change max tokens for answers

app.py

@@ -72,7 +72,7 @@ vectorstore_graphs = get_azure_search_vectorstore(embeddings=embeddings_function
 vectorstore_region = get_azure_search_vectorstore(embeddings=embeddings_function, index_name="climateqa-v2")
 
 
-llm = get_llm(provider="azure", max_tokens=1024, temperature=0.0, streaming=True)
+llm = get_llm(provider="azure", max_tokens=3000, temperature=0.0, streaming=True)
 if os.environ["GRADIO_ENV"] == "local":
     reranker = get_reranker("nano")
 else:

climateqa/engine/llm/azure.py

@@ -109,9 +109,19 @@ def get_llm(
     model_kwargs = kwargs.pop("model_kwargs", {})
     model_kwargs["max_completion_tokens"] = max_tokens
     
-    # GPT-5/o1 models only support temperature=1, so we force it to 1
-    # (ignoring any temperature value passed by the caller)
-    fixed_temperature = 1.0
+    # Some models only support temperature=1:
+    # - o1 reasoning models (o1, o1-mini, o1-preview)
+    # - GPT-5 models (gpt-5, gpt-5-mini, etc.)
+    deployment_lower = deployment_name.lower()
+    is_o1_model = "o1" in deployment_lower and "gpt" not in deployment_lower
+    is_gpt5_model = "gpt-5" in deployment_lower or "gpt5" in deployment_lower
+    requires_fixed_temperature = is_o1_model or is_gpt5_model
+    
+    if requires_fixed_temperature:
+        print(f"Using fixed temperature=1.0 for model: {deployment_name} (model only supports temperature=1)")
+        effective_temperature = 1.0
+    else:
+        effective_temperature = temperature
     
     # Use Service Principal authentication if credentials are available
     if all([client_id, client_credential, tenant_id]):
@@ -121,7 +131,7 @@ def get_llm(
             azure_deployment=deployment_name,
             api_version=AZURE_OPENAI_API_VERSION,
             azure_ad_token_provider=_get_azure_ad_token,
-            temperature=fixed_temperature,
+            temperature=effective_temperature,
             verbose=verbose,
             streaming=streaming,
             timeout=timeout,
@@ -142,7 +152,7 @@ def get_llm(
             azure_deployment=deployment_name,
             api_version=AZURE_OPENAI_API_VERSION,
             api_key=api_key,
-            temperature=fixed_temperature,
+            temperature=effective_temperature,
             verbose=verbose,
             streaming=streaming,
             timeout=timeout,