diff --git a/python/src/reag/client.py b/python/src/reag/client.py
index 7d174b6..10e6900 100644
--- a/python/src/reag/client.py
+++ b/python/src/reag/client.py
@@ -156,24 +156,27 @@ def format_doc(doc: Document) -> str:
 
             results = []
             for batch in batches:
+                tasks = []
+                # Create tasks for parallel processing within the batch
                 for document in batch:
-                    system = (
-                        f"{self.system}\n\n# Available source\n\n{format_doc(document)}"
+                    system = f"{self.system}\n\n# Available source\n\n{format_doc(document)}"
+                    tasks.append(
+                        acompletion(
+                            model=self.model,
+                            messages=[
+                                {"role": "system", "content": system},
+                                {"role": "user", "content": prompt},
+                            ],
+                            response_format=self.schema,
+                        )
                     )
 
-                    # Use litellm for model completion with the Pydantic schema
-                    response = await acompletion(
-                        model=self.model,
-                        messages=[
-                            {"role": "system", "content": system},
-                            {"role": "user", "content": prompt},
-                        ],
-                        response_format=self.schema,
-                    )
+                # Process all documents in the batch concurrently
+                batch_responses = await asyncio.gather(*tasks)
 
-                    message_content = response.choices[
-                        0
-                    ].message.content  # Might be a JSON string
+                # Process the responses
+                for document, response in zip(batch, batch_responses):
+                    message_content = response.choices[0].message.content
 
                     try:
                         # Ensure it's parsed as a dict
@@ -190,17 +193,15 @@ def format_doc(doc: Document) -> str:
                             QueryResult(
                                 content=data["source"].get("content", ""),
                                 reasoning=data["source"].get("reasoning", ""),
-                                is_irrelevant=data["source"].get(
-                                    "is_irrelevant", False
-                                ),
+                                is_irrelevant=data["source"].get("is_irrelevant", False),
                                 document=document,
                             )
                         )
                     except json.JSONDecodeError:
                         print("Error: Could not parse response:", message_content)
-                        continue  # Skip this iteration if parsing fails
+                        continue
 
-                return results
+            return results  # Moved outside the batch loop to return all results
 
         except Exception as e:
             raise Exception(f"Query failed: {str(e)}")
diff --git a/python/tests/test_client.py b/python/tests/test_client.py
index 10b57be..de91999 100644
--- a/python/tests/test_client.py
+++ b/python/tests/test_client.py
@@ -1,5 +1,5 @@
 import pytest
-from src.reag.client import ReagClient, Document, QueryResult
+from reag.client import ReagClient, Document, QueryResult
 
 
 @pytest.mark.asyncio