diff --git a/src/server/routes/query.rs b/src/server/routes/query.rs
index cd1705b..b540db4 100644
--- a/src/server/routes/query.rs
+++ b/src/server/routes/query.rs
@@ -5,7 +5,7 @@ use async_openai::types::{
     ChatCompletionRequestSystemMessage, ChatCompletionRequestUserMessage,
     CreateChatCompletionRequestArgs, ResponseFormat, ResponseFormatJsonSchema,
 };
-use axum::{response::IntoResponse, Extension, Json};
+use axum::{http::StatusCode, response::IntoResponse, Extension, Json};
 use serde::Deserialize;
 use serde_json::json;
 use std::sync::Arc;
@@ -16,6 +16,22 @@ pub struct QueryInput {
     query: String,
 }
 
+#[derive(Debug, Deserialize)]
+struct Reference {
+    reference: String,
+}
+
+#[derive(Debug, Deserialize)]
+struct LLMResponseFormat {
+    answer: String,
+    references: Vec<Reference>,
+}
+// impl IntoResponse for LLMResponseFormat {
+//     fn into_response(self) -> axum::response::Response {
+//         (StatusCode::OK, Json(self))
+//     }
+// }
+
 pub async fn query_handler(
     Extension(db_client): Extension<Arc<SurrealDbClient>>,
     Json(query): Json<QueryInput>,
@@ -127,12 +143,16 @@ pub async fn query_handler(
         .first()
         .and_then(|choice| choice.message.content.as_ref())
         .ok_or(ApiError::QueryError(
-            "No content found in LLM response".to_string(),
+            "No content found in LLM response".into(),
         ))?;
 
     info!("{:?}", answer);
 
-    // info!("{:#?}", entities_json);
+    let parsed: LLMResponseFormat = serde_json::from_str(answer).map_err(|e| {
+        ApiError::QueryError(format!("Failed to parse LLM response into analysis: {}", e))
+    })?;
 
-    Ok(answer.clone().into_response())
+    info!("{:?}", parsed);
+
+    Ok(parsed.answer.into_response())
 }