Add GPT-5 evals and "minimal" to reasoning dropdown

app/backend/approaches/chatreadretrieveread.py

@@ -162,7 +162,7 @@ async def run_search_approach(
                 ),  # Setting too low risks malformed JSON, setting too high may affect performance
                 temperature=0.0,  # Minimize creativity for search query generation
                 tools=tools,
-                reasoning_effort="low",  # Minimize reasoning for search query generation
+                reasoning_effort="minimal",  # Minimize reasoning for search query generation
             ),
         )
 
@@ -202,7 +202,7 @@ async def run_search_approach(
                     model=self.chatgpt_model,
                     deployment=self.chatgpt_deployment,
                     usage=chat_completion.usage,
-                    reasoning_effort="low",
+                    reasoning_effort="minimal",
                 ),
                 ThoughtStep(
                     "Search using generated search query",

app/backend/requirements.txt

@@ -212,7 +212,7 @@ multidict==6.0.5
     #   yarl
 oauthlib==3.2.2
     # via requests-oauthlib
-openai==1.63.0
+openai==1.99.8
     # via -r requirements.in
 opentelemetry-api==1.31.1
     # via

app/frontend/src/components/Settings/Settings.tsx

@@ -335,6 +335,7 @@ export const Settings = ({
                     }
                     aria-labelledby={reasoningEffortFieldId}
                     options={[
+                        { key: "minimal", text: t("labels.reasoningEffortOptions.minimal") },
                         { key: "low", text: t("labels.reasoningEffortOptions.low") },
                         { key: "medium", text: t("labels.reasoningEffortOptions.medium") },
                         { key: "high", text: t("labels.reasoningEffortOptions.high") }

app/frontend/src/locales/da/translation.json

@@ -89,6 +89,7 @@
         "useQueryRewriting": "Brug forespørgselsomskrivning til informationsgenfinding",
         "reasoningEffort": "Ræsonnementsindsats",
         "reasoningEffortOptions": {
+            "minimal": "Minimal (kun GPT-5)",
             "low": "Lav",
             "medium": "Medium",
             "high": "Høj"

app/frontend/src/locales/en/translation.json

@@ -99,6 +99,7 @@
         "useQueryRewriting": "Use query rewriting for retrieval",
         "reasoningEffort": "Reasoning effort",
         "reasoningEffortOptions": {
+            "minimal": "Minimal (GPT-5 only)",
             "low": "Low",
             "medium": "Medium",
             "high": "High"

app/frontend/src/locales/es/translation.json

@@ -93,6 +93,7 @@
         "useQueryRewriting": "Utiliza la reescritura de consultas para la recuperación",
         "reasoningEffort": "Esfuerzo de razonamiento",
         "reasoningEffortOptions": {
+            "minimal": "Mínimo (solo GPT-5)",
             "low": "Bajo",
             "medium": "Medio",
             "high": "Alto"

app/frontend/src/locales/fr/translation.json

@@ -95,6 +95,7 @@
         "useQueryRewriting": "Utilisez la réécriture des requêtes pour la récupération",
         "reasoningEffort": "Effort de raisonnement",
         "reasoningEffortOptions": {
+            "minimal": "Minimal (GPT-5 uniquement)",
             "low": "Faible",
             "medium": "Moyen",
             "high": "Élevé"

app/frontend/src/locales/it/translation.json

@@ -93,6 +93,7 @@
         "useQueryRewriting": "Usa la riscrittura delle query per il recupero",
         "reasoningEffort": "Sforzo di ragionamento",
         "reasoningEffortOptions": {
+            "minimal": "Minimo (solo GPT-5)",
             "low": "Basso",
             "medium": "Medio",
             "high": "Alto"

app/frontend/src/locales/ja/translation.json

@@ -93,6 +93,7 @@
         "useQueryRewriting": "検索のためにクエリの書き換えを使用する",
         "reasoningEffort": "推論の労力",
         "reasoningEffortOptions": {
+            "minimal": "最小（GPT-5 のみ）",
             "low": "低",
             "medium": "中",
             "high": "高"

app/frontend/src/locales/nl/translation.json

@@ -93,6 +93,7 @@
         "useQueryRewriting": "Gebruik de herformulering van zoekopdrachten om informatie op te halen",
         "reasoningEffort": "Redeneerinspanning",
         "reasoningEffortOptions": {
+            "minimal": "Minimaal (alleen GPT-5)",
             "low": "Laag",
             "medium": "Gemiddeld",
             "high": "Hoog"

app/frontend/src/locales/ptBR/translation.json

@@ -95,6 +95,7 @@
         "useQueryRewriting": "Utilize a reescrita de consultas para a recuperação",
         "reasoningEffort": "Esforço de raciocínio",
         "reasoningEffortOptions": {
+            "minimal": "Mínimo (apenas GPT-5)",
             "low": "Baixo",
             "medium": "Médio",
             "high": "Alto"

app/frontend/src/locales/tr/translation.json

@@ -93,6 +93,7 @@
         "useQueryRewriting": "Bilgi erişimi için sorgu yeniden yazımını kullanın",
         "reasoningEffort": "Akıl yürütme çabası",
         "reasoningEffortOptions": {
+            "minimal": "En az (yalnızca GPT-5)",
             "low": "Düşük",
             "medium": "Orta",
             "high": "Yüksek"

evals/evaluate_config.json

@@ -6,20 +6,24 @@
     "target_parameters": {
         "overrides": {
             "top": 3,
+            "max_subqueries": 10,
+            "results_merge_strategy": "interleaved",
             "temperature": 0.3,
             "minimum_reranker_score": 0,
             "minimum_search_score": 0,
             "retrieval_mode": "hybrid",
             "semantic_ranker": true,
             "semantic_captions": false,
+            "query_rewriting": false,
+            "reasoning_effort": "minimal",
             "suggest_followup_questions": false,
             "use_oid_security_filter": false,
             "use_groups_security_filter": false,
-            "vector_fields": [
-                "embedding"
-            ],
+            "vector_fields": "textEmbeddingOnly",
             "use_gpt4v": false,
             "gpt4v_input": "textAndImages",
+            "language": "en",
+            "use_agentic_retrieval": false,
             "seed": 1
         }
     },

evals/requirements.txt

@@ -3,4 +3,4 @@ rich
 ragas==0.2.13
 rapidfuzz==3.13.0
 langchain==0.3.17
-git+https://github.com/mattgotteiner/ai-rag-chat-evaluator@patch-1
+git+https://github.com/Azure-Samples/ai-rag-chat-evaluator@2025-06-02

evals/results/gpt5-emb3l/config.json

@@ -0,0 +1,32 @@
+{
+    "testdata_path": "ground_truth.jsonl",
+    "results_dir": "results/experiment<TIMESTAMP>",
+    "requested_metrics": ["gpt_groundedness", "gpt_relevance", "answer_length", "latency", "citations_matched", "any_citation"],
+    "target_url": "http://localhost:50505/chat",
+    "target_parameters": {
+        "overrides": {
+            "top": 3,
+            "max_subqueries": 10,
+            "results_merge_strategy": "interleaved",
+            "temperature": 0.3,
+            "minimum_reranker_score": 0,
+            "minimum_search_score": 0,
+            "retrieval_mode": "hybrid",
+            "semantic_ranker": true,
+            "semantic_captions": false,
+            "query_rewriting": false,
+            "reasoning_effort": "minimal",
+            "suggest_followup_questions": false,
+            "use_oid_security_filter": false,
+            "use_groups_security_filter": false,
+            "vector_fields": "textEmbeddingOnly",
+            "use_gpt4v": false,
+            "gpt4v_input": "textAndImages",
+            "language": "en",
+            "use_agentic_retrieval": false,
+            "seed": 1
+        }
+    },
+    "target_response_answer_jmespath": "message.content",
+    "target_response_context_jmespath": "context.data_points.text"
+}