fix(desktop): resolve 2 release-blocking P1 defects

P1-04: GenerationPipeline hardcoded model="default" causing classroom
generation 404. Added model field to GenerationPipeline struct, passed
from kernel config via with_driver(driver, model). Static scene
generation now receives model parameter.

P1-03: LLM API concurrent 500 DATABASE_ERROR. Added transient DB error
retry (PoolTimedOut/Io) in create_relay_task with 200ms backoff.
Recommend setting ZCLAW_DB_MIN_CONNECTIONS=10 for burst resilience.

crates/zclaw-kernel/src/generation/mod.rs

@@ -248,6 +248,7 @@ pub struct GenerationPipeline {
     scenes: Arc<RwLock<Vec<GeneratedScene>>>,
     agents_store: Arc<RwLock<Vec<AgentProfile>>>,
     driver: Option<Arc<dyn LlmDriver>>,
+    model: String,
 }
 
 impl GenerationPipeline {
@@ -265,12 +266,14 @@ impl GenerationPipeline {
             scenes: Arc::new(RwLock::new(Vec::new())),
             agents_store: Arc::new(RwLock::new(Vec::new())),
             driver: None,
+            model: "default".to_string(),
         }
     }
 
-    pub fn with_driver(driver: Arc<dyn LlmDriver>) -> Self {
+    pub fn with_driver(driver: Arc<dyn LlmDriver>, model: String) -> Self {
         Self {
             driver: Some(driver),
+            model,
             ..Self::new()
         }
     }
@@ -353,7 +356,7 @@ impl GenerationPipeline {
                 let item = item.clone();
                 async move {
                     if let Some(d) = driver {
-                        Self::generate_scene_with_llm_static(d.as_ref(), &item, i).await
+                        Self::generate_scene_with_llm_static(d.as_ref(), &self.model, &item, i).await
                     } else {
                         Self::generate_scene_for_item_static(&item, i)
                     }
@@ -413,7 +416,7 @@ impl GenerationPipeline {
         request: &GenerationRequest,
     ) -> Result<Vec<OutlineItem>> {
         let llm_request = CompletionRequest {
-            model: "default".to_string(),
+            model: self.model.clone(),
             system: Some(self.get_outline_system_prompt()),
             messages: vec![zclaw_types::Message::User {
                 content: prompt.to_string(),
@@ -469,6 +472,7 @@ Use Chinese if the topic is in Chinese. Include vivid metaphors and analogies."#
 
     async fn generate_scene_with_llm_static(
         driver: &dyn LlmDriver,
+        model: &str,
         item: &OutlineItem,
         order: usize,
     ) -> Result<GeneratedScene> {
@@ -488,7 +492,7 @@ Use Chinese if the topic is in Chinese. Include vivid metaphors and analogies."#
         );
 
         let llm_request = CompletionRequest {
-            model: "default".to_string(),
+            model: model.to_string(),
             system: Some(Self::get_scene_system_prompt_static()),
             messages: vec![zclaw_types::Message::User {
                 content: prompt,

crates/zclaw-saas/src/relay/service.rs

@@ -33,6 +33,11 @@ fn is_retryable_error(e: &reqwest::Error) -> bool {
 
 // ============ Relay Task Management ============
 
+/// 判断 sqlx 错误是否为可重试的瞬态错误（连接池耗尽、临时网络故障）
+fn is_transient_db_error(e: &sqlx::Error) -> bool {
+    matches!(e, sqlx::Error::PoolTimedOut | sqlx::Error::Io(_))
+}
+
 pub async fn create_relay_task(
     db: &PgPool,
     account_id: &str,
@@ -47,8 +52,21 @@ pub async fn create_relay_task(
     let request_hash = hash_request(request_body);
     let max_attempts = max_attempts.max(1).min(5);
 
-    // INSERT ... RETURNING 合并两次 DB 往返为一次
-    let row: RelayTaskRow = sqlx::query_as(
+    let query = sqlx::query_as::<_, RelayTaskRow>(
+        "INSERT INTO relay_tasks (id, account_id, provider_id, model_id, request_hash, request_body, status, priority, attempt_count, max_attempts, queued_at, created_at)
+         VALUES ($1, $2, $3, $4, $5, $6, 'queued', $7, 0, $8, $9, $9)
+         RETURNING id, account_id, provider_id, model_id, status, priority, attempt_count, max_attempts, input_tokens, output_tokens, error_message, queued_at, started_at, completed_at, created_at"
+    )
+    .bind(&id).bind(account_id).bind(provider_id).bind(model_id)
+    .bind(&request_hash).bind(request_body).bind(priority).bind(max_attempts as i64).bind(&now);
+
+    // 对瞬时 DB 错误（连接池耗尽/超时）重试一次
+    let row = match query.fetch_one(db).await {
+        Ok(row) => row,
+        Err(e) if is_transient_db_error(&e) => {
+            tracing::warn!("Transient DB error in create_relay_task, retrying: {}", e);
+            tokio::time::sleep(Duration::from_millis(200)).await;
+            sqlx::query_as::<_, RelayTaskRow>(
                 "INSERT INTO relay_tasks (id, account_id, provider_id, model_id, request_hash, request_body, status, priority, attempt_count, max_attempts, queued_at, created_at)
                  VALUES ($1, $2, $3, $4, $5, $6, 'queued', $7, 0, $8, $9, $9)
                  RETURNING id, account_id, provider_id, model_id, status, priority, attempt_count, max_attempts, input_tokens, output_tokens, error_message, queued_at, started_at, completed_at, created_at"
@@ -56,7 +74,10 @@ pub async fn create_relay_task(
             .bind(&id).bind(account_id).bind(provider_id).bind(model_id)
             .bind(&request_hash).bind(request_body).bind(priority).bind(max_attempts as i64).bind(&now)
             .fetch_one(db)
-    .await?;
+            .await?
+        }
+        Err(e) => return Err(e.into()),
+    };
 
     Ok(RelayTaskInfo {
         id: row.id, account_id: row.account_id, provider_id: row.provider_id, model_id: row.model_id,

desktop/src-tauri/src/classroom_commands/generate.rs

@@ -132,7 +132,7 @@ pub async fn classroom_generate(
     let pipeline = {
         let ks = kernel_state.lock().await;
         if let Some(kernel) = ks.as_ref() {
-            GenerationPipeline::with_driver(kernel.driver())
+            GenerationPipeline::with_driver(kernel.driver(), kernel.config().model().to_string())
         } else {
             GenerationPipeline::new()
         }