avoid double serialization

Author: mcharytoniuk

src/balancer/chunk_forwarding_session_controller/mod.rs

@@ -40,7 +40,7 @@ where
 {
     async fn send_response(&mut self, message: OutgoingMessage) -> anyhow::Result<()> {
         let transformed_message = self.transformer.transform(message).await?;
-        let stringified_message = serde_json::to_string(&transformed_message)?;
+        let stringified_message = self.transformer.stringify(&transformed_message)?;
 
         self.chunk_tx.send(stringified_message)?;
 

src/balancer/chunk_forwarding_session_controller/transforms_outgoing_message.rs

@@ -9,4 +9,8 @@ pub trait TransformsOutgoingMessage {
     type TransformedMessage: Serialize;
 
     async fn transform(&self, message: OutgoingMessage) -> Result<Self::TransformedMessage>;
+
+    fn stringify(&self, message: &Self::TransformedMessage) -> Result<String> {
+        Ok(serde_json::to_string(message)?)
+    }
 }

src/balancer/compatibility/openai_service/http_route/post_chat_completions.rs

@@ -143,9 +143,13 @@ impl TransformsOutgoingMessage for OpenAICombinedResponseTransfomer {
             _ => Err(anyhow!("Unexpected message type: {:?}", message)),
         }
     }
+
+    fn stringify(&self, message: &Self::TransformedMessage) -> anyhow::Result<String> {
+        Ok(message.clone())
+    }
 }
 
-#[post("/v1/chat_completions")]
+#[post("/v1/chat/completions")]
 async fn respond(
     app_data: web::Data<AppData>,
     openai_params: web::Json<OpenAICompletionRequestParams>,