fix: use configured model for all LiteLLM completions

complete_as and complete_as_user were passing internal logical model names
(jd-generator, profile-writer, etc.) directly to LiteLLM, which only
knows the configured model alias (askash-main). Route all completions
through self.model so LiteLLM can resolve the correct backend model.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

src/providers/llm/litellm_provider.rs

@@ -142,17 +142,20 @@ impl AiProvider for LiteLLMProvider {
 
     async fn complete_as(
         &self,
-        model: &str,
+        _model: &str,
         system_prompt: &str,
         user_prompt: &str,
     ) -> Result<String, AppError> {
-        self.chat(&self.api_key, model, system_prompt, user_prompt).await
+        // LiteLLM routes everything through the configured model (e.g. askash-main).
+        // Logical model names like "jd-generator" are internal identifiers only.
+        let model = self.model.clone();
+        self.chat(&self.api_key, &model, system_prompt, user_prompt).await
     }
 
     async fn complete_as_user(
         &self,
         user_id: Option<&str>,
-        model: &str,
+        _model: &str,
         system_prompt: &str,
         user_prompt: &str,
     ) -> Result<String, AppError> {
@@ -160,7 +163,8 @@ impl AiProvider for LiteLLMProvider {
             (Some(uid), Some(client)) => client.get_key(uid).await,
             _ => None,
         };
-        self.chat(key.as_deref().unwrap_or(&self.api_key), model, system_prompt, user_prompt)
+        let model = self.model.clone();
+        self.chat(key.as_deref().unwrap_or(&self.api_key), &model, system_prompt, user_prompt)
             .await
     }
 }