) enum ExecutionResult {
     Single {
         stream: ProtoStream,
     },
@@ -579,7 +513,7 @@ pub enum ExecutionResult {
 
 /// Final processed response
 #[derive(Debug)]
-pub enum FinalResponse {
+pub(crate) enum FinalResponse {
     Chat(ChatCompletionResponse),
     /// Generate response is a Vec of GenerateResponse (n=1 returns single item, n>1 returns multiple)
     Generate(Vec<GenerateResponse>),

sgl-model-gateway/src/routers/grpc/harmony/builder.rs

@@ -113,7 +113,7 @@ fn has_custom_tools(tool_types: &[&str]) -> bool {
 ///
 /// Converts OpenAI-format requests into Harmony-encoded format with input_ids,
 /// stop tokens, and selection text for worker routing.
-pub struct HarmonyBuilder {
+pub(crate) struct HarmonyBuilder {
     encoding: &'static HarmonyEncoding,
 }
 

sgl-model-gateway/src/routers/grpc/harmony/detector.rs

@@ -5,7 +5,7 @@ use crate::core::{Worker, WorkerRegistry};
 /// Harmony model detector
 ///
 /// Detects if a model name indicates support for Harmony encoding/parsing.
-pub struct HarmonyDetector;
+pub(crate) struct HarmonyDetector;
 
 impl HarmonyDetector {
     /// Check if a worker is a Harmony/GPT-OSS model.

sgl-model-gateway/src/routers/grpc/harmony/mod.rs

@@ -29,28 +29,22 @@
 //! }
 //! ```
 
-pub mod builder;
-pub mod detector;
-pub mod parser;
-pub mod processor;
-pub mod responses;
-pub mod stages;
-pub mod streaming;
-pub mod types;
+pub(crate) mod builder;
+pub(crate) mod detector;
+pub(crate) mod parser;
+pub(crate) mod processor;
+pub(crate) mod responses;
+pub(crate) mod stages;
+pub(crate) mod streaming;
+pub(crate) mod types;
 
-// Re-export main types for convenience
-pub use builder::HarmonyBuilder;
-pub use detector::HarmonyDetector;
-pub use parser::HarmonyParserAdapter;
-pub use processor::{HarmonyResponseProcessor, ResponsesIterationResult};
-pub use responses::{
+// Re-export types that are accessed via harmony::TypeName
+pub(crate) use builder::HarmonyBuilder;
+pub(crate) use detector::HarmonyDetector;
+pub(crate) use parser::HarmonyParserAdapter;
+pub(crate) use processor::{HarmonyResponseProcessor, ResponsesIterationResult};
+pub(crate) use responses::{
     serve_harmony_responses, serve_harmony_responses_stream, HarmonyResponsesContext,
 };
-pub use stages::{
-    HarmonyPreparationStage, HarmonyRequestBuildingStage, HarmonyResponseProcessingStage,
-};
-pub use streaming::HarmonyStreamingProcessor;
-pub use types::{
-    FunctionDelta, HarmonyBuildOutput, HarmonyChannelDelta, HarmonyChannelOutput, HarmonyMessage,
-    ToolCallDelta,
-};
+pub(crate) use streaming::HarmonyStreamingProcessor;
+pub(crate) use types::HarmonyMessage;

sgl-model-gateway/src/routers/grpc/harmony/parser.rs

@@ -20,7 +20,7 @@ fn get_harmony_encoding() -> &'static HarmonyEncoding {
 ///
 /// Wraps openai_harmony::StreamableParser and provides methods for parsing
 /// complete responses and streaming chunks.
-pub struct HarmonyParserAdapter {
+pub(crate) struct HarmonyParserAdapter {
     parser: StreamableParser,
     prev_recipient: Option<String>,
     reasoning_token_count: u32,
@@ -517,6 +517,7 @@ impl HarmonyParserAdapter {
     /// Reset parser state
     ///
     /// Resets the parser to initial state for reuse
+    #[allow(dead_code)]
     pub fn reset(&mut self) -> Result<(), String> {
         // Create a new parser instance (StreamableParser doesn't have a reset method)
         let encoding = get_harmony_encoding();

sgl-model-gateway/src/routers/grpc/harmony/processor.rs

@@ -29,7 +29,7 @@ use crate::{
 ///
 /// Collects all output tokens from execution and parses them using
 /// HarmonyParserAdapter to extract the complete response.
-pub struct HarmonyResponseProcessor;
+pub(crate) struct HarmonyResponseProcessor;
 
 impl HarmonyResponseProcessor {
     /// Create a new Harmony response processor
@@ -155,7 +155,7 @@ impl Default for HarmonyResponseProcessor {
 ///
 /// Used by the MCP tool loop to determine whether to continue
 /// executing tools or return the final response.
-pub enum ResponsesIterationResult {
+pub(crate) enum ResponsesIterationResult {
     /// Tool calls found in commentary channel - continue MCP loop
     ToolCallsFound {
         tool_calls: Vec<ToolCall>,

sgl-model-gateway/src/routers/grpc/harmony/responses/context.rs

@@ -2,8 +2,6 @@
 
 use std::sync::Arc;
 
-use tokio::sync::mpsc;
-
 use crate::{
     data_connector::{ConversationItemStorage, ConversationStorage, ResponseStorage},
     mcp::McpManager,
@@ -15,7 +13,7 @@ use crate::{
 /// Contains all dependencies needed for multi-turn Responses API execution.
 /// Cheap to clone (all Arc references).
 #[derive(Clone)]
-pub struct HarmonyResponsesContext {
+pub(crate) struct HarmonyResponsesContext {
     /// Pipeline for executing Harmony requests
     pub pipeline: Arc<RequestPipeline>,
 
@@ -33,9 +31,6 @@ pub struct HarmonyResponsesContext {
 
     /// Conversation item storage for persisting conversation items
     pub conversation_item_storage: Arc<dyn ConversationItemStorage>,
-
-    /// Optional streaming sender (for future streaming support)
-    pub stream_tx: Option<mpsc::UnboundedSender<Result<String, String>>>,
 }
 
 impl HarmonyResponsesContext {
@@ -55,28 +50,6 @@ impl HarmonyResponsesContext {
             response_storage,
             conversation_storage,
             conversation_item_storage,
-            stream_tx: None,
-        }
-    }
-
-    /// Create with streaming support
-    pub fn with_streaming(
-        pipeline: Arc<RequestPipeline>,
-        components: Arc<SharedComponents>,
-        mcp_manager: Arc<McpManager>,
-        response_storage: Arc<dyn ResponseStorage>,
-        conversation_storage: Arc<dyn ConversationStorage>,
-        conversation_item_storage: Arc<dyn ConversationItemStorage>,
-        stream_tx: mpsc::UnboundedSender<Result<String, String>>,
-    ) -> Self {
-        Self {
-            pipeline,
-            components,
-            mcp_manager,
-            response_storage,
-            conversation_storage,
-            conversation_item_storage,
-            stream_tx: Some(stream_tx),
         }
     }
 }

sgl-model-gateway/src/routers/grpc/harmony/responses/execution.rs

@@ -20,7 +20,7 @@ use crate::{
 /// Tool execution result
 ///
 /// Contains the result of executing a single MCP tool.
-pub struct ToolResult {
+pub(crate) struct ToolResult {
     /// Tool call ID (for matching with request)
     pub call_id: String,
 
@@ -202,7 +202,7 @@ pub(super) async fn execute_mcp_tools(
 ///
 /// Converts MCP Tool entries (from rmcp SDK) to ResponseTool format so the model
 /// knows about available MCP tools when making tool calls.
-pub fn convert_mcp_tools_to_response_tools(mcp_tools: &[mcp::Tool]) -> Vec<ResponseTool> {
+pub(crate) fn convert_mcp_tools_to_response_tools(mcp_tools: &[mcp::Tool]) -> Vec<ResponseTool> {
     mcp_tools
         .iter()
         .map(|tool_info| ResponseTool {

sgl-model-gateway/src/routers/grpc/harmony/responses/mod.rs

@@ -19,14 +19,14 @@
 //! - `execution` - MCP tool execution logic
 //! - `common` - Shared helpers and state tracking
 
-mod common;
-mod context;
-mod execution;
-mod non_streaming;
-mod streaming;
+pub(crate) mod common;
+pub(crate) mod context;
+pub(crate) mod execution;
+pub(crate) mod non_streaming;
+pub(crate) mod streaming;
 
-// Public exports
-pub use context::HarmonyResponsesContext;
-pub use execution::{convert_mcp_tools_to_response_tools, ToolResult};
-pub use non_streaming::serve_harmony_responses;
-pub use streaming::serve_harmony_responses_stream;
+// Re-export types accessed via harmony::responses::TypeName
+pub(crate) use context::HarmonyResponsesContext;
+pub(crate) use execution::ToolResult;
+pub(crate) use non_streaming::serve_harmony_responses;
+pub(crate) use streaming::serve_harmony_responses_stream;

sgl-model-gateway/src/routers/grpc/harmony/responses/non_streaming.rs

@@ -46,7 +46,7 @@ use crate::{
 ///    - Build next request with tool results
 ///    - Repeat from step 1 (full pipeline re-execution)
 /// 4. If no tool calls, return final response
-pub async fn serve_harmony_responses(
+pub(crate) async fn serve_harmony_responses(
     ctx: &HarmonyResponsesContext,
     request: ResponsesRequest,
 ) -> Result<ResponsesResponse, Response> {

sgl-model-gateway/src/routers/grpc/harmony/responses/streaming.rs

@@ -36,7 +36,7 @@ use crate::{
 ///
 /// This is the streaming equivalent of `serve_harmony_responses()`.
 /// Emits SSE events for lifecycle, MCP list_tools, and per-iteration streaming.
-pub async fn serve_harmony_responses_stream(
+pub(crate) async fn serve_harmony_responses_stream(
     ctx: &HarmonyResponsesContext,
     request: ResponsesRequest,
 ) -> Response {

sgl-model-gateway/src/routers/grpc/harmony/stages/mod.rs

@@ -5,10 +5,10 @@
 //! - HarmonyRequestBuildingStage: Token-based request building
 //! - HarmonyResponseProcessingStage: Harmony channel parsing
 
-pub mod preparation;
-pub mod request_building;
-pub mod response_processing;
+pub(crate) mod preparation;
+pub(crate) mod request_building;
+pub(crate) mod response_processing;
 
-pub use preparation::HarmonyPreparationStage;
-pub use request_building::HarmonyRequestBuildingStage;
-pub use response_processing::HarmonyResponseProcessingStage;
+pub(crate) use preparation::HarmonyPreparationStage;
+pub(crate) use request_building::HarmonyRequestBuildingStage;
+pub(crate) use response_processing::HarmonyResponseProcessingStage;

sgl-model-gateway/src/routers/grpc/harmony/stages/preparation.rs

@@ -26,7 +26,7 @@ use crate::{
 ///
 /// Replaces the regular PreparationStage for Harmony models.
 /// Converts chat/generate requests to Harmony-encoded token_ids and extraction_text.
-pub struct HarmonyPreparationStage {
+pub(crate) struct HarmonyPreparationStage {
     builder: HarmonyBuilder,
 }
 
@@ -387,7 +387,9 @@ impl HarmonyPreparationStage {
 /// - Without reasoning: triggers on `<|channel|>final` (goes directly to final channel)
 ///
 /// This is used for the Responses API text.format field (json_object or json_schema).
-pub fn build_text_format_structural_tag(schema: &serde_json::Value) -> Result<String, String> {
+pub(crate) fn build_text_format_structural_tag(
+    schema: &serde_json::Value,
+) -> Result<String, String> {
     let structural_tag = json!({
         "format": {
             "type": "triggered_tags",

sgl-model-gateway/src/routers/grpc/harmony/stages/request_building.rs

@@ -18,7 +18,7 @@ use crate::routers::{
 ///
 /// Takes the Harmony-encoded input_ids from preparation and builds a proto::GenerateRequest.
 /// Unlike regular request building, this uses token_ids directly (Harmony encoding handles messages).
-pub struct HarmonyRequestBuildingStage {
+pub(crate) struct HarmonyRequestBuildingStage {
     inject_pd_metadata: bool,
 }
 

sgl-model-gateway/src/routers/grpc/harmony/stages/response_processing.rs

@@ -19,7 +19,7 @@ use crate::routers::{
 ///
 /// Takes output tokens from execution and parses them using HarmonyParserAdapter
 /// to extract analysis, tool calls, and final response text from Harmony channels.
-pub struct HarmonyResponseProcessingStage {
+pub(crate) struct HarmonyResponseProcessingStage {
     processor: HarmonyResponseProcessor,
     streaming_processor: Arc<HarmonyStreamingProcessor>,
 }

sgl-model-gateway/src/routers/grpc/harmony/streaming.rs

@@ -105,7 +105,7 @@ impl ToolCallMode {
 ///
 /// Returns an SSE stream that parses Harmony tokens incrementally and
 /// emits ChatCompletionChunk events for streaming responses.
-pub struct HarmonyStreamingProcessor;
+pub(crate) struct HarmonyStreamingProcessor;
 
 impl HarmonyStreamingProcessor {
     /// Create a new Harmony streaming processor

sgl-model-gateway/src/routers/grpc/harmony/types.rs

@@ -10,11 +10,12 @@ use crate::protocols::common::ToolCall;
 ///
 /// Represents messages in the Harmony encoding format with role and content.
 #[derive(Debug, Clone, Serialize, Deserialize)]
-pub struct HarmonyMessage {
+pub(crate) struct HarmonyMessage {
     pub role: String,
     pub content: String,
 }
 
+#[allow(dead_code)]
 impl HarmonyMessage {
     pub fn new(role: impl Into<String>, content: impl Into<String>) -> Self {
         Self {
@@ -67,7 +68,7 @@ impl HarmonyMessage {
 /// Contains the encoded input_ids, stop tokens, selection text for worker routing,
 /// and the Harmony message history.
 #[derive(Debug, Clone)]
-pub struct HarmonyBuildOutput {
+pub(crate) struct HarmonyBuildOutput {
     /// Encoded token IDs to send to the model
     pub input_ids: Vec<u32>,
 
@@ -85,7 +86,7 @@ pub struct HarmonyBuildOutput {
 ///
 /// Represents the complete response after parsing analysis, commentary, and final channels.
 #[derive(Debug, Clone)]
-pub struct HarmonyChannelOutput {
+pub(crate) struct HarmonyChannelOutput {
     /// Analysis/reasoning content (from analysis channel)
     pub analysis: Option<String>,
 
@@ -109,7 +110,8 @@ pub struct HarmonyChannelOutput {
 ///
 /// Represents incremental updates as tokens are parsed from the stream.
 #[derive(Debug, Clone)]
-pub struct HarmonyChannelDelta {
+#[allow(dead_code)]
+pub(crate) struct HarmonyChannelDelta {
     /// Delta for analysis/reasoning content
     pub analysis_delta: Option<String>,
 
@@ -125,7 +127,7 @@ pub struct HarmonyChannelDelta {
 
 /// Tool call delta for streaming
 #[derive(Debug, Clone, Serialize, Deserialize)]
-pub struct ToolCallDelta {
+pub(crate) struct ToolCallDelta {
     pub index: usize,
     pub id: Option<String>,
     pub function: Option<FunctionDelta>,
@@ -133,7 +135,7 @@ pub struct ToolCallDelta {
 
 /// Function call delta for streaming
 #[derive(Debug, Clone, Serialize, Deserialize)]
-pub struct FunctionDelta {
+pub(crate) struct FunctionDelta {
     pub name: Option<String>,
     pub arguments: Option<String>,
 }

sgl-model-gateway/src/routers/grpc/mod.rs

@@ -2,21 +2,22 @@
 
 use crate::{grpc_client::sglang_proto::MultimodalInputs, protocols::common::StringOrArray};
 
-pub mod client;
-pub mod common;
-pub mod context;
-pub mod harmony;
-pub mod pd_router;
-pub mod pipeline;
-pub mod proto_wrapper;
-pub mod regular;
-pub mod router;
-pub mod utils;
+pub mod client; // Used by core/
+pub(crate) mod common;
+pub(crate) mod context;
+pub(crate) mod harmony;
+pub(crate) mod pd_router; // Used by routers/factory
+pub(crate) mod pipeline;
+pub(crate) mod proto_wrapper;
+pub(crate) mod regular;
+pub(crate) mod router; // Used by routers/factory
+pub(crate) mod utils; // Used by routers/http
 
 /// Processed chat messages ready for gRPC generation
 #[derive(Debug)]
-pub struct ProcessedMessages {
+pub(crate) struct ProcessedMessages {
     pub text: String,
     pub multimodal_inputs: Option<MultimodalInputs>,
+    #[allow(dead_code)]
     pub stop_sequences: Option<StringOrArray>,
 }

sgl-model-gateway/src/routers/grpc/pipeline.rs

@@ -48,7 +48,7 @@ use crate::{
 /// Orchestrates all stages from request preparation to response delivery.
 /// Configured differently for regular vs PD mode.
 #[derive(Clone)]
-pub struct RequestPipeline {
+pub(crate) struct RequestPipeline {
     stages: Arc<Vec<Box<dyn PipelineStage>>>,
     /// Backend type for metrics labeling
     backend_type: &'static str,
@@ -129,6 +129,7 @@ impl RequestPipeline {
     }
 
     /// Create a Harmony PD (prefill-decode) pipeline
+    #[allow(dead_code)]
     pub fn new_harmony_pd(
         worker_registry: Arc<WorkerRegistry>,
         policy_registry: Arc<PolicyRegistry>,
@@ -369,9 +370,6 @@ impl RequestPipeline {
         components: Arc<SharedComponents>,
     ) -> Response {
         let start = Instant::now();
-        // Clone model_id for metrics before moving into context
-        // GenerateRequest doesn't have a model field, so we use model_id
-        let model_for_metrics = model_id.clone();
         let streaming = request.stream;
 
         // Record request start
@@ -379,12 +377,12 @@ impl RequestPipeline {
             metrics_labels::ROUTER_GRPC,
             self.backend_type,
             metrics_labels::CONNECTION_GRPC,
-            model_for_metrics.as_deref().unwrap_or("unknown"),
+            model_id.as_deref().unwrap_or(UNKNOWN_MODEL_ID),
             metrics_labels::ENDPOINT_GENERATE,
             bool_to_static_str(streaming),
         );
 
-        let mut ctx = RequestContext::for_generate(request, headers, model_id, components);
+        let mut ctx = RequestContext::for_generate(request, headers, model_id.clone(), components);
 
         for stage in self.stages.iter() {
             match stage.execute(&mut ctx).await {
@@ -393,7 +391,7 @@ impl RequestPipeline {
                         metrics_labels::ROUTER_GRPC,
                         self.backend_type,
                         metrics_labels::CONNECTION_GRPC,
-                        model_for_metrics.as_deref().unwrap_or("unknown"),
+                        model_id.as_deref().unwrap_or(UNKNOWN_MODEL_ID),
                         metrics_labels::ENDPOINT_GENERATE,
                         start.elapsed(),
                     );
@@ -405,7 +403,7 @@ impl RequestPipeline {
                         metrics_labels::ROUTER_GRPC,
                         self.backend_type,
                         metrics_labels::CONNECTION_GRPC,
-                        model_for_metrics.as_deref().unwrap_or("unknown"),
+                        model_id.as_deref().unwrap_or(UNKNOWN_MODEL_ID),
                         metrics_labels::ENDPOINT_GENERATE,
                         error_type_from_status(response.status()),
                     );
@@ -425,7 +423,7 @@ impl RequestPipeline {
                     metrics_labels::ROUTER_GRPC,
                     self.backend_type,
                     metrics_labels::CONNECTION_GRPC,
-                    model_for_metrics.as_deref().unwrap_or("unknown"),
+                    model_id.as_deref().unwrap_or(UNKNOWN_MODEL_ID),
                     metrics_labels::ENDPOINT_GENERATE,
                     start.elapsed(),
                 );
@@ -442,7 +440,7 @@ impl RequestPipeline {
                     metrics_labels::ROUTER_GRPC,
                     self.backend_type,
                     metrics_labels::CONNECTION_GRPC,
-                    model_for_metrics.as_deref().unwrap_or("unknown"),
+                    model_id.as_deref().unwrap_or(UNKNOWN_MODEL_ID),
                     metrics_labels::ENDPOINT_GENERATE,
                     metrics_labels::ERROR_INTERNAL,
                 );
@@ -457,7 +455,7 @@ impl RequestPipeline {
                     metrics_labels::ROUTER_GRPC,
                     self.backend_type,
                     metrics_labels::CONNECTION_GRPC,
-                    model_for_metrics.as_deref().unwrap_or("unknown"),
+                    model_id.as_deref().unwrap_or(UNKNOWN_MODEL_ID),
                     metrics_labels::ENDPOINT_GENERATE,
                     metrics_labels::ERROR_INTERNAL,
                 );
@@ -541,9 +539,7 @@ impl RequestPipeline {
             ctx.state.response.final_response
         );
         match ctx.state.response.final_response {
-            Some(FinalResponse::Embedding(_)) => {
-                error!("execute_embeddings: Embedding FinalResponse found, but pipeline finished without returning response directly. This should be handled by the last stage.");
-                // Already handled in ResponseProcessingStage, but just in case
+            Some(FinalResponse::Embedding(response)) => {
                 Metrics::record_router_duration(
                     metrics_labels::ROUTER_GRPC,
                     self.backend_type,
@@ -552,11 +548,7 @@ impl RequestPipeline {
                     metrics_labels::ENDPOINT_EMBEDDINGS,
                     start.elapsed(),
                 );
-                // The response should have been returned by the last stage
-                error::internal_error(
-                    "pipeline_fallthrough",
-                    "Pipeline finished without returning response",
-                )
+                axum::Json(response).into_response()
             }
             Some(_) => {
                 error!(function = "execute_embeddings", "Wrong response type");
@@ -647,8 +639,7 @@ impl RequestPipeline {
             ctx.state.response.final_response
         );
         match ctx.state.response.final_response {
-            Some(FinalResponse::Classify(_)) => {
-                error!("execute_classify: Classify FinalResponse found, but pipeline finished without returning response directly. This should be handled by the last stage.");
+            Some(FinalResponse::Classify(response)) => {
                 Metrics::record_router_duration(
                     metrics_labels::ROUTER_GRPC,
                     self.backend_type,
@@ -657,10 +648,7 @@ impl RequestPipeline {
                     metrics_labels::ENDPOINT_CLASSIFY,
                     start.elapsed(),
                 );
-                error::internal_error(
-                    "pipeline_fallthrough",
-                    "Pipeline finished without returning response",
-                )
+                axum::Json(response).into_response()
             }
             Some(_) => {
                 error!(function = "execute_classify", "Wrong response type");

sgl-model-gateway/src/routers/grpc/proto_wrapper.rs

@@ -20,20 +20,7 @@ pub enum ProtoRequest {
 }
 
 impl ProtoRequest {
-    pub fn as_generate(&self) -> &ProtoGenerateRequest {
-        match self {
-            Self::Generate(req) => req,
-            _ => panic!("Expected Generate request"),
-        }
-    }
-
-    pub fn as_embed(&self) -> &ProtoEmbedRequest {
-        match self {
-            Self::Embed(req) => req,
-            _ => panic!("Expected Embed request"),
-        }
-    }
-
+    /// Get request ID from either variant
     pub fn request_id(&self) -> &str {
         match self {
             Self::Generate(req) => req.request_id(),

sgl-model-gateway/src/routers/grpc/regular/mod.rs

@@ -3,7 +3,7 @@
 //! This module contains all code specific to regular tokenizer-based models,
 //! including pipeline stages, response processing, and streaming.
 
-pub mod processor;
-pub mod responses;
-pub mod stages;
-pub mod streaming;
+pub(crate) mod processor;
+pub(crate) mod responses;
+pub(crate) mod stages;
+pub(crate) mod streaming;

sgl-model-gateway/src/routers/grpc/regular/processor.rs

@@ -34,7 +34,7 @@ use crate::{
 
 /// Unified response processor for both routers
 #[derive(Clone)]
-pub struct ResponseProcessor {
+pub(crate) struct ResponseProcessor {
     pub tool_parser_factory: ToolParserFactory,
     pub reasoning_parser_factory: ReasoningParserFactory,
     pub configured_tool_parser: Option<String>,

sgl-model-gateway/src/routers/grpc/regular/responses/context.rs

@@ -19,7 +19,7 @@ use crate::{
 ///
 /// This struct enables cancelling both the Rust task AND the Python scheduler processing.
 /// The client field is lazily initialized during pipeline execution.
-pub struct BackgroundTaskInfo {
+pub(crate) struct BackgroundTaskInfo {
     /// Tokio task handle for aborting the Rust task
     pub handle: JoinHandle<()>,
     /// gRPC request_id sent to Python scheduler (chatcmpl-* prefix)
@@ -32,7 +32,7 @@ pub struct BackgroundTaskInfo {
 ///
 /// All fields are Arc/shared references, so cloning this context is cheap.
 #[derive(Clone)]
-pub struct ResponsesContext {
+pub(crate) struct ResponsesContext {
     /// Chat pipeline for executing requests
     pub pipeline: Arc<RequestPipeline>,
 
@@ -40,6 +40,7 @@ pub struct ResponsesContext {
     pub components: Arc<SharedComponents>,
 
     /// Worker registry for validation
+    #[allow(dead_code)]
     pub worker_registry: Arc<WorkerRegistry>,
 
     /// Response storage backend

sgl-model-gateway/src/routers/grpc/regular/responses/conversions.rs

@@ -33,7 +33,7 @@ use crate::{
 /// - `tools` → function tools extracted from ResponseTools
 /// - `tool_choice` → passed through from request
 /// - Response-specific fields (previous_response_id, conversation) are handled by router
-pub fn responses_to_chat(req: &ResponsesRequest) -> Result<ChatCompletionRequest, String> {
+pub(crate) fn responses_to_chat(req: &ResponsesRequest) -> Result<ChatCompletionRequest, String> {
     let mut messages = Vec::new();
 
     // 1. Add system message if instructions provided
@@ -271,7 +271,7 @@ fn map_text_to_response_format(text: &Option<TextConfig>) -> Option<ResponseForm
 /// - `choices[0].message` → `output` array (convert to ResponseOutputItem::Message)
 /// - `choices[0].finish_reason` → determines `status` (stop/length → Completed)
 /// - `created` timestamp → `created_at`
-pub fn chat_to_responses(
+pub(crate) fn chat_to_responses(
     chat_resp: &ChatCompletionResponse,
     original_req: &ResponsesRequest,
     response_id_override: Option<String>,

sgl-model-gateway/src/routers/grpc/regular/responses/handlers.rs

@@ -46,7 +46,7 @@ use crate::{
 /// Main handler for POST /v1/responses
 ///
 /// Validates request, determines execution mode (sync/streaming), and delegates
-pub async fn route_responses(
+pub(crate) async fn route_responses(
     ctx: &ResponsesContext,
     request: Arc<ResponsesRequest>,
     headers: Option<http::HeaderMap>,

sgl-model-gateway/src/routers/grpc/regular/responses/mod.rs

@@ -19,5 +19,5 @@ mod non_streaming;
 mod streaming;
 
 // Public exports
-pub use context::{BackgroundTaskInfo, ResponsesContext};
-pub use handlers::route_responses;
+pub(crate) use context::ResponsesContext;
+pub(crate) use handlers::route_responses;

sgl-model-gateway/src/routers/grpc/regular/stages/chat/mod.rs

@@ -7,6 +7,6 @@ mod preparation;
 mod request_building;
 mod response_processing;
 
-pub use preparation::ChatPreparationStage;
-pub use request_building::ChatRequestBuildingStage;
-pub use response_processing::ChatResponseProcessingStage;
+pub(crate) use preparation::ChatPreparationStage;
+pub(crate) use request_building::ChatRequestBuildingStage;
+pub(crate) use response_processing::ChatResponseProcessingStage;

sgl-model-gateway/src/routers/grpc/regular/stages/chat/preparation.rs

@@ -22,7 +22,7 @@ use crate::{
 ///
 /// Extracts chat-specific preparation logic from the old unified PreparationStage.
 /// This is a direct extraction without architectural changes.
-pub struct ChatPreparationStage;
+pub(crate) struct ChatPreparationStage;
 
 #[async_trait]
 impl PipelineStage for ChatPreparationStage {

sgl-model-gateway/src/routers/grpc/regular/stages/chat/request_building.rs

@@ -18,7 +18,7 @@ use crate::routers::{
 /// Chat request building stage
 ///
 /// Extracts chat-specific request building logic from the old unified RequestBuildingStage.
-pub struct ChatRequestBuildingStage {
+pub(crate) struct ChatRequestBuildingStage {
     inject_pd_metadata: bool,
 }
 

sgl-model-gateway/src/routers/grpc/regular/stages/chat/response_processing.rs

@@ -19,9 +19,7 @@ use crate::routers::{
 };
 
 /// Chat response processing stage
-///
-/// Extracts chat-specific response processing logic from the old unified ResponseProcessingStage.
-pub struct ChatResponseProcessingStage {
+pub(crate) struct ChatResponseProcessingStage {
     processor: processor::ResponseProcessor,
     streaming_processor: Arc<streaming::StreamingProcessor>,
 }

sgl-model-gateway/src/routers/grpc/regular/stages/classify/mod.rs

@@ -4,6 +4,6 @@
 //! as the scheduler treats classify as an embedding request and returns logits.
 //! Only response processing is classify-specific (softmax + label mapping).
 
-pub mod response_processing;
+pub(crate) mod response_processing;
 
-pub use response_processing::ClassifyResponseProcessingStage;
+pub(crate) use response_processing::ClassifyResponseProcessingStage;

sgl-model-gateway/src/routers/grpc/regular/stages/classify/response_processing.rs

@@ -10,10 +10,7 @@
 use std::collections::HashMap;
 
 use async_trait::async_trait;
-use axum::{
-    response::{IntoResponse, Response},
-    Json,
-};
+use axum::response::Response;
 use tracing::error;
 
 use crate::{
@@ -37,7 +34,7 @@ use crate::{
 ///
 /// The stage is stateless - id2label mapping is obtained from the
 /// selected worker's model card at runtime.
-pub struct ClassifyResponseProcessingStage;
+pub(crate) struct ClassifyResponseProcessingStage;
 
 impl ClassifyResponseProcessingStage {
     /// Create a new classify response processing stage.
@@ -205,11 +202,10 @@ impl PipelineStage for ClassifyResponseProcessingStage {
             usage,
         );
 
-        // Store in context
-        ctx.state.response.final_response = Some(FinalResponse::Classify(response.clone()));
+        // Store in context for pipeline to extract
+        ctx.state.response.final_response = Some(FinalResponse::Classify(response));
 
-        // Return HTTP response
-        Ok(Some(Json(response).into_response()))
+        Ok(None)
     }
 
     fn name(&self) -> &'static str {

sgl-model-gateway/src/routers/grpc/regular/stages/embedding/mod.rs

@@ -1,3 +1,3 @@
-pub mod preparation;
-pub mod request_building;
-pub mod response_processing;
+pub(crate) mod preparation;
+pub(crate) mod request_building;
+pub(crate) mod response_processing;

sgl-model-gateway/src/routers/grpc/regular/stages/embedding/preparation.rs

@@ -16,7 +16,7 @@ use crate::{
     },
 };
 
-pub struct EmbeddingPreparationStage;
+pub(crate) struct EmbeddingPreparationStage;
 
 impl EmbeddingPreparationStage {
     pub fn new() -> Self {

sgl-model-gateway/src/routers/grpc/regular/stages/embedding/request_building.rs

@@ -15,7 +15,7 @@ use crate::routers::{
 };
 
 /// Request building stage for embedding requests
-pub struct EmbeddingRequestBuildingStage;
+pub(crate) struct EmbeddingRequestBuildingStage;
 
 impl EmbeddingRequestBuildingStage {
     pub fn new() -> Self {

sgl-model-gateway/src/routers/grpc/regular/stages/embedding/response_processing.rs

@@ -1,10 +1,7 @@
 //! Response processing stage for embedding requests
 
 use async_trait::async_trait;
-use axum::{
-    response::{IntoResponse, Response},
-    Json,
-};
+use axum::response::Response;
 use tracing::error;
 
 use crate::{
@@ -20,7 +17,7 @@ use crate::{
 };
 
 /// Response processing stage for embedding requests
-pub struct EmbeddingResponseProcessingStage;
+pub(crate) struct EmbeddingResponseProcessingStage;
 
 impl EmbeddingResponseProcessingStage {
     pub fn new() -> Self {
@@ -65,12 +62,10 @@ impl PipelineStage for EmbeddingResponseProcessingStage {
             .convert_response(ctx, proto_response)
             .map_err(|boxed_err| *boxed_err)?;
 
-        // Store in context
-        ctx.state.response.final_response =
-            Some(FinalResponse::Embedding(embedding_response.clone()));
+        // Store in context for pipeline to extract
+        ctx.state.response.final_response = Some(FinalResponse::Embedding(embedding_response));
 
-        // Return the HTTP response directly
-        Ok(Some(Json(embedding_response).into_response()))
+        Ok(None)
     }
 
     fn name(&self) -> &'static str {

sgl-model-gateway/src/routers/grpc/regular/stages/generate/mod.rs

@@ -7,6 +7,6 @@ mod preparation;
 mod request_building;
 mod response_processing;
 
-pub use preparation::GeneratePreparationStage;
-pub use request_building::GenerateRequestBuildingStage;
-pub use response_processing::GenerateResponseProcessingStage;
+pub(crate) use preparation::GeneratePreparationStage;
+pub(crate) use request_building::GenerateRequestBuildingStage;
+pub(crate) use response_processing::GenerateResponseProcessingStage;

sgl-model-gateway/src/routers/grpc/regular/stages/generate/preparation.rs

@@ -23,7 +23,7 @@ use crate::{
 ///
 /// Extracts generate-specific preparation logic from the old unified PreparationStage.
 /// This is a direct extraction without architectural changes.
-pub struct GeneratePreparationStage;
+pub(crate) struct GeneratePreparationStage;
 
 #[async_trait]
 impl PipelineStage for GeneratePreparationStage {

sgl-model-gateway/src/routers/grpc/regular/stages/generate/request_building.rs

@@ -18,7 +18,7 @@ use crate::routers::{
 /// Generate request building stage
 ///
 /// Extracts generate-specific request building logic from the old unified RequestBuildingStage.
-pub struct GenerateRequestBuildingStage {
+pub(crate) struct GenerateRequestBuildingStage {
     inject_pd_metadata: bool,
 }
 

sgl-model-gateway/src/routers/grpc/regular/stages/generate/response_processing.rs

@@ -18,7 +18,7 @@ use crate::routers::{
 /// Generate response processing stage
 ///
 /// Extracts generate-specific response processing logic from the old unified ResponseProcessingStage.
-pub struct GenerateResponseProcessingStage {
+pub(crate) struct GenerateResponseProcessingStage {
     processor: processor::ResponseProcessor,
     streaming_processor: Arc<streaming::StreamingProcessor>,
 }

sgl-model-gateway/src/routers/grpc/regular/stages/mod.rs

@@ -2,19 +2,15 @@
 //!
 //! This module defines stages specific to regular tokenizer-based models.
 
-pub mod chat;
-pub mod classify;
-pub mod embedding;
-pub mod generate;
-mod preparation;
-mod request_building;
-mod response_processing;
+pub(crate) mod chat;
+pub(crate) mod classify;
+pub(crate) mod embedding;
+pub(crate) mod generate;
+pub(crate) mod preparation;
+pub(crate) mod request_building;
+pub(crate) mod response_processing;
 
-pub use chat::{ChatPreparationStage, ChatRequestBuildingStage, ChatResponseProcessingStage};
-pub use classify::ClassifyResponseProcessingStage;
-pub use generate::{
-    GeneratePreparationStage, GenerateRequestBuildingStage, GenerateResponseProcessingStage,
-};
-pub use preparation::PreparationStage;
-pub use request_building::RequestBuildingStage;
-pub use response_processing::ResponseProcessingStage;
+// Re-export main stages used by pipeline
+pub(crate) use preparation::PreparationStage;
+pub(crate) use request_building::RequestBuildingStage;
+pub(crate) use response_processing::ResponseProcessingStage;

sgl-model-gateway/src/routers/grpc/regular/stages/preparation.rs

@@ -20,7 +20,7 @@ use crate::routers::{
 };
 
 /// Preparation stage (delegates to endpoint-specific implementations)
-pub struct PreparationStage {
+pub(crate) struct PreparationStage {
     chat_stage: ChatPreparationStage,
     generate_stage: GeneratePreparationStage,
     embedding_stage: EmbeddingPreparationStage,

sgl-model-gateway/src/routers/grpc/regular/stages/request_building.rs

@@ -17,7 +17,7 @@ use crate::routers::{
 };
 
 /// Request building stage (delegates to endpoint-specific implementations)
-pub struct RequestBuildingStage {
+pub(crate) struct RequestBuildingStage {
     chat_stage: ChatRequestBuildingStage,
     generate_stage: GenerateRequestBuildingStage,
     embedding_stage: EmbeddingRequestBuildingStage,

sgl-model-gateway/src/routers/grpc/regular/stages/response_processing.rs

@@ -21,7 +21,7 @@ use crate::routers::{
 };
 
 /// Response processing stage (delegates to endpoint-specific implementations)
-pub struct ResponseProcessingStage {
+pub(crate) struct ResponseProcessingStage {
     chat_stage: ChatResponseProcessingStage,
     generate_stage: GenerateResponseProcessingStage,
     embedding_stage: EmbeddingResponseProcessingStage,

sgl-model-gateway/src/routers/grpc/regular/streaming.rs

@@ -38,7 +38,7 @@ use crate::{
 
 /// Shared streaming processor for both single and dual dispatch modes
 #[derive(Clone)]
-pub struct StreamingProcessor {
+pub(crate) struct StreamingProcessor {
     tool_parser_factory: ToolParserFactory,
     reasoning_parser_factory: ReasoningParserFactory,
     configured_tool_parser: Option<String>,
@@ -1324,7 +1324,9 @@ impl StreamingProcessor {
 }
 
 /// Build SSE response with proper headers
-pub fn build_sse_response(rx: mpsc::UnboundedReceiver<Result<Bytes, io::Error>>) -> Response {
+pub(crate) fn build_sse_response(
+    rx: mpsc::UnboundedReceiver<Result<Bytes, io::Error>>,
+) -> Response {
     let stream = UnboundedReceiverStream::new(rx);
     let mut response = Response::new(Body::from_stream(stream));
     *response.status_mut() = StatusCode::OK;

sgl-model-gateway/src/routers/grpc/utils.rs

@@ -49,7 +49,7 @@ use crate::{
 /// preparation stages (chat, generate, embedding).
 ///
 /// Returns the tokenizer Arc, which is also cached in `ctx.state.tokenizer`.
-pub fn resolve_tokenizer(
+pub(crate) fn resolve_tokenizer(
     ctx: &mut RequestContext,
     stage_name: &str,
 ) -> Result<Arc<dyn Tokenizer>, Box<Response>> {
@@ -87,7 +87,9 @@ pub fn resolve_tokenizer(
 }
 
 /// Get gRPC client from worker, returning appropriate error response on failure
-pub async fn get_grpc_client_from_worker(worker: &Arc<dyn Worker>) -> Result<GrpcClient, Response> {
+pub(crate) async fn get_grpc_client_from_worker(
+    worker: &Arc<dyn Worker>,
+) -> Result<GrpcClient, Response> {
     // Get cached client from worker (or create one if not cached yet)
     let client_arc = worker
         .get_grpc_client()
@@ -157,7 +159,7 @@ fn process_tool_call_arguments(messages: &mut [Value]) -> Result<(), String> {
 }
 
 /// Process messages based on content format for ANY message type
-pub fn process_content_format(
+pub(crate) fn process_content_format(
     messages: &[ChatMessage],
     content_format: ChatTemplateContentFormat,
 ) -> Result<Vec<Value>, String> {
@@ -227,7 +229,7 @@ fn transform_content_field(content_value: &mut Value, content_format: ChatTempla
 
 /// Generate tool constraints for structured generation
 /// Note: tools should already be filtered if needed (by allowed_tools or specific function)
-pub fn generate_tool_constraints(
+pub(crate) fn generate_tool_constraints(
     tools: &[Tool],
     tool_choice: &Option<ToolChoice>,
     _model: &str,
@@ -343,7 +345,7 @@ fn build_required_array_schema(tools: &[Tool]) -> Result<String, String> {
 ///
 /// Returns filtered tools if filtering is needed, otherwise returns None.
 /// Used by both Chat API and Responses API (Harmony) for constraint generation.
-pub fn filter_tools_by_tool_choice(
+pub(crate) fn filter_tools_by_tool_choice(
     tools: &[Tool],
     tool_choice: &Option<ToolChoice>,
 ) -> Option<Vec<Tool>> {
@@ -377,7 +379,7 @@ pub fn filter_tools_by_tool_choice(
 ///
 /// Note: Tool existence is validated earlier in ChatCompletionRequest::validate(),
 /// so this function assumes tool_choice references valid tools.
-pub fn filter_chat_request_by_tool_choice(
+pub(crate) fn filter_chat_request_by_tool_choice(
     body: &ChatCompletionRequest,
 ) -> std::borrow::Cow<'_, ChatCompletionRequest> {
     if let Some(tools) = &body.tools {
@@ -394,7 +396,7 @@ pub fn filter_chat_request_by_tool_choice(
 
 /// Process chat messages and apply template (shared by both routers)
 /// Requires HuggingFace tokenizer with chat template support
-pub fn process_chat_messages(
+pub(crate) fn process_chat_messages(
     request: &ChatCompletionRequest,
     tokenizer: &dyn Tokenizer,
 ) -> Result<ProcessedMessages, String> {
@@ -515,7 +517,7 @@ pub fn process_chat_messages(
 }
 
 /// Create a StopSequenceDecoder from stop parameters
-pub fn create_stop_decoder(
+pub(crate) fn create_stop_decoder(
     tokenizer: &Arc<dyn Tokenizer>,
     stop: Option<&StringOrArray>,
     stop_token_ids: Option<&Vec<u32>>,
@@ -557,7 +559,7 @@ pub fn create_stop_decoder(
 }
 
 /// Parse tool calls from JSON schema constrained response
-pub fn parse_json_schema_response(
+pub(crate) fn parse_json_schema_response(
     processed_text: &str,
     tool_choice: &Option<ToolChoice>,
     model: &str,
@@ -646,7 +648,7 @@ pub fn parse_json_schema_response(
 /// # Returns
 /// * `Ok(Vec<GenerateComplete>)` - All complete responses collected from the stream
 /// * `Err(Response)` - Error response if the stream fails or returns an error
-pub async fn collect_stream_responses(
+pub(crate) async fn collect_stream_responses(
     stream: &mut ProtoStream,
     worker_name: &str,
 ) -> Result<Vec<ProtoGenerateComplete>, Response> {
@@ -691,7 +693,7 @@ pub async fn collect_stream_responses(
 
 /// Count the number of tool calls in the request message history
 /// This is used for KimiK2 format which needs globally unique indices
-pub fn get_history_tool_calls_count(request: &ChatCompletionRequest) -> usize {
+pub(crate) fn get_history_tool_calls_count(request: &ChatCompletionRequest) -> usize {
     request
         .messages
         .iter()
@@ -715,7 +717,7 @@ pub fn get_history_tool_calls_count(request: &ChatCompletionRequest) -> usize {
 ///
 /// # Returns
 /// A unique ID string. KimiK2 uses `functions.{name}:{global_index}`, others use `call_{uuid}`
-pub fn generate_tool_call_id(
+pub(crate) fn generate_tool_call_id(
     model: &str,
     tool_name: &str,
     tool_index: usize,
@@ -737,7 +739,7 @@ pub fn generate_tool_call_id(
 }
 
 /// Check if a reasoning parser is available for the given model
-pub fn check_reasoning_parser_availability(
+pub(crate) fn check_reasoning_parser_availability(
     reasoning_parser_factory: &ReasoningParserFactory,
     configured_parser: Option<&str>,
     model: &str,
@@ -752,7 +754,7 @@ pub fn check_reasoning_parser_availability(
 }
 
 /// Check if a tool parser is available for the given model
-pub fn check_tool_parser_availability(
+pub(crate) fn check_tool_parser_availability(
     tool_parser_factory: &ToolParserFactory,
     configured_parser: Option<&str>,
     model: &str,
@@ -769,7 +771,7 @@ pub fn check_tool_parser_availability(
 /// If a parser name is explicitly configured, use that parser.
 /// Otherwise, auto-detect based on the model name.
 /// Get a pooled reasoning parser (for non-streaming where state doesn't matter)
-pub fn get_reasoning_parser(
+pub(crate) fn get_reasoning_parser(
     reasoning_parser_factory: &ReasoningParserFactory,
     configured_parser: Option<&str>,
     model: &str,
@@ -793,7 +795,7 @@ pub fn get_reasoning_parser(
 }
 
 /// Create a fresh reasoning parser instance (for streaming where state isolation is needed)
-pub fn create_reasoning_parser(
+pub(crate) fn create_reasoning_parser(
     reasoning_parser_factory: &ReasoningParserFactory,
     configured_parser: Option<&str>,
     model: &str,
@@ -821,7 +823,7 @@ pub fn create_reasoning_parser(
 /// If a parser name is explicitly configured, use that parser.
 /// Otherwise, auto-detect based on the model name.
 /// Get a pooled tool parser (for non-streaming where state doesn't matter)
-pub fn get_tool_parser(
+pub(crate) fn get_tool_parser(
     tool_parser_factory: &ToolParserFactory,
     configured_parser: Option<&str>,
     model: &str,
@@ -845,7 +847,7 @@ pub fn get_tool_parser(
 }
 
 /// Create a fresh tool parser instance (for streaming where state isolation is needed)
-pub fn create_tool_parser(
+pub(crate) fn create_tool_parser(
     tool_parser_factory: &ToolParserFactory,
     configured_parser: Option<&str>,
     model: &str,
@@ -872,7 +874,7 @@ pub fn create_tool_parser(
 ///
 /// This function decodes token IDs using the tokenizer and builds the logprobs structure
 /// expected by the OpenAI API format.
-pub fn convert_proto_to_openai_logprobs(
+pub(crate) fn convert_proto_to_openai_logprobs(
     proto_logprobs: &OutputLogProbs,
     tokenizer: &Arc<dyn Tokenizer>,
 ) -> Result<ChatLogProbs, String> {
@@ -949,7 +951,9 @@ pub fn convert_proto_to_openai_logprobs(
 ///
 /// Generate format: [[logprob, token_id, ...], [logprob, token_id, ...], ...]
 /// Each inner vec contains [logprob (f64), token_id (i32), ...]
-pub fn convert_generate_output_logprobs(proto_logprobs: &OutputLogProbs) -> Vec<Vec<Option<f64>>> {
+pub(crate) fn convert_generate_output_logprobs(
+    proto_logprobs: &OutputLogProbs,
+) -> Vec<Vec<Option<f64>>> {
     proto_logprobs
         .token_logprobs
         .iter()
@@ -962,7 +966,9 @@ pub fn convert_generate_output_logprobs(proto_logprobs: &OutputLogProbs) -> Vec<
 ///
 /// Generate format: [[logprob, token_id, ...], [logprob, token_id, ...], ...]
 /// First token has null logprob: [[null, token_id], [logprob, token_id], ...]
-pub fn convert_generate_input_logprobs(proto_logprobs: &InputLogProbs) -> Vec<Vec<Option<f64>>> {
+pub(crate) fn convert_generate_input_logprobs(
+    proto_logprobs: &InputLogProbs,
+) -> Vec<Vec<Option<f64>>> {
     proto_logprobs
         .token_logprobs
         .iter()
@@ -985,7 +991,10 @@ pub fn convert_generate_input_logprobs(proto_logprobs: &InputLogProbs) -> Vec<Ve
 /// - Any other JSON -> Other(...)
 ///
 /// For backward compatibility, also handles simple string "stop" -> Stop
-pub fn parse_finish_reason(reason_str: &str, completion_tokens: i32) -> GenerateFinishReason {
+pub(crate) fn parse_finish_reason(
+    reason_str: &str,
+    completion_tokens: i32,
+) -> GenerateFinishReason {
     if reason_str == "stop" {
         return GenerateFinishReason::Stop;
     }
@@ -1010,7 +1019,7 @@ pub fn parse_finish_reason(reason_str: &str, completion_tokens: i32) -> Generate
 // ============================================================================
 
 /// Map route path to endpoint label for metrics
-pub fn route_to_endpoint(route: &str) -> &'static str {
+pub(crate) fn route_to_endpoint(route: &str) -> &'static str {
     match route {
         "/v1/chat/completions" => metrics_labels::ENDPOINT_CHAT,
         "/generate" => metrics_labels::ENDPOINT_GENERATE,
@@ -1022,7 +1031,7 @@ pub fn route_to_endpoint(route: &str) -> &'static str {
 }
 
 /// Map HTTP status code to error type label for metrics
-pub fn error_type_from_status(status: StatusCode) -> &'static str {
+pub(crate) fn error_type_from_status(status: StatusCode) -> &'static str {
     match status.as_u16() {
         400 => metrics_labels::ERROR_VALIDATION,
         404 => metrics_labels::ERROR_NO_WORKERS,