[model-gateway] bug fix on module name (#16332)

sgl-model-gateway/Cargo.toml

@@ -14,7 +14,7 @@ vendored-openssl = ["openssl/vendored"]
 unused_qualifications = "warn"
 
 [lib]
-name = "sgl_model_gateway"
+name = "smg"
 crate-type = ["rlib"]
 
 [[bin]]

sgl-model-gateway/benches/manual_policy_benchmark.rs

@@ -1,7 +1,7 @@
 use std::{sync::Arc, thread};
 
 use criterion::{black_box, criterion_group, criterion_main, BenchmarkId, Criterion, Throughput};
-use sgl_model_gateway::{
+use smg::{
     core::{BasicWorkerBuilder, Worker, WorkerType},
     policies::{LoadBalancingPolicy, ManualPolicy, SelectWorkerInfo},
 };

sgl-model-gateway/benches/request_processing.rs

@@ -2,7 +2,7 @@ use std::time::Instant;
 
 use criterion::{black_box, criterion_group, criterion_main, BenchmarkId, Criterion, Throughput};
 use serde_json::{from_str, to_string, to_value, to_vec};
-use sgl_model_gateway::{
+use smg::{
     core::{BasicWorker, BasicWorkerBuilder, Worker, WorkerType},
     protocols::{
         chat::{ChatCompletionRequest, ChatMessage, MessageContent},

sgl-model-gateway/benches/router_registry_bench.rs

@@ -1,9 +1,7 @@
 use std::{collections::HashMap, sync::Arc};
 
 use criterion::{black_box, criterion_group, criterion_main, BenchmarkId, Criterion};
-use sgl_model_gateway::core::{
-    BasicWorkerBuilder, CircuitBreakerConfig, WorkerRegistry, WorkerType,
-};
+use smg::core::{BasicWorkerBuilder, CircuitBreakerConfig, WorkerRegistry, WorkerType};
 
 // Helper to populate registry
 fn setup_registry(count: usize) -> Arc<WorkerRegistry> {

sgl-model-gateway/benches/tokenizer_benchmark.rs

@@ -13,7 +13,7 @@ use std::{
 };
 
 use criterion::{black_box, criterion_group, BenchmarkId, Criterion, Throughput};
-use sgl_model_gateway::tokenizer::{
+use smg::tokenizer::{
     cache::{CacheConfig, CachedTokenizer},
     huggingface::HuggingFaceTokenizer,
     sequence::Sequence,
@@ -31,11 +31,9 @@ fn get_tokenizer_path() -> &'static PathBuf {
         // with special: true, normalized: false - perfect for demonstrating L1 cache
         let rt = tokio::runtime::Runtime::new().expect("Failed to create tokio runtime");
         let tokenizer_dir = rt.block_on(async {
-            sgl_model_gateway::tokenizer::hub::download_tokenizer_from_hf(
-                "Qwen/Qwen3-4B-Instruct-2507",
-            )
-            .await
-            .expect("Failed to download Qwen3-4B-Instruct tokenizer from HuggingFace")
+            smg::tokenizer::hub::download_tokenizer_from_hf("Qwen/Qwen3-4B-Instruct-2507")
+                .await
+                .expect("Failed to download Qwen3-4B-Instruct tokenizer from HuggingFace")
         });
 
         // The download_tokenizer_from_hf returns the directory containing tokenizer.json

sgl-model-gateway/benches/tool_parser_benchmark.rs

@@ -19,7 +19,7 @@ use std::{
 
 use criterion::{black_box, criterion_group, BenchmarkId, Criterion, Throughput};
 use serde_json::json;
-use sgl_model_gateway::{
+use smg::{
     protocols::common::{Function, Tool},
     tool_parser::{JsonParser, ParserFactory as ToolParserFactory, ToolParser},
 };

sgl-model-gateway/benches/tree_benchmark.rs

@@ -25,7 +25,7 @@ use rand::{
     rng as thread_rng, Rng,
 };
 // Import the tree module
-use sgl_model_gateway::policies::tree::Tree;
+use smg::policies::tree::Tree;
 
 // Global results storage for summary
 lazy_static::lazy_static! {

sgl-model-gateway/bindings/golang/src/client.rs

@@ -8,11 +8,11 @@ use tokio::runtime::Runtime;
 use once_cell::sync::Lazy;
 use uuid::Uuid;
 
-use sgl_model_gateway::tokenizer::create_tokenizer_from_file;
-use sgl_model_gateway::tokenizer::traits::Tokenizer;
-use sgl_model_gateway::grpc_client::sglang_scheduler::SglangSchedulerClient;
-use sgl_model_gateway::protocols::chat::ChatCompletionRequest;
-use sgl_model_gateway::routers::grpc::utils::{process_chat_messages, generate_tool_constraints};
+use smg::tokenizer::create_tokenizer_from_file;
+use smg::tokenizer::traits::Tokenizer;
+use smg::grpc_client::sglang_scheduler::SglangSchedulerClient;
+use smg::protocols::chat::ChatCompletionRequest;
+use smg::routers::grpc::utils::{process_chat_messages, generate_tool_constraints};
 
 use super::error::{SglErrorCode, set_error_message};
 use super::grpc_converter::sgl_grpc_response_converter_create;

sgl-model-gateway/bindings/golang/src/grpc_converter.rs

@@ -9,12 +9,12 @@ use serde_json::Value;
 use tokio::runtime::Runtime;
 use once_cell::sync::Lazy;
 
-use sgl_model_gateway::tokenizer::traits::Tokenizer;
-use sgl_model_gateway::tokenizer::stream::DecodeStream;
-use sgl_model_gateway::tool_parser::ToolParser;
-use sgl_model_gateway::protocols::common::{Tool, ToolChoice, ToolChoiceValue, ToolCallDelta, FunctionCallDelta, Usage, StringOrArray};
-use sgl_model_gateway::tokenizer::stop::StopSequenceDecoder;
-use sgl_model_gateway::grpc_client::sglang_proto as proto;
+use smg::tokenizer::traits::Tokenizer;
+use smg::tokenizer::stream::DecodeStream;
+use smg::tool_parser::ToolParser;
+use smg::protocols::common::{Tool, ToolChoice, ToolChoiceValue, ToolCallDelta, FunctionCallDelta, Usage, StringOrArray};
+use smg::tokenizer::stop::StopSequenceDecoder;
+use smg::grpc_client::sglang_proto as proto;
 
 use super::error::{SglErrorCode, set_error_message, clear_error_message};
 use super::tokenizer::TokenizerHandle;
@@ -22,9 +22,9 @@ use super::utils::generate_tool_call_id;
 
 /// Global parser factory (initialized once)
 // Use the re-exported ParserFactory from tool_parser module
-static PARSER_FACTORY: Lazy<sgl_model_gateway::tool_parser::ParserFactory> = Lazy::new(|| {
+static PARSER_FACTORY: Lazy<smg::tool_parser::ParserFactory> = Lazy::new(|| {
     // ParserFactory is re-exported from tool_parser::factory, so we can use it directly
-    sgl_model_gateway::tool_parser::ParserFactory::default()
+    smg::tool_parser::ParserFactory::default()
 });
 
 /// Global tokio runtime for async operations
@@ -151,7 +151,7 @@ pub unsafe extern "C" fn sgl_grpc_response_converter_create(
     // Create stop decoder if needed
     let stop_decoder = if stop.is_some() || stop_token_ids.is_some() {
         Some(Arc::new(tokio::sync::Mutex::new(
-            sgl_model_gateway::routers::grpc::utils::create_stop_decoder(
+            smg::routers::grpc::utils::create_stop_decoder(
                 &tokenizer,
                 stop.as_ref(),
                 stop_token_ids.as_ref(),
@@ -389,9 +389,9 @@ pub(crate) async fn convert_proto_chunk_to_openai(
     request_id: &str,
     created: u64,
     system_fingerprint: Option<&str>,
-) -> Result<Option<sgl_model_gateway::protocols::chat::ChatCompletionStreamResponse>, String> {
-    use sgl_model_gateway::grpc_client::sglang_proto::generate_response::Response::*;
-    use sgl_model_gateway::protocols::chat::{ChatCompletionStreamResponse, ChatMessageDelta, ChatStreamChoice};
+) -> Result<Option<smg::protocols::chat::ChatCompletionStreamResponse>, String> {
+    use smg::grpc_client::sglang_proto::generate_response::Response::*;
+    use smg::protocols::chat::{ChatCompletionStreamResponse, ChatMessageDelta, ChatStreamChoice};
 
     match proto_response.response {
         Some(Chunk(chunk)) => {
@@ -427,19 +427,19 @@ pub(crate) async fn convert_proto_chunk_to_openai(
                 let mut text = String::new();
                 for &token_id in &chunk.token_ids {
                     match decoder_guard.process_token(token_id).unwrap_or_else(|_| {
-                        sgl_model_gateway::tokenizer::stop::SequenceDecoderOutput::Held
+                        smg::tokenizer::stop::SequenceDecoderOutput::Held
                     }) {
-                        sgl_model_gateway::tokenizer::stop::SequenceDecoderOutput::Text(t) => {
+                        smg::tokenizer::stop::SequenceDecoderOutput::Text(t) => {
                             text.push_str(&t);
                         }
-                        sgl_model_gateway::tokenizer::stop::SequenceDecoderOutput::StoppedWithText(t) => {
+                        smg::tokenizer::stop::SequenceDecoderOutput::StoppedWithText(t) => {
                             text.push_str(&t);
                             break;
                         }
-                        sgl_model_gateway::tokenizer::stop::SequenceDecoderOutput::Stopped => {
+                        smg::tokenizer::stop::SequenceDecoderOutput::Stopped => {
                             break;
                         }
-                        sgl_model_gateway::tokenizer::stop::SequenceDecoderOutput::Held => {}
+                        smg::tokenizer::stop::SequenceDecoderOutput::Held => {}
                     }
                 }
                 text

sgl-model-gateway/bindings/golang/src/postprocessor.rs

@@ -14,7 +14,7 @@ use std::ptr;
 use std::sync::Arc;
 use serde_json::Value;
 
-use sgl_model_gateway::grpc_client::sglang_proto as proto;
+use smg::grpc_client::sglang_proto as proto;
 
 use super::error::{SglErrorCode, set_error_message};
 use super::grpc_converter::GrpcResponseConverterHandle;

sgl-model-gateway/bindings/golang/src/preprocessor.rs

@@ -12,9 +12,9 @@ use std::os::raw::{c_char, c_int};
 use std::ptr;
 use std::os::raw::c_uint;
 
-use sgl_model_gateway::tokenizer::create_tokenizer_from_file;
-use sgl_model_gateway::protocols::chat::ChatCompletionRequest;
-use sgl_model_gateway::routers::grpc::utils::{process_chat_messages, generate_tool_constraints};
+use smg::tokenizer::create_tokenizer_from_file;
+use smg::protocols::chat::ChatCompletionRequest;
+use smg::routers::grpc::utils::{process_chat_messages, generate_tool_constraints};
 
 use super::error::{SglErrorCode, set_error_message};
 use super::memory::{sgl_free_string, sgl_free_token_ids};

sgl-model-gateway/bindings/golang/src/stream.rs

@@ -23,7 +23,7 @@ use tokio::runtime::Runtime;
 use once_cell::sync::Lazy;
 use futures_util::StreamExt;
 
-use sgl_model_gateway::grpc_client::{sglang_proto as proto, sglang_scheduler::{SglangSchedulerClient, AbortOnDropStream}};
+use smg::grpc_client::{sglang_proto as proto, sglang_scheduler::{SglangSchedulerClient, AbortOnDropStream}};
 
 use super::error::{SglErrorCode, set_error_message};
 use super::grpc_converter::{GrpcResponseConverterHandle, convert_proto_chunk_to_openai};

sgl-model-gateway/bindings/golang/src/tokenizer.rs

@@ -6,7 +6,7 @@ use std::ptr;
 use std::sync::Arc;
 use serde_json::Value;
 
-use sgl_model_gateway::tokenizer::{
+use smg::tokenizer::{
     create_tokenizer_from_file,
     traits::Tokenizer as TokenizerTrait,
     chat_template::ChatTemplateParams,

sgl-model-gateway/bindings/golang/src/tool_parser.rs

@@ -9,8 +9,8 @@ use serde_json::{json, Value};
 use tokio::runtime::Runtime;
 use once_cell::sync::Lazy;
 
-use sgl_model_gateway::tool_parser::{ParserFactory, ToolParser};
-use sgl_model_gateway::protocols::common::Tool;
+use smg::tool_parser::{ParserFactory, ToolParser};
+use smg::protocols::common::Tool;
 
 use super::error::{SglErrorCode, set_error_message, clear_error_message};
 use super::utils::generate_tool_call_id;

sgl-model-gateway/bindings/python/src/lib.rs

@@ -1,5 +1,5 @@
 use pyo3::prelude::*;
-use sgl_model_gateway::*;
+use smg::*;
 use once_cell::sync::OnceCell;
 use std::collections::HashMap;
 

sgl-model-gateway/src/core/model_card.rs

@@ -95,7 +95,7 @@ impl std::fmt::Display for ProviderType {
 /// # Example
 ///
 /// ```
-/// use sgl_model_gateway::core::{ModelCard, ModelType, ProviderType};
+/// use smg::core::{ModelCard, ModelType, ProviderType};
 ///
 /// let card = ModelCard::new("meta-llama/Llama-3.1-8B-Instruct")
 ///     .with_display_name("Llama 3.1 8B Instruct")

sgl-model-gateway/src/main.rs

@@ -1,7 +1,7 @@
 use std::collections::HashMap;
 
 use clap::{ArgAction, Parser, Subcommand, ValueEnum};
-use sgl_model_gateway::{
+use smg::{
     auth::{ApiKeyEntry, ControlPlaneAuthConfig, JwtConfig, Role},
     config::{
         CircuitBreakerConfig, ConfigError, ConfigResult, DiscoveryConfig, HealthCheckConfig,

sgl-model-gateway/src/multimodal/vision/mod.rs

@@ -15,7 +15,7 @@
 //! # Usage
 //!
 //! ```rust,ignore
-//! use sgl_model_gateway::multimodal::vision::{
+//! use smg::multimodal::vision::{
 //!     PreProcessorConfig,
 //!     processors::LlavaProcessor,
 //!     ImagePreProcessor,

sgl-model-gateway/src/observability/otel_trace.rs

@@ -116,10 +116,8 @@ pub fn otel_tracing_init(enable: bool, otlp_endpoint: Option<&str>) -> Result<()
         .with_batch_config(batch_config)
         .build();
 
-    let resource = Resource::default().merge(&Resource::new(vec![KeyValue::new(
-        "service.name",
-        "smg",
-    )]));
+    let resource =
+        Resource::default().merge(&Resource::new(vec![KeyValue::new("service.name", "smg")]));
 
     let provider = TracerProvider::builder()
         .with_span_processor(span_processor)

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

@@ -18,7 +18,7 @@
 //! ## Usage
 //!
 //! ```ignore
-//! use sgl_model_gateway::routers::grpc::harmony::{HarmonyDetector, HarmonyBuilder};
+//! use smg::routers::grpc::harmony::{HarmonyDetector, HarmonyBuilder};
 //!
 //! // Detect if model supports Harmony
 //! if HarmonyDetector::is_harmony_model("gpt-4o") {

sgl-model-gateway/src/tokenizer/README.md

@@ -138,7 +138,7 @@ as of `sgl-model-gateway/src/tokenizer/*`.
 ## Usage Examples
 ```rust
 use std::sync::Arc;
-use sgl_model_gateway::tokenizer::{
+use smg::tokenizer::{
     create_tokenizer, SequenceDecoderOutput, StopSequenceDecoderBuilder, Tokenizer,
 };
 
@@ -172,7 +172,7 @@ for &token in encoding.token_ids() {
 
 ```rust
 // Apply a chat template when one is bundled with the tokenizer
-use sgl_model_gateway::tokenizer::{chat_template::ChatTemplateParams, HuggingFaceTokenizer};
+use smg::tokenizer::{chat_template::ChatTemplateParams, HuggingFaceTokenizer};
 
 let mut hf = HuggingFaceTokenizer::from_file_with_chat_template(
     "./tokenizer.json",

sgl-model-gateway/tests/api_endpoints_test.rs

@@ -10,7 +10,7 @@ use axum::{
 use common::mock_worker::{HealthStatus, MockWorker, MockWorkerConfig, WorkerType};
 use reqwest::Client;
 use serde_json::json;
-use sgl_model_gateway::{
+use smg::{
     app_context::AppContext,
     config::{RouterConfig, RoutingMode},
     core::Job,
@@ -1028,7 +1028,7 @@ mod responses_endpoint_tests {
         let app = ctx.create_app().await;
 
         // Directly store a response in the storage to test the retrieval endpoint
-        use sgl_model_gateway::data_connector::{ResponseId, StoredResponse};
+        use smg::data_connector::{ResponseId, StoredResponse};
         let mut stored_response = StoredResponse::new(None);
         stored_response.id = ResponseId::from("resp_test_input_items");
         stored_response.input = json!([

sgl-model-gateway/tests/auth_integration_test.rs

@@ -18,9 +18,7 @@ use jsonwebtoken::{encode, EncodingKey, Header};
 use rsa::{traits::PublicKeyParts, RsaPrivateKey};
 use serde::{Deserialize, Serialize};
 use serde_json::json;
-use sgl_model_gateway::auth::{
-    ApiKeyEntry, ControlPlaneAuthConfig, ControlPlaneAuthState, JwtConfig, Role,
-};
+use smg::auth::{ApiKeyEntry, ControlPlaneAuthConfig, ControlPlaneAuthState, JwtConfig, Role};
 use tokio::net::TcpListener;
 
 const TEST_KEY_ID: &str = "test-key-1";
@@ -572,7 +570,7 @@ async fn test_audit_logging_disabled() {
 
 #[tokio::test]
 async fn test_jwt_jti_replay_protection() {
-    use sgl_model_gateway::auth::JwtValidator;
+    use smg::auth::JwtValidator;
 
     let (addr, _server) = start_mock_jwks_server().await;
 
@@ -612,7 +610,7 @@ async fn test_jwt_jti_replay_protection() {
 
 #[tokio::test]
 async fn test_jwt_different_tokens_no_replay() {
-    use sgl_model_gateway::auth::JwtValidator;
+    use smg::auth::JwtValidator;
 
     let (addr, _server) = start_mock_jwks_server().await;
 

sgl-model-gateway/tests/cache_aware_backward_compat_test.rs

@@ -1,6 +1,6 @@
 use std::{collections::HashMap, sync::Arc};
 
-use sgl_model_gateway::{
+use smg::{
     core::{BasicWorkerBuilder, Worker, WorkerType},
     policies::{CacheAwareConfig, CacheAwarePolicy, LoadBalancingPolicy, SelectWorkerInfo},
 };

sgl-model-gateway/tests/chat_template_format_detection.rs

@@ -1,4 +1,4 @@
-use sgl_model_gateway::{
+use smg::{
     protocols::chat::{ChatMessage, MessageContent},
     tokenizer::chat_template::{
         detect_chat_template_content_format, ChatTemplateContentFormat, ChatTemplateParams,

sgl-model-gateway/tests/chat_template_integration.rs

@@ -1,4 +1,4 @@
-use sgl_model_gateway::{
+use smg::{
     protocols::{
         chat::{ChatMessage, MessageContent},
         common::{ContentPart, ImageUrl},

sgl-model-gateway/tests/chat_template_loading.rs

@@ -2,7 +2,7 @@
 mod tests {
     use std::fs;
 
-    use sgl_model_gateway::{
+    use smg::{
         protocols::chat::{ChatMessage, MessageContent},
         tokenizer::{chat_template::ChatTemplateParams, huggingface::HuggingFaceTokenizer},
     };
@@ -78,7 +78,7 @@ mod tests {
             .map(|msg| serde_json::to_value(msg).unwrap())
             .collect();
 
-        use sgl_model_gateway::tokenizer::chat_template::ChatTemplateParams;
+        use smg::tokenizer::chat_template::ChatTemplateParams;
         let params = ChatTemplateParams {
             add_generation_prompt: true,
             ..Default::default()

sgl-model-gateway/tests/common/mod.rs

@@ -14,7 +14,7 @@ use std::{
 };
 
 use serde_json::json;
-use sgl_model_gateway::{
+use smg::{
     app_context::AppContext,
     config::{RouterConfig, RoutingMode},
     core::{
@@ -95,17 +95,14 @@ pub async fn create_test_context(config: RouterConfig) -> Arc<AppContext> {
 
     // Initialize JobQueue after AppContext is created
     let weak_context = Arc::downgrade(&app_context);
-    let job_queue = sgl_model_gateway::core::JobQueue::new(
-        sgl_model_gateway::core::JobQueueConfig::default(),
-        weak_context,
-    );
+    let job_queue = smg::core::JobQueue::new(smg::core::JobQueueConfig::default(), weak_context);
     app_context
         .worker_job_queue
         .set(job_queue)
         .expect("JobQueue should only be initialized once");
 
     // Initialize WorkflowEngine and register workflows
-    use sgl_model_gateway::{
+    use smg::{
         core::steps::{create_worker_registration_workflow, create_worker_removal_workflow},
         workflow::WorkflowEngine,
     };
@@ -142,7 +139,7 @@ pub async fn create_test_context(config: RouterConfig) -> Arc<AppContext> {
     }
 
     // Initialize MCP manager with empty config
-    use sgl_model_gateway::mcp::{McpConfig, McpManager};
+    use smg::mcp::{McpConfig, McpManager};
     let empty_config = McpConfig {
         servers: vec![],
         pool: Default::default(),
@@ -231,17 +228,14 @@ pub async fn create_test_context_with_parsers(config: RouterConfig) -> Arc<AppCo
 
     // Initialize JobQueue after AppContext is created
     let weak_context = Arc::downgrade(&app_context);
-    let job_queue = sgl_model_gateway::core::JobQueue::new(
-        sgl_model_gateway::core::JobQueueConfig::default(),
-        weak_context,
-    );
+    let job_queue = smg::core::JobQueue::new(smg::core::JobQueueConfig::default(), weak_context);
     app_context
         .worker_job_queue
         .set(job_queue)
         .expect("JobQueue should only be initialized once");
 
     // Initialize WorkflowEngine and register workflows
-    use sgl_model_gateway::{
+    use smg::{
         core::steps::{create_worker_registration_workflow, create_worker_removal_workflow},
         workflow::WorkflowEngine,
     };
@@ -278,7 +272,7 @@ pub async fn create_test_context_with_parsers(config: RouterConfig) -> Arc<AppCo
     }
 
     // Initialize MCP manager with empty config
-    use sgl_model_gateway::mcp::{McpConfig, McpManager};
+    use smg::mcp::{McpConfig, McpManager};
     let empty_config = McpConfig {
         servers: vec![],
         pool: Default::default(),
@@ -303,7 +297,7 @@ pub async fn create_test_context_with_mcp_config(
     config: RouterConfig,
     mcp_config_path: &str,
 ) -> Arc<AppContext> {
-    use sgl_model_gateway::mcp::{McpConfig, McpManager};
+    use smg::mcp::{McpConfig, McpManager};
 
     let client = reqwest::Client::new();
 
@@ -367,17 +361,14 @@ pub async fn create_test_context_with_mcp_config(
 
     // Initialize JobQueue after AppContext is created
     let weak_context = Arc::downgrade(&app_context);
-    let job_queue = sgl_model_gateway::core::JobQueue::new(
-        sgl_model_gateway::core::JobQueueConfig::default(),
-        weak_context,
-    );
+    let job_queue = smg::core::JobQueue::new(smg::core::JobQueueConfig::default(), weak_context);
     app_context
         .worker_job_queue
         .set(job_queue)
         .expect("JobQueue should only be initialized once");
 
     // Initialize WorkflowEngine and register workflows
-    use sgl_model_gateway::{
+    use smg::{
         core::steps::{create_worker_registration_workflow, create_worker_removal_workflow},
         workflow::WorkflowEngine,
     };

sgl-model-gateway/tests/common/test_app.rs

@@ -2,7 +2,7 @@ use std::sync::{Arc, OnceLock};
 
 use axum::Router;
 use reqwest::Client;
-use sgl_model_gateway::{
+use smg::{
     app_context::AppContext,
     config::RouterConfig,
     core::{

sgl-model-gateway/tests/load_guard_raii_test.rs

@@ -11,9 +11,7 @@ use axum::{body::Body, response::Response};
 use bytes::Bytes;
 use futures_util::StreamExt;
 use http_body_util::BodyExt;
-use sgl_model_gateway::core::{
-    attach_guards_to_response, BasicWorkerBuilder, Worker, WorkerLoadGuard,
-};
+use smg::core::{attach_guards_to_response, BasicWorkerBuilder, Worker, WorkerLoadGuard};
 use tokio::sync::mpsc;
 use tokio_stream::wrappers::UnboundedReceiverStream;
 

sgl-model-gateway/tests/mcp_test.rs

@@ -13,7 +13,7 @@ use std::collections::HashMap;
 
 use common::mock_mcp_server::MockMCPServer;
 use serde_json::json;
-use sgl_model_gateway::mcp::{McpConfig, McpError, McpManager, McpServerConfig, McpTransport};
+use smg::mcp::{McpConfig, McpError, McpManager, McpServerConfig, McpTransport};
 
 /// Create a new mock server for testing (each test gets its own)
 async fn create_mock_server() -> MockMCPServer {

sgl-model-gateway/tests/metrics_aggregator_test.rs

@@ -1,4 +1,4 @@
-use sgl_model_gateway::core::metrics_aggregator::{aggregate_metrics, MetricPack};
+use smg::core::metrics_aggregator::{aggregate_metrics, MetricPack};
 
 #[test]
 fn test_aggregate_simple() {

sgl-model-gateway/tests/multimodal_tracker_test.rs

@@ -2,7 +2,7 @@ use std::{collections::HashMap, path::PathBuf, sync::Arc, time::Duration};
 
 use base64::{engine::general_purpose::STANDARD as BASE64_STANDARD, Engine};
 use reqwest::Client;
-use sgl_model_gateway::multimodal::{
+use smg::multimodal::{
     AsyncMultiModalTracker, ChatContentPart, ConversationSegment, ImageFetchConfig, ImageSource,
     MediaConnector, MediaConnectorConfig, MediaSource, Modality, TrackerConfig,
 };

sgl-model-gateway/tests/otel_tracing_test.rs

@@ -17,7 +17,7 @@ use opentelemetry_proto::tonic::collector::trace::v1::{
 use portpicker::pick_unused_port;
 use serde_json::json;
 use serial_test::serial;
-use sgl_model_gateway::{
+use smg::{
     config::{RouterConfig, TraceConfig},
     core::Job,
     observability::{logging, otel_trace},
@@ -161,14 +161,14 @@ async fn test_router_with_tracing() {
             log_dir: None,
             colorize: false,
             log_file_name: "test-otel".to_string(),
-            log_targets: Some(vec!["sgl_model_gateway".to_string()]),
+            log_targets: Some(vec!["smg".to_string()]),
         },
         Some(trace_config),
     );
     println!("Logging initialized with OTEL layer");
 
     // 5. Create a span and sleep for a while
-    let _span = info_span!(target: "sgl_model_gateway::otel-trace", "test_router_with_tracing");
+    let _span = info_span!(target: "smg::otel-trace", "test_router_with_tracing");
     tokio::time::sleep(Duration::from_secs(1)).await;
     drop(_span);
 
@@ -315,7 +315,7 @@ async fn test_grpc_trace_context_injection() {
     // 4. Test within a span context
     tracing::subscriber::with_default(subscriber, || {
         // Create a span that will be exported to OTEL
-        let span = info_span!(target: "sgl_model_gateway::otel-trace", "test_grpc_span");
+        let span = info_span!(target: "smg::otel-trace", "test_grpc_span");
         let _guard = span.enter();
 
         // Create empty gRPC metadata

sgl-model-gateway/tests/parser_endpoints_test.rs

@@ -10,7 +10,7 @@ use axum::{
 use common::mock_worker::{MockWorker, MockWorkerConfig};
 use reqwest::Client;
 use serde_json::json;
-use sgl_model_gateway::{
+use smg::{
     app_context::AppContext,
     config::{RouterConfig, RoutingMode},
     routers::{RouterFactory, RouterTrait},

sgl-model-gateway/tests/policy_registry_integration.rs

@@ -2,7 +2,7 @@
 
 use std::{collections::HashMap, sync::Arc};
 
-use sgl_model_gateway::{
+use smg::{
     config::PolicyConfig, core::WorkerRegistry, policies::PolicyRegistry,
     protocols::worker_spec::WorkerConfigRequest, routers::router_manager::RouterManager,
 };
@@ -115,7 +115,7 @@ async fn test_policy_registry_with_router_manager() {
 
 #[test]
 fn test_policy_registry_cleanup() {
-    use sgl_model_gateway::{config::PolicyConfig, policies::PolicyRegistry};
+    use smg::{config::PolicyConfig, policies::PolicyRegistry};
 
     let registry = PolicyRegistry::new(PolicyConfig::RoundRobin);
 
@@ -142,7 +142,7 @@ fn test_policy_registry_cleanup() {
 
 #[test]
 fn test_policy_registry_multiple_models() {
-    use sgl_model_gateway::{config::PolicyConfig, policies::PolicyRegistry};
+    use smg::{config::PolicyConfig, policies::PolicyRegistry};
 
     let registry = PolicyRegistry::new(PolicyConfig::RoundRobin);
 

sgl-model-gateway/tests/request_formats_test.rs

@@ -5,7 +5,7 @@ use std::sync::Arc;
 use common::mock_worker::{HealthStatus, MockWorker, MockWorkerConfig, WorkerType};
 use reqwest::Client;
 use serde_json::json;
-use sgl_model_gateway::{
+use smg::{
     config::{RouterConfig, RoutingMode},
     routers::{RouterFactory, RouterTrait},
 };

sgl-model-gateway/tests/responses_api_test.rs

@@ -1,7 +1,7 @@
 // Integration test for Responses API
 
 use axum::http::StatusCode;
-use sgl_model_gateway::protocols::{
+use smg::protocols::{
     common::{GenerationRequest, ToolChoice, ToolChoiceValue, UsageInfo},
     responses::{
         ReasoningEffort, ResponseInput, ResponseReasoningParam, ResponseTool, ResponseToolType,
@@ -14,7 +14,7 @@ use common::{
     mock_mcp_server::MockMCPServer,
     mock_worker::{HealthStatus, MockWorker, MockWorkerConfig, WorkerType},
 };
-use sgl_model_gateway::{
+use smg::{
     config::RouterConfig,
     routers::{conversations, RouterFactory},
 };
@@ -397,7 +397,7 @@ fn test_usage_conversion() {
         completion_tokens: 25,
         total_tokens: 40,
         reasoning_tokens: Some(8),
-        prompt_tokens_details: Some(sgl_model_gateway::protocols::common::PromptTokenUsageInfo {
+        prompt_tokens_details: Some(smg::protocols::common::PromptTokenUsageInfo {
             cached_tokens: 3,
         }),
     };
@@ -785,7 +785,7 @@ async fn test_max_tool_calls_limit() {
 async fn setup_streaming_mcp_test() -> (
     MockMCPServer,
     MockWorker,
-    Box<dyn sgl_model_gateway::routers::RouterTrait>,
+    Box<dyn smg::routers::RouterTrait>,
     tempfile::TempDir,
 ) {
     let mcp = MockMCPServer::start().await.expect("start mcp");

sgl-model-gateway/tests/spec/chat_completion.rs

@@ -1,5 +1,5 @@
 use serde_json::json;
-use sgl_model_gateway::protocols::{
+use smg::protocols::{
     chat::{ChatCompletionRequest, ChatMessage, MessageContent},
     common::{
         Function, FunctionCall, FunctionChoice, StreamOptions, Tool, ToolChoice, ToolChoiceValue,

sgl-model-gateway/tests/spec/chat_message.rs

@@ -1,5 +1,5 @@
 use serde_json::json;
-use sgl_model_gateway::protocols::chat::{ChatMessage, MessageContent};
+use smg::protocols::chat::{ChatMessage, MessageContent};
 
 #[test]
 fn test_chat_message_tagged_by_role_system() {

sgl-model-gateway/tests/spec/embedding.rs

@@ -1,5 +1,5 @@
 use serde_json::{from_str, json, to_string};
-use sgl_model_gateway::protocols::{common::GenerationRequest, embedding::EmbeddingRequest};
+use smg::protocols::{common::GenerationRequest, embedding::EmbeddingRequest};
 
 #[test]
 fn test_embedding_request_serialization_string_input() {

sgl-model-gateway/tests/spec/rerank.rs

@@ -1,7 +1,7 @@
 use std::collections::HashMap;
 
 use serde_json::{from_str, to_string, Number, Value};
-use sgl_model_gateway::protocols::{
+use smg::protocols::{
     common::{GenerationRequest, StringOrArray, UsageInfo},
     rerank::{RerankRequest, RerankResponse, RerankResult, V1RerankReqInput},
 };

sgl-model-gateway/tests/spec/responses.rs

@@ -1,5 +1,5 @@
 use serde_json::json;
-use sgl_model_gateway::protocols::{
+use smg::protocols::{
     common::{Function, StringOrArray, ToolChoice, ToolChoiceValue},
     responses::{
         IncludeField, ResponseInput, ResponseInputOutputItem, ResponseTool, ResponseToolType,
@@ -963,7 +963,7 @@ fn test_validate_input_items_structure() {
 /// Test tool_choice defaults to auto when tools are present
 #[test]
 fn test_normalize_tool_choice_auto() {
-    use sgl_model_gateway::protocols::validated::Normalizable;
+    use smg::protocols::validated::Normalizable;
 
     let mut request = ResponsesRequest {
         input: ResponseInput::Text("test".to_string()),
@@ -1004,7 +1004,7 @@ fn test_normalize_tool_choice_auto() {
 /// Test tool_choice defaults to none when tools array is empty
 #[test]
 fn test_normalize_tool_choice_none() {
-    use sgl_model_gateway::protocols::validated::Normalizable;
+    use smg::protocols::validated::Normalizable;
 
     let mut request = ResponsesRequest {
         input: ResponseInput::Text("test".to_string()),
@@ -1031,7 +1031,7 @@ fn test_normalize_tool_choice_none() {
 /// Test tool_choice is not overridden if already set
 #[test]
 fn test_normalize_tool_choice_no_override() {
-    use sgl_model_gateway::protocols::validated::Normalizable;
+    use smg::protocols::validated::Normalizable;
 
     let mut request = ResponsesRequest {
         input: ResponseInput::Text("test".to_string()),
@@ -1068,7 +1068,7 @@ fn test_normalize_tool_choice_no_override() {
 /// Test parallel_tool_calls defaults to true when tools are present
 #[test]
 fn test_normalize_parallel_tool_calls() {
-    use sgl_model_gateway::protocols::validated::Normalizable;
+    use smg::protocols::validated::Normalizable;
 
     let mut request = ResponsesRequest {
         input: ResponseInput::Text("test".to_string()),
@@ -1107,7 +1107,7 @@ fn test_normalize_parallel_tool_calls() {
 /// Test parallel_tool_calls is not set when tools are absent
 #[test]
 fn test_normalize_parallel_tool_calls_no_tools() {
-    use sgl_model_gateway::protocols::validated::Normalizable;
+    use smg::protocols::validated::Normalizable;
 
     let mut request = ResponsesRequest {
         input: ResponseInput::Text("test".to_string()),
@@ -1127,7 +1127,7 @@ fn test_normalize_parallel_tool_calls_no_tools() {
 /// Test parallel_tool_calls is not overridden if already set
 #[test]
 fn test_normalize_parallel_tool_calls_no_override() {
-    use sgl_model_gateway::protocols::validated::Normalizable;
+    use smg::protocols::validated::Normalizable;
 
     let mut request = ResponsesRequest {
         input: ResponseInput::Text("test".to_string()),
@@ -1162,7 +1162,7 @@ fn test_normalize_parallel_tool_calls_no_override() {
 /// Test store defaults to true
 #[test]
 fn test_normalize_store_default() {
-    use sgl_model_gateway::protocols::validated::Normalizable;
+    use smg::protocols::validated::Normalizable;
 
     let mut request = ResponsesRequest {
         input: ResponseInput::Text("test".to_string()),
@@ -1182,7 +1182,7 @@ fn test_normalize_store_default() {
 /// Test store is not overridden if already set
 #[test]
 fn test_normalize_store_no_override() {
-    use sgl_model_gateway::protocols::validated::Normalizable;
+    use smg::protocols::validated::Normalizable;
 
     let mut request = ResponsesRequest {
         input: ResponseInput::Text("test".to_string()),

sgl-model-gateway/tests/streaming_tests.rs

@@ -6,7 +6,7 @@ use common::mock_worker::{HealthStatus, MockWorker, MockWorkerConfig, WorkerType
 use futures_util::StreamExt;
 use reqwest::Client;
 use serde_json::json;
-use sgl_model_gateway::{
+use smg::{
     config::{RouterConfig, RoutingMode},
     routers::{RouterFactory, RouterTrait},
 };

sgl-model-gateway/tests/test_openai_routing.rs

@@ -17,7 +17,7 @@ use axum::{
     Json, Router,
 };
 use serde_json::json;
-use sgl_model_gateway::{
+use smg::{
     config::{
         ConfigError, ConfigValidator, HistoryBackend, OracleConfig, RouterConfig, RoutingMode,
     },
@@ -579,14 +579,11 @@ async fn test_router_factory_openai_mode() {
         worker_urls: vec!["https://api.openai.com".to_string()],
     };
 
-    let router_config = RouterConfig::new(
-        routing_mode,
-        sgl_model_gateway::config::PolicyConfig::Random,
-    );
+    let router_config = RouterConfig::new(routing_mode, smg::config::PolicyConfig::Random);
 
     let app_context = common::create_test_context(router_config).await;
 
-    let router = sgl_model_gateway::routers::RouterFactory::create_router(&app_context).await;
+    let router = smg::routers::RouterFactory::create_router(&app_context).await;
     assert!(
         router.is_ok(),
         "Router factory should create OpenAI router successfully"

sgl-model-gateway/tests/test_pd_routing.rs

@@ -1,7 +1,7 @@
 #[cfg(test)]
 mod test_pd_routing {
     use serde_json::json;
-    use sgl_model_gateway::{
+    use smg::{
         app_context::AppContext,
         config::{PolicyConfig, RouterConfig, RoutingMode},
         core::{BasicWorkerBuilder, Worker, WorkerType},
@@ -39,7 +39,7 @@ mod test_pd_routing {
 
     #[test]
     fn test_worker_types() {
-        use sgl_model_gateway::core::{BasicWorkerBuilder, Worker, WorkerType};
+        use smg::core::{BasicWorkerBuilder, Worker, WorkerType};
 
         let prefill_worker: Box<dyn Worker> = Box::new(
             BasicWorkerBuilder::new("http://prefill:8080")
@@ -215,7 +215,7 @@ mod test_pd_routing {
             let app_context = {
                 use std::sync::{Arc, OnceLock};
 
-                use sgl_model_gateway::{
+                use smg::{
                     core::{LoadMonitor, WorkerRegistry},
                     data_connector::{
                         MemoryConversationItemStorage, MemoryConversationStorage,
@@ -674,7 +674,7 @@ mod test_pd_routing {
 
     #[test]
     fn test_bootstrap_injection_with_benchmark_requests() {
-        use sgl_model_gateway::core::{BasicWorkerBuilder, Worker, WorkerType};
+        use smg::core::{BasicWorkerBuilder, Worker, WorkerType};
 
         let mut benchmark_request = json!({
             "input_ids": vec![vec![1, 2, 3, 4]; 16], // Batch size 16

sgl-model-gateway/tests/tokenizer_cache_correctness_test.rs

@@ -9,7 +9,7 @@ use std::{
     sync::{Arc, OnceLock},
 };
 
-use sgl_model_gateway::tokenizer::{
+use smg::tokenizer::{
     cache::{CacheConfig, CachedTokenizer},
     hub::download_tokenizer_from_hf,
     huggingface::HuggingFaceTokenizer,

sgl-model-gateway/tests/tokenizer_integration.rs

@@ -7,7 +7,7 @@ mod common;
 use std::sync::Arc;
 
 use common::{ensure_tokenizer_cached, EXPECTED_HASHES, TEST_PROMPTS};
-use sgl_model_gateway::tokenizer::{
+use smg::tokenizer::{
     factory, huggingface::HuggingFaceTokenizer, sequence::Sequence, stop::*, stream::DecodeStream,
     traits::*,
 };
@@ -279,7 +279,7 @@ fn test_batch_encoding() {
 
 #[test]
 fn test_special_tokens() {
-    use sgl_model_gateway::tokenizer::traits::Tokenizer as TokenizerTrait;
+    use smg::tokenizer::traits::Tokenizer as TokenizerTrait;
 
     let tokenizer_path = ensure_tokenizer_cached();
     let tokenizer = HuggingFaceTokenizer::from_file(tokenizer_path.to_str().unwrap())
@@ -408,7 +408,7 @@ fn test_load_chat_template_from_local_file() {
 
 #[tokio::test]
 async fn test_tinyllama_embedded_template() {
-    use sgl_model_gateway::tokenizer::hub::download_tokenizer_from_hf;
+    use smg::tokenizer::hub::download_tokenizer_from_hf;
 
     // Skip in CI without HF_TOKEN
 
@@ -444,7 +444,7 @@ async fn test_tinyllama_embedded_template() {
 
 #[tokio::test]
 async fn test_qwen3_next_embedded_template() {
-    use sgl_model_gateway::tokenizer::hub::download_tokenizer_from_hf;
+    use smg::tokenizer::hub::download_tokenizer_from_hf;
 
     // Test 3: Qwen3-Next has chat template in tokenizer_config.json
     match download_tokenizer_from_hf("Qwen/Qwen3-Next-80B-A3B-Instruct").await {
@@ -476,7 +476,7 @@ async fn test_qwen3_next_embedded_template() {
 
 #[tokio::test]
 async fn test_qwen3_vl_json_template_priority() {
-    use sgl_model_gateway::tokenizer::hub::download_tokenizer_from_hf;
+    use smg::tokenizer::hub::download_tokenizer_from_hf;
 
     // Test 4: Qwen3-VL has both tokenizer_config.json template and chat_template.json
     // Should prioritize chat_template.json
@@ -518,7 +518,7 @@ async fn test_qwen3_vl_json_template_priority() {
 
 #[tokio::test]
 async fn test_llava_separate_jinja_template() {
-    use sgl_model_gateway::tokenizer::hub::download_tokenizer_from_hf;
+    use smg::tokenizer::hub::download_tokenizer_from_hf;
 
     // Test 5: llava has chat_template.jinja as a separate file, not in tokenizer_config.json
     match download_tokenizer_from_hf("llava-hf/llava-1.5-7b-hf").await {

sgl-model-gateway/tests/tool_parser_deepseek.rs

@@ -1,6 +1,6 @@
 //! DeepSeek V3 Parser Integration Tests
 
-use sgl_model_gateway::tool_parser::{DeepSeekParser, ToolParser};
+use smg::tool_parser::{DeepSeekParser, ToolParser};
 
 mod common;
 use common::create_test_tools;

sgl-model-gateway/tests/tool_parser_edge_cases.rs

@@ -2,9 +2,7 @@
 //!
 //! Tests for malformed input, edge cases, and error recovery
 
-use sgl_model_gateway::tool_parser::{
-    JsonParser, MistralParser, PythonicParser, QwenParser, ToolParser,
-};
+use smg::tool_parser::{JsonParser, MistralParser, PythonicParser, QwenParser, ToolParser};
 
 mod common;
 use common::create_test_tools;

sgl-model-gateway/tests/tool_parser_fallback.rs

@@ -3,7 +3,7 @@
 //! When tool call parsing fails, the original text should be preserved as normal text
 //! rather than being lost. This ensures graceful degradation.
 
-use sgl_model_gateway::tool_parser::{
+use smg::tool_parser::{
     DeepSeekParser, JsonParser, LlamaParser, MistralParser, QwenParser, ToolParser,
 };
 

sgl-model-gateway/tests/tool_parser_glm47_moe.rs

@@ -1,6 +1,6 @@
 //! GLM-4.7 MoE Parser Integration Tests
 
-use sgl_model_gateway::tool_parser::{Glm4MoeParser, ToolParser};
+use smg::tool_parser::{Glm4MoeParser, ToolParser};
 
 mod common;
 use common::create_test_tools;

sgl-model-gateway/tests/tool_parser_glm4_moe.rs

@@ -1,6 +1,6 @@
 //! GLM-4 MoE Parser Integration Tests
 
-use sgl_model_gateway::tool_parser::{Glm4MoeParser, ToolParser};
+use smg::tool_parser::{Glm4MoeParser, ToolParser};
 
 mod common;
 use common::create_test_tools;

sgl-model-gateway/tests/tool_parser_json.rs

@@ -3,7 +3,7 @@
 //! Tests for the JSON parser which handles OpenAI, Claude, and generic JSON formats
 
 use serde_json::json;
-use sgl_model_gateway::tool_parser::{JsonParser, ToolParser};
+use smg::tool_parser::{JsonParser, ToolParser};
 
 mod common;
 use common::{create_test_tools, streaming_helpers::*};
@@ -166,7 +166,7 @@ async fn test_json_format_detection() {
 // Streaming tests for JSON array format
 #[tokio::test]
 async fn test_json_array_streaming_required_mode() {
-    use sgl_model_gateway::protocols::common::Tool;
+    use smg::protocols::common::Tool;
 
     // Test that simulates the exact streaming pattern from required mode
     let mut parser = JsonParser::new();
@@ -174,7 +174,7 @@ async fn test_json_array_streaming_required_mode() {
     // Define test tools
     let tools = vec![Tool {
         tool_type: "function".to_string(),
-        function: sgl_model_gateway::protocols::common::Function {
+        function: smg::protocols::common::Function {
             name: "get_weather".to_string(),
             description: Some("Get weather".to_string()),
             parameters: serde_json::json!({}),
@@ -240,7 +240,7 @@ async fn test_json_array_streaming_required_mode() {
 
 #[tokio::test]
 async fn test_json_array_multiple_tools_streaming() {
-    use sgl_model_gateway::protocols::common::Tool;
+    use smg::protocols::common::Tool;
 
     // Test with multiple tools in array
     let mut parser = JsonParser::new();
@@ -248,7 +248,7 @@ async fn test_json_array_multiple_tools_streaming() {
     let tools = vec![
         Tool {
             tool_type: "function".to_string(),
-            function: sgl_model_gateway::protocols::common::Function {
+            function: smg::protocols::common::Function {
                 name: "get_weather".to_string(),
                 description: Some("Get weather".to_string()),
                 parameters: serde_json::json!({}),
@@ -257,7 +257,7 @@ async fn test_json_array_multiple_tools_streaming() {
         },
         Tool {
             tool_type: "function".to_string(),
-            function: sgl_model_gateway::protocols::common::Function {
+            function: smg::protocols::common::Function {
                 name: "get_news".to_string(),
                 description: Some("Get news".to_string()),
                 parameters: serde_json::json!({}),
@@ -305,14 +305,14 @@ async fn test_json_array_multiple_tools_streaming() {
 
 #[tokio::test]
 async fn test_json_array_closing_bracket_separate_chunk() {
-    use sgl_model_gateway::protocols::common::Tool;
+    use smg::protocols::common::Tool;
 
     // Test case where the closing ] comes as a separate chunk
     let mut parser = JsonParser::new();
 
     let tools = vec![Tool {
         tool_type: "function".to_string(),
-        function: sgl_model_gateway::protocols::common::Function {
+        function: smg::protocols::common::Function {
             name: "get_weather".to_string(),
             description: Some("Get weather".to_string()),
             parameters: json!({}),
@@ -366,14 +366,14 @@ async fn test_json_array_closing_bracket_separate_chunk() {
 
 #[tokio::test]
 async fn test_json_single_object_with_trailing_text() {
-    use sgl_model_gateway::protocols::common::Tool;
+    use smg::protocols::common::Tool;
 
     // Test single object format (no array) with trailing text
     let mut parser = JsonParser::new();
 
     let tools = vec![Tool {
         tool_type: "function".to_string(),
-        function: sgl_model_gateway::protocols::common::Function {
+        function: smg::protocols::common::Function {
             name: "get_weather".to_string(),
             description: Some("Get weather".to_string()),
             parameters: serde_json::json!({}),
@@ -420,14 +420,14 @@ async fn test_json_single_object_with_trailing_text() {
 
 #[tokio::test]
 async fn test_json_single_object_with_bracket_in_text() {
-    use sgl_model_gateway::protocols::common::Tool;
+    use smg::protocols::common::Tool;
 
     // Test that ] in normal text is NOT stripped for single object format
     let mut parser = JsonParser::new();
 
     let tools = vec![Tool {
         tool_type: "function".to_string(),
-        function: sgl_model_gateway::protocols::common::Function {
+        function: smg::protocols::common::Function {
             name: "get_weather".to_string(),
             description: Some("Get weather".to_string()),
             parameters: serde_json::json!({}),
@@ -472,14 +472,14 @@ async fn test_json_single_object_with_bracket_in_text() {
 
 #[tokio::test]
 async fn test_json_array_bracket_in_text_after_tools() {
-    use sgl_model_gateway::protocols::common::Tool;
+    use smg::protocols::common::Tool;
 
     // Test that ] in normal text AFTER array tools is preserved
     let mut parser = JsonParser::new();
 
     let tools = vec![Tool {
         tool_type: "function".to_string(),
-        function: sgl_model_gateway::protocols::common::Function {
+        function: smg::protocols::common::Function {
             name: "get_weather".to_string(),
             description: Some("Get weather".to_string()),
             parameters: serde_json::json!({}),

sgl-model-gateway/tests/tool_parser_kimik2.rs

@@ -1,6 +1,6 @@
 //! Kimi K2 Parser Integration Tests
 
-use sgl_model_gateway::tool_parser::{KimiK2Parser, ToolParser};
+use smg::tool_parser::{KimiK2Parser, ToolParser};
 
 mod common;
 use common::create_test_tools;

sgl-model-gateway/tests/tool_parser_llama.rs

@@ -2,7 +2,7 @@
 //!
 //! Tests for the Llama parser which handles <|python_tag|> format and plain JSON
 
-use sgl_model_gateway::tool_parser::{LlamaParser, ToolParser};
+use smg::tool_parser::{LlamaParser, ToolParser};
 
 mod common;
 use common::{create_test_tools, streaming_helpers::*};

sgl-model-gateway/tests/tool_parser_minimax_m2.rs

@@ -1,6 +1,6 @@
 //! MiniMax M2 Parser Integration Tests
 
-use sgl_model_gateway::tool_parser::{MinimaxM2Parser, ToolParser};
+use smg::tool_parser::{MinimaxM2Parser, ToolParser};
 
 mod common;
 use common::create_test_tools;

sgl-model-gateway/tests/tool_parser_mistral.rs

@@ -3,7 +3,7 @@
 //! Tests for the Mistral parser which handles [TOOL_CALLS] format
 
 use serde_json::json;
-use sgl_model_gateway::tool_parser::{MistralParser, ToolParser};
+use smg::tool_parser::{MistralParser, ToolParser};
 
 #[tokio::test]
 async fn test_mistral_single_tool() {
@@ -158,14 +158,14 @@ Let me execute these searches for you."#;
 
 #[tokio::test]
 async fn test_mistral_streaming_closing_bracket() {
-    use sgl_model_gateway::protocols::common::Tool;
+    use smg::protocols::common::Tool;
 
     // Test that closing ] is stripped for Mistral array format
     let mut parser = MistralParser::new();
 
     let tools = vec![Tool {
         tool_type: "function".to_string(),
-        function: sgl_model_gateway::protocols::common::Function {
+        function: smg::protocols::common::Function {
             name: "get_weather".to_string(),
             description: Some("Get weather".to_string()),
             parameters: json!({}),
@@ -218,14 +218,14 @@ async fn test_mistral_streaming_closing_bracket() {
 
 #[tokio::test]
 async fn test_mistral_streaming_bracket_in_text_after_tools() {
-    use sgl_model_gateway::protocols::common::Tool;
+    use smg::protocols::common::Tool;
 
     // Test that ] in normal text AFTER tool calls is preserved
     let mut parser = MistralParser::new();
 
     let tools = vec![Tool {
         tool_type: "function".to_string(),
-        function: sgl_model_gateway::protocols::common::Function {
+        function: smg::protocols::common::Function {
             name: "get_weather".to_string(),
             description: Some("Get weather".to_string()),
             parameters: json!({}),

sgl-model-gateway/tests/tool_parser_mixed_edge_cases.rs

@@ -3,7 +3,7 @@
 //! Tests for edge cases across parsers and mixed format scenarios
 
 use serde_json::json;
-use sgl_model_gateway::tool_parser::{
+use smg::tool_parser::{
     JsonParser, LlamaParser, MistralParser, PythonicParser, QwenParser, ToolParser,
 };
 

sgl-model-gateway/tests/tool_parser_partial_json.rs

@@ -2,7 +2,7 @@
 //!
 //! Tests for the partial JSON parser with allow_partial_strings flag behavior
 
-use sgl_model_gateway::tool_parser::partial_json::PartialJson;
+use smg::tool_parser::partial_json::PartialJson;
 
 #[test]
 fn test_partial_string_flag_disallows_incomplete_strings() {

sgl-model-gateway/tests/tool_parser_pythonic.rs

@@ -3,7 +3,7 @@
 //! Tests for the Pythonic parser which handles Python function call syntax
 
 use serde_json::json;
-use sgl_model_gateway::tool_parser::{PythonicParser, ToolParser};
+use smg::tool_parser::{PythonicParser, ToolParser};
 
 mod common;
 use common::create_test_tools;

sgl-model-gateway/tests/tool_parser_qwen.rs

@@ -3,7 +3,7 @@
 //! Tests for the Qwen parser which handles <tool_call>...</tool_call> format
 
 use serde_json::json;
-use sgl_model_gateway::tool_parser::{QwenParser, ToolParser};
+use smg::tool_parser::{QwenParser, ToolParser};
 
 mod common;
 use common::{create_test_tools, streaming_helpers::*};

sgl-model-gateway/tests/tool_parser_step3.rs

@@ -1,6 +1,6 @@
 //! Step3 Parser Integration Tests
 
-use sgl_model_gateway::tool_parser::{Step3Parser, ToolParser};
+use smg::tool_parser::{Step3Parser, ToolParser};
 
 mod common;
 use common::create_test_tools;

sgl-model-gateway/tests/vision_golden_tests.rs

@@ -17,7 +17,7 @@
 use std::{fs::File, io::Read, path::Path};
 
 use ndarray::{Array4, Array5};
-use sgl_model_gateway::multimodal::vision::{
+use smg::multimodal::vision::{
     image_processor::ModelSpecificValue, ImagePreProcessor, Llama4VisionProcessor, LlavaProcessor,
     Phi3VisionProcessor, Phi4VisionProcessor, PixtralProcessor, PreProcessorConfig,
     Qwen2VLProcessor, Qwen3VLProcessor,

sgl-model-gateway/tests/wasm_test.rs

@@ -15,7 +15,7 @@ use axum::{
     extract::Request,
     http::{header::CONTENT_TYPE, StatusCode},
 };
-use sgl_model_gateway::{
+use smg::{
     app_context::AppContext,
     config::RouterConfig,
     core::{
@@ -101,17 +101,14 @@ async fn create_test_context_with_wasm() -> Arc<AppContext> {
 
     // Initialize JobQueue after AppContext is created
     let weak_context = Arc::downgrade(&app_context);
-    let job_queue = sgl_model_gateway::core::JobQueue::new(
-        sgl_model_gateway::core::JobQueueConfig::default(),
-        weak_context,
-    );
+    let job_queue = smg::core::JobQueue::new(smg::core::JobQueueConfig::default(), weak_context);
     app_context
         .worker_job_queue
         .set(job_queue)
         .expect("JobQueue should only be initialized once");
 
     // Initialize WorkflowEngine and register workflows
-    use sgl_model_gateway::{
+    use smg::{
         core::steps::{create_worker_registration_workflow, create_worker_removal_workflow},
         workflow::WorkflowEngine,
     };
@@ -134,7 +131,7 @@ async fn create_test_context_with_wasm() -> Arc<AppContext> {
         .expect("WorkflowEngine should only be initialized once");
 
     // Initialize MCP manager with empty config
-    use sgl_model_gateway::mcp::{McpConfig, McpManager};
+    use smg::mcp::{McpConfig, McpManager};
     let empty_config = McpConfig {
         servers: vec![],
         pool: Default::default(),
@@ -206,7 +203,7 @@ async fn create_test_app_with_wasm() -> (axum::Router, Arc<AppContext>, TempDir)
 
     let app = build_app(
         app_state,
-        sgl_model_gateway::middleware::AuthConfig { api_key: None },
+        smg::middleware::AuthConfig { api_key: None },
         None, // No control plane auth for tests
         256 * 1024 * 1024,
         request_id_headers,
@@ -231,7 +228,7 @@ async fn test_wasm_api_add_module() {
             file_path: wasm_file_path.clone(),
             module_type: WasmModuleType::Middleware,
             attach_points: vec![WasmModuleAttachPoint::Middleware(
-                sgl_model_gateway::wasm::module::MiddlewareAttachPoint::OnRequest,
+                smg::wasm::module::MiddlewareAttachPoint::OnRequest,
             )],
             add_result: None,
         }],
@@ -257,7 +254,7 @@ async fn test_wasm_api_add_module() {
     let module_result = &response_json.modules[0].add_result;
 
     // Print error for debugging
-    if let Some(sgl_model_gateway::wasm::module::WasmModuleAddResult::Error(err)) = module_result {
+    if let Some(smg::wasm::module::WasmModuleAddResult::Error(err)) = module_result {
         eprintln!("Module registration failed: {}", err);
     }
 
@@ -278,9 +275,7 @@ async fn test_wasm_api_add_module() {
         let modules = wasm_manager.get_modules().expect("Failed to get modules");
         assert!(!modules.is_empty(), "Module should be registered");
 
-        if let Some(sgl_model_gateway::wasm::module::WasmModuleAddResult::Success(uuid)) =
-            module_result
-        {
+        if let Some(smg::wasm::module::WasmModuleAddResult::Success(uuid)) = module_result {
             let module = wasm_manager
                 .get_module(*uuid)
                 .expect("Failed to get module");
@@ -299,7 +294,7 @@ async fn test_wasm_api_add_module_invalid_file() {
             file_path: "/nonexistent/path/to/module.component.wasm".to_string(),
             module_type: WasmModuleType::Middleware,
             attach_points: vec![WasmModuleAttachPoint::Middleware(
-                sgl_model_gateway::wasm::module::MiddlewareAttachPoint::OnRequest,
+                smg::wasm::module::MiddlewareAttachPoint::OnRequest,
             )],
             add_result: None,
         }],
@@ -331,7 +326,7 @@ async fn test_wasm_api_add_module_invalid_file() {
     assert!(module_result.is_some());
 
     // Verify it's an error result
-    if let Some(sgl_model_gateway::wasm::module::WasmModuleAddResult::Error(_)) = module_result {
+    if let Some(smg::wasm::module::WasmModuleAddResult::Error(_)) = module_result {
         // Expected error
     } else {
         panic!("Expected error result for invalid file path");
@@ -354,7 +349,7 @@ async fn test_wasm_api_add_module_invalid_wasm() {
             file_path: invalid_wasm_path.to_str().unwrap().to_string(),
             module_type: WasmModuleType::Middleware,
             attach_points: vec![WasmModuleAttachPoint::Middleware(
-                sgl_model_gateway::wasm::module::MiddlewareAttachPoint::OnRequest,
+                smg::wasm::module::MiddlewareAttachPoint::OnRequest,
             )],
             add_result: None,
         }],
@@ -386,7 +381,7 @@ async fn test_wasm_api_add_module_invalid_wasm() {
     assert!(module_result.is_some());
 
     // Verify it's an error result
-    if let Some(sgl_model_gateway::wasm::module::WasmModuleAddResult::Error(_)) = module_result {
+    if let Some(smg::wasm::module::WasmModuleAddResult::Error(_)) = module_result {
         // Expected error
     } else {
         panic!("Expected error result for invalid WASM file");
@@ -405,7 +400,7 @@ async fn test_wasm_api_list_modules() {
             file_path: wasm_file_path.clone(),
             module_type: WasmModuleType::Middleware,
             attach_points: vec![WasmModuleAttachPoint::Middleware(
-                sgl_model_gateway::wasm::module::MiddlewareAttachPoint::OnRequest,
+                smg::wasm::module::MiddlewareAttachPoint::OnRequest,
             )],
             add_result: None,
         }],
@@ -473,7 +468,7 @@ async fn test_wasm_api_remove_module() {
             file_path: wasm_file_path.clone(),
             module_type: WasmModuleType::Middleware,
             attach_points: vec![WasmModuleAttachPoint::Middleware(
-                sgl_model_gateway::wasm::module::MiddlewareAttachPoint::OnRequest,
+                smg::wasm::module::MiddlewareAttachPoint::OnRequest,
             )],
             add_result: None,
         }],
@@ -503,24 +498,23 @@ async fn test_wasm_api_remove_module() {
     tokio::time::sleep(Duration::from_millis(500)).await;
 
     // Get the module UUID
-    let module_uuid =
-        if let Some(sgl_model_gateway::wasm::module::WasmModuleAddResult::Success(uuid)) =
-            &response_json.modules[0].add_result
-        {
-            *uuid
+    let module_uuid = if let Some(smg::wasm::module::WasmModuleAddResult::Success(uuid)) =
+        &response_json.modules[0].add_result
+    {
+        *uuid
+    } else {
+        // If we can't get UUID from response, try to find it from manager
+        if let Some(wasm_manager) = app_context.wasm_manager.as_ref() {
+            let modules = wasm_manager.get_modules().expect("Failed to get modules");
+            modules
+                .iter()
+                .find(|m| m.module_meta.name == "test_module_remove")
+                .map(|m| m.module_uuid)
+                .expect("Module should be registered")
         } else {
-            // If we can't get UUID from response, try to find it from manager
-            if let Some(wasm_manager) = app_context.wasm_manager.as_ref() {
-                let modules = wasm_manager.get_modules().expect("Failed to get modules");
-                modules
-                    .iter()
-                    .find(|m| m.module_meta.name == "test_module_remove")
-                    .map(|m| m.module_uuid)
-                    .expect("Module should be registered")
-            } else {
-                panic!("WASM manager not available");
-            }
-        };
+            panic!("WASM manager not available");
+        }
+    };
 
     // Now remove the module
     let remove_response = app
@@ -599,7 +593,7 @@ async fn test_wasm_module_duplicate_sha256() {
             file_path: wasm_file_path.clone(),
             module_type: WasmModuleType::Middleware,
             attach_points: vec![WasmModuleAttachPoint::Middleware(
-                sgl_model_gateway::wasm::module::MiddlewareAttachPoint::OnRequest,
+                smg::wasm::module::MiddlewareAttachPoint::OnRequest,
             )],
             add_result: None,
         }],
@@ -630,7 +624,7 @@ async fn test_wasm_module_duplicate_sha256() {
             file_path: wasm_file_path.clone(), // Same file
             module_type: WasmModuleType::Middleware,
             attach_points: vec![WasmModuleAttachPoint::Middleware(
-                sgl_model_gateway::wasm::module::MiddlewareAttachPoint::OnRequest,
+                smg::wasm::module::MiddlewareAttachPoint::OnRequest,
             )],
             add_result: None,
         }],
@@ -662,9 +656,7 @@ async fn test_wasm_module_duplicate_sha256() {
     assert!(module_result.is_some());
 
     // Verify it's an error result (duplicate)
-    if let Some(sgl_model_gateway::wasm::module::WasmModuleAddResult::Error(err_msg)) =
-        module_result
-    {
+    if let Some(smg::wasm::module::WasmModuleAddResult::Error(err_msg)) = module_result {
         assert!(
             err_msg.contains("duplicate")
                 || err_msg.contains("Duplicate")
@@ -692,7 +684,7 @@ async fn test_wasm_module_execution() {
         .expect("Workflow engine should be initialized");
 
     // Create workflow context for registration
-    use sgl_model_gateway::{
+    use smg::{
         core::steps::WasmModuleConfigRequest,
         workflow::{WorkflowContext, WorkflowId, WorkflowInstanceId},
     };
@@ -702,7 +694,7 @@ async fn test_wasm_module_execution() {
         file_path: wasm_file_path.clone(),
         module_type: WasmModuleType::Middleware,
         attach_points: vec![WasmModuleAttachPoint::Middleware(
-            sgl_model_gateway::wasm::module::MiddlewareAttachPoint::OnRequest,
+            smg::wasm::module::MiddlewareAttachPoint::OnRequest,
         )],
         add_result: None,
     };
@@ -736,14 +728,14 @@ async fn test_wasm_module_execution() {
             .expect("Failed to get workflow status");
 
         match state.status {
-            sgl_model_gateway::workflow::WorkflowStatus::Completed => {
+            smg::workflow::WorkflowStatus::Completed => {
                 // Extract module UUID from context
                 if let Some(uuid_arc) = state.context.get::<Uuid>("module_uuid") {
                     module_uuid = Some(*uuid_arc.as_ref());
                 }
                 break;
             }
-            sgl_model_gateway::workflow::WorkflowStatus::Failed => {
+            smg::workflow::WorkflowStatus::Failed => {
                 panic!("Workflow failed: {:?}", state);
             }
             _ => {
@@ -764,7 +756,7 @@ async fn test_wasm_module_execution() {
     let (initial_total, initial_success, initial_failed, _, _) = wasm_manager.get_metrics();
 
     // Execute the module
-    use sgl_model_gateway::wasm::{
+    use smg::wasm::{
         spec::sgl::model_gateway::middleware_types,
         types::{WasmComponentInput, WasmComponentOutput},
     };
@@ -780,9 +772,8 @@ async fn test_wasm_module_execution() {
     };
 
     let input = WasmComponentInput::MiddlewareRequest(request);
-    let attach_point = WasmModuleAttachPoint::Middleware(
-        sgl_model_gateway::wasm::module::MiddlewareAttachPoint::OnRequest,
-    );
+    let attach_point =
+        WasmModuleAttachPoint::Middleware(smg::wasm::module::MiddlewareAttachPoint::OnRequest);
 
     // Execute the module
     let result = wasm_manager

sgl-model-gateway/tests/workflow_test.rs

@@ -8,7 +8,7 @@ use std::{
     time::Duration,
 };
 
-use sgl_model_gateway::workflow::*;
+use smg::workflow::*;
 use tokio::time::sleep;
 
 // Test step that counts invocations