refactor ai serving modules and crates

apps/aether-gateway/src/execution_runtime/kiro_web_search.rs

@@ -937,7 +937,7 @@ fn build_web_search_sse_body(
     input_tokens: u64,
 ) -> Result<Vec<u8>, serde_json::Error> {
     let events = build_web_search_events(model, query, tool_use_id, search_results, input_tokens);
-    crate::ai_pipeline_api::encode_kiro_sse_events(events)
+    crate::ai_serving::api::encode_kiro_sse_events(events)
 }
 
 fn build_web_search_events(

apps/aether-gateway/src/execution_runtime/mod.rs

@@ -33,6 +33,7 @@ pub(crate) use self::fallback::{
 pub(crate) use crate::orchestration::{
     append_local_failover_policy_to_value, LocalFailoverAnalysis, LocalFailoverDecision,
 };
+pub(crate) use aether_ai_serving::{ConversionMode, ExecutionStrategy};
 pub use server::{
     build_execution_runtime_router, build_execution_runtime_router_with_request_concurrency_limit,
     build_execution_runtime_router_with_request_gates, serve_execution_runtime_tcp,
@@ -52,46 +53,6 @@ pub(crate) use transport::{
     DirectUpstreamStreamExecution, ExecutionRuntimeTransportError,
 };
 
-#[derive(Debug, Clone, Copy, PartialEq, Eq, Serialize, Deserialize)]
-#[serde(rename_all = "snake_case")]
-pub(crate) enum ExecutionStrategy {
-    GatewayAffinityForward,
-    RawPublicProxy,
-    LocalSameFormat,
-    LocalCrossFormat,
-}
-
-impl ExecutionStrategy {
-    pub(crate) const fn as_str(self) -> &'static str {
-        match self {
-            Self::GatewayAffinityForward => "gateway_affinity_forward",
-            Self::RawPublicProxy => "raw_public_proxy",
-            Self::LocalSameFormat => "local_same_format",
-            Self::LocalCrossFormat => "local_cross_format",
-        }
-    }
-}
-
-#[derive(Debug, Clone, Copy, PartialEq, Eq, Serialize, Deserialize)]
-#[serde(rename_all = "snake_case")]
-pub(crate) enum ConversionMode {
-    None,
-    RequestOnly,
-    ResponseOnly,
-    Bidirectional,
-}
-
-impl ConversionMode {
-    pub(crate) const fn as_str(self) -> &'static str {
-        match self {
-            Self::None => "none",
-            Self::RequestOnly => "request_only",
-            Self::ResponseOnly => "response_only",
-            Self::Bidirectional => "bidirectional",
-        }
-    }
-}
-
 #[derive(Debug, Clone, PartialEq, Serialize, Deserialize)]
 pub(crate) struct ClientIntent {
     pub(crate) client_contract: String,

apps/aether-gateway/src/execution_runtime/stream/execution.rs

@@ -39,7 +39,7 @@ use self::execution_failures::{
     build_stream_failure_from_execution_error, build_stream_failure_report,
     handle_prefetch_stream_failure, submit_midstream_stream_failure, StreamFailureReport,
 };
-use crate::ai_pipeline_api::{
+use crate::ai_serving::api::{
     maybe_bridge_standard_sync_json_to_stream, maybe_build_provider_private_stream_normalizer,
     maybe_build_stream_response_rewriter, normalize_provider_private_report_context,
     StreamingStandardTerminalObserver,

apps/aether-gateway/src/execution_runtime/stream_pump.rs

@@ -13,7 +13,7 @@ use futures_util::{Stream, StreamExt};
 use serde_json::Value;
 use tracing::warn;
 
-use crate::ai_pipeline_api::{
+use crate::ai_serving::api::{
     maybe_bridge_standard_sync_json_to_stream, maybe_build_provider_private_stream_normalizer,
     normalize_provider_private_report_context, StreamingStandardTerminalObserver,
 };
@@ -399,7 +399,7 @@ fn should_treat_upstream_response_as_stream(
     report_context
         .get("envelope_name")
         .and_then(Value::as_str)
-        .is_some_and(|value| value.eq_ignore_ascii_case(crate::ai_pipeline::KIRO_ENVELOPE_NAME))
+        .is_some_and(|value| value.eq_ignore_ascii_case(crate::ai_serving::KIRO_ENVELOPE_NAME))
 }
 
 fn should_buffer_non_stream_response(
@@ -495,7 +495,7 @@ fn maybe_bridge_non_sse_sync_json_to_stream(
     body_bytes: &[u8],
     provider_api_format: &str,
     report_context: &Value,
-) -> Result<Option<crate::ai_pipeline::SyncToStreamBridgeOutcome>, GatewayError> {
+) -> Result<Option<crate::ai_serving::SyncToStreamBridgeOutcome>, GatewayError> {
     if !(200..300).contains(&status_code) || body_bytes.is_empty() {
         return Ok(None);
     }
@@ -584,7 +584,7 @@ fn format_error_chain(err: &(dyn std::error::Error + 'static)) -> String {
 fn observe_stream_chunk(
     observer: &mut StreamingStandardTerminalObserver,
     report_context: &Value,
-    private_stream_normalizer: Option<&mut crate::ai_pipeline::ProviderPrivateStreamNormalizer<'_>>,
+    private_stream_normalizer: Option<&mut crate::ai_serving::ProviderPrivateStreamNormalizer<'_>>,
     observer_buffered: &mut Vec<u8>,
     chunk: &[u8],
 ) {
@@ -608,7 +608,7 @@ fn observe_stream_chunk(
 fn finalize_stream_terminal_summary(
     observer: &mut StreamingStandardTerminalObserver,
     report_context: &Value,
-    private_stream_normalizer: Option<&mut crate::ai_pipeline::ProviderPrivateStreamNormalizer<'_>>,
+    private_stream_normalizer: Option<&mut crate::ai_serving::ProviderPrivateStreamNormalizer<'_>>,
     observer_buffered: &mut Vec<u8>,
 ) -> Option<ExecutionStreamTerminalSummary> {
     if let Some(normalizer) = private_stream_normalizer {

apps/aether-gateway/src/execution_runtime/submission.rs

@@ -1,6 +1,6 @@
 #[cfg(test)]
-use crate::ai_pipeline_api::core_success_background_report_kind;
-use crate::ai_pipeline_api::{
+use crate::ai_serving::api::core_success_background_report_kind;
+use crate::ai_serving::api::{
     build_core_error_body_for_client_format, core_error_background_report_kind,
     core_error_default_client_api_format, is_core_error_finalize_kind,
     maybe_compile_sync_finalize_response,

apps/aether-gateway/src/execution_runtime/sync/execution.rs

@@ -15,7 +15,7 @@ use axum::http::Response;
 use base64::Engine as _;
 use tracing::warn;
 
-use crate::ai_pipeline_api::{
+use crate::ai_serving::api::{
     implicit_sync_finalize_report_kind, maybe_build_sync_finalize_outcome,
     LocalCoreSyncFinalizeOutcome,
 };

apps/aether-gateway/src/execution_runtime/tests.rs

@@ -3,12 +3,12 @@ use axum::http::Request;
 use base64::Engine as _;
 use serde_json::json;
 
-use crate::ai_pipeline_api::{
+use crate::ai_serving::api::{
     build_gemini_stream_plan_from_decision, build_gemini_sync_plan_from_decision,
     build_openai_responses_stream_plan_from_decision,
     build_openai_responses_sync_plan_from_decision, build_passthrough_sync_plan_from_decision,
     build_standard_stream_plan_from_decision, build_standard_sync_plan_from_decision,
-    GatewayControlSyncDecisionResponse,
+    AiExecutionDecision,
 };
 use crate::execution_runtime::submission::{
     build_best_effort_local_core_error_body, has_nested_error,
@@ -34,10 +34,8 @@ fn test_parts() -> http::request::Parts {
     parts
 }
 
-fn missing_exact_provider_request_payload(
-    decision_kind: &str,
-) -> GatewayControlSyncDecisionResponse {
-    GatewayControlSyncDecisionResponse {
+fn missing_exact_provider_request_payload(decision_kind: &str) -> AiExecutionDecision {
+    AiExecutionDecision {
         action: "execution_runtime".to_string(),
         decision_kind: Some(decision_kind.to_string()),
         execution_strategy: Some("local_same_format".to_string()),