diff --git a/hackathon/api/routes.py b/hackathon/api/routes.py
index bb06318..5080700 100644
--- a/hackathon/api/routes.py
+++ b/hackathon/api/routes.py
@@ -326,6 +326,7 @@ def _extract_sample_data(answer: str, correct_answer) -> tuple[float, list[AISam
 async def provider_run(ai_provider: AIProvider, body: AIRunBody, api_key: str = Header(default=None)):
     _validate_body_model(ai_provider, body)
 
+
     param = ProviderParam(
         sample_id=body.sample_id,
         provider_model=body.provider_model,
@@ -336,6 +337,64 @@ async def provider_run(ai_provider: AIProvider, body: AIRunBody, api_key: str =
         top_p=body.top_p,
         top_k=body.top_k,
     )
+
+    from dataclasses import dataclass, asdict
+    # import json
+    # with open(f'C:/dev/hackathon-quantminds-2023/tmp/tmp.txt', 'w') as f:
+    #     json.dump(param.dict(), f)
+    # # param = ProviderParam(**param.dict())
+    # from dacite import from_dict
+    # param = from_dict(data_class=ProviderParam, data=param.dict())
+    provider = get_provider(ai_provider, api_key)
+    question = provider.build_question(prompt=param.prompt, context=param.context)
+    provider_answers = await get_provider(ai_provider, api_key).run([param])
+    answer = provider_answers[0].answer if provider_answers else ""
+
+    param = ProviderParam(
+        sample_id=body.sample_id,
+        provider_model=body.provider_model,
+        prompt=None,
+        context=None,
+        seed=body.seed,
+        temperature=body.temperature,
+        top_p=body.top_p,
+        top_k=body.top_k,
+        # question=(f"I am about to ask a question. Before responding, can you please start by repeating for me this current prompt, before giving your answer to the question, as I cannot see my own prompt easily.\n"
+        #           f"My question is: given the following QUESTION, and the following ANSWER that you gave,"
+        #           f"can you please explain to me how you came to the ANSWER. "
+        #           f"\n"
+        #           f"QUESTION: {question}\n"
+        #           f"ANSWER: {answer}"),
+        question="When you answer, please begin by repeating the content of this prompt in full, from the next line.\n\n"
+                 "Below I will refer to a question I asked you as 'QUESTION', and an answer you gave as 'ANSWER'. Can you please explain the answer you gave for each key.\n"
+                 "Please repeat the full JSON from 'ANSWER' in your current answer, or if there is no JSON in 'ANSWER', say so.\n\n"
+                 f"QUESTION: {question}\n\n"
+                 f"ANSWER: {answer}"
+    )
+
+    # param = ProviderParam(
+    #     sample_id=body.sample_id,
+    #     provider_model=body.provider_model,
+    #     prompt=None,
+    #     context=None,
+    #     seed=body.seed,
+    #     temperature=body.temperature,
+    #     top_p=body.top_p,
+    #     top_k=body.top_k,
+    #     # question=(f"I am about to ask a question. Before responding, can you please start by repeating for me this current prompt, before giving your answer to the question, as I cannot see my own prompt easily.\n"
+    #     #           f"My question is: given the following QUESTION, and the following ANSWER that you gave,"
+    #     #           f"can you please explain to me how you came to the ANSWER. "
+    #     #           f"\n"
+    #     #           f"QUESTION: {question}\n"
+    #     #           f"ANSWER: {answer}"),
+    #     question="When you answer, please begin by repeating the content of this prompt in full, from the next line.\n\n"
+    #              "Please check that the following JSON is consistent with the following term sheet and make any modifications that are required to the JSON values (not keys) to make it consistent.\n"
+    #              "Please output only the JSON, modified if necessary:\n"
+    #              #"Please repeat the full JSON from 'ANSWER' in your current answer, or if there is no JSON in 'ANSWER', say so.\n\n"
+    #              f"JSON: {answer}\n\n"
+    #              f"TERM SHEET: {body.input}"
+    # )
+
     provider_answers = await get_provider(ai_provider, api_key).run([param])
     answer = provider_answers[0].answer if provider_answers else ""
 
diff --git a/hackathon/providers/base_provider.py b/hackathon/providers/base_provider.py
index b7a8e59..d2352e0 100644
--- a/hackathon/providers/base_provider.py
+++ b/hackathon/providers/base_provider.py
@@ -14,7 +14,7 @@
 
 import abc
 import asyncio
-from dataclasses import dataclass
+from dataclasses import dataclass, asdict
 from typing import Final, Optional
 
 
@@ -28,7 +28,9 @@ class ProviderParam:
     temperature: Optional[float] = None
     top_p: Optional[float] = None
     top_k: Optional[int] = None
-
+    question: Optional[str] = None
+    # def dict(self):
+    #     return {k: str(v) for k, v in asdict(self).items()}
 
 @dataclass
 class ProviderAnswer:
diff --git a/hackathon/providers/openai_provider.py b/hackathon/providers/openai_provider.py
index e1a1e12..af9318b 100644
--- a/hackathon/providers/openai_provider.py
+++ b/hackathon/providers/openai_provider.py
@@ -23,7 +23,7 @@
 
 
 class OpenAIProvider(BaseProvider):
-    REQUEST_TIMEOUT: Final[int] = 60
+    REQUEST_TIMEOUT: Final[int] = 120
 
     async def run(self, params: list[ProviderParam]) -> list[ProviderAnswer]:
         async with aiohttp.ClientSession() as session:
@@ -43,7 +43,10 @@ async def get_answer(self, param: ProviderParam) -> ProviderAnswer:
 
     @retry(reraise=True, stop=stop_after_attempt(BaseProvider.RETRY_ATTEMPT))
     async def _openai_create(self, param: ProviderParam):
-        question = self.build_question(prompt=param.prompt, context=param.context)
+        if not param.question:
+            question = self.build_question(prompt=param.prompt, context=param.context)
+        else:
+            question = param.question
         messages = [{"role": "user", "content": question}]
         return await openai.ChatCompletion.acreate(
             model=param.provider_model,