Fix preserved state across OpenAI requests when backend_body / backend_params are used (#236)

Signed-off-by: Mark Kurtz <[email protected]>
Signed-off-by: dalthecow <[email protected]>
Signed-off-by: dependabot[bot] <[email protected]>
Co-authored-by: Benjamin Blue <[email protected]>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
Co-authored-by: Samuel Monson <[email protected]>
Co-authored-by: Copilot <[email protected]>

Author: 5 people

src/guidellm/backend/openai.py

@@ -1,4 +1,5 @@
 import base64
+import copy
 import json
 import time
 from collections.abc import AsyncGenerator
@@ -428,9 +429,9 @@ def _extra_body(self, endpoint_type: EndpointType) -> dict[str, Any]:
             or MODELS in self.extra_body
             or TEXT_COMPLETIONS in self.extra_body
         ):
-            return self.extra_body.get(endpoint_type, {})
+            return copy.deepcopy(self.extra_body.get(endpoint_type, {}))
 
-        return self.extra_body
+        return copy.deepcopy(self.extra_body)
 
     def _completions_payload(
         self,