add

agentgraph/extraction/graph_utilities/knowledge_graph_merger.py

@@ -70,7 +70,7 @@ class KnowledgeGraphMerger:
     Uses CrewAI and specialized agents to perform entity resolution and relationship consolidation.
     """
     
-    def __init__(self, model: str = "gpt-4o-mini"):
+    def __init__(self, model: str = "gpt-5-mini"):
         """
         Initialize the knowledge graph merger.
         

agentgraph/input/text_processing/chunking_service.py

@@ -30,7 +30,7 @@ class ChunkingService:
     trace content with optimal parameters.
     """
     
-    def __init__(self, default_batch_size: int = 3, default_model: str = "gpt-4o-mini"):
+    def __init__(self, default_batch_size: int = 3, default_model: str = "gpt-5-mini"):
         """
         Initialize the chunking service with default parameters.
         
@@ -452,4 +452,4 @@ class ChunkingService:
         from .trace_line_processor import TraceLineNumberProcessor
         processor = TraceLineNumberProcessor()
         numbered_content, _ = processor.add_line_numbers(chunk_content, start_line=start_line)
-        return numbered_content 
+        return numbered_content 

agentgraph/methods/baseline/clustering_method.py

@@ -49,7 +49,7 @@ logging.getLogger("chromadb").setLevel(logging.WARNING)
 verbose_level = 0
 
 # Set environment variables
-os.environ["OPENAI_MODEL_NAME"] = "gpt-4o-mini"
+os.environ["OPENAI_MODEL_NAME"] = "gpt-5-mini"
 
 
 class ClusteringKnowledgeExtractionMethod(BaseKnowledgeExtractionMethod):
@@ -279,7 +279,9 @@ class ClusteringKnowledgeExtractionMethod(BaseKnowledgeExtractionMethod):
 
         # Normalize model name to match pricing keys
         model_key = model_name.lower()
-        if "gpt-4o-mini" in model_key:
+        if "gpt-5-mini" in model_key:
+            model_key = "gpt-5-mini"
+        elif "gpt-4o-mini" in model_key:
             model_key = "gpt-4o-mini"
         elif "gpt-4o" in model_key:
             model_key = "gpt-4o"
@@ -301,8 +303,8 @@ class ClusteringKnowledgeExtractionMethod(BaseKnowledgeExtractionMethod):
             model_key = "claude-3-haiku"
 
         if model_key not in pricing:
-            # Default to gpt-4o-mini pricing if model not found
-            model_key = "gpt-4o-mini"
+            # Default to gpt-5-mini pricing if model not found
+            model_key = "gpt-5-mini"
 
         rates = pricing[model_key]
 

agentgraph/methods/baseline/direct_llm_method.py

@@ -44,7 +44,7 @@ logging.getLogger("httpx").setLevel(logging.WARNING)
 class DirectLLMKnowledgeExtractor(BaseKnowledgeExtractionMethod):
     """Direct LLM knowledge extraction method using OpenAI API with structured output."""
 
-    def __init__(self, model: str = "gpt-4o-mini", **kwargs):
+    def __init__(self, model: str = "gpt-5-mini", **kwargs):
         super().__init__("direct_llm_method", **kwargs)
         self.client = OpenAI()
         self.model = model

agentgraph/methods/baseline/hybrid_method.py

@@ -47,7 +47,7 @@ logging.getLogger("chromadb").setLevel(logging.WARNING)
 verbose_level = 0
 
 # Set environment variables
-os.environ["OPENAI_MODEL_NAME"] = "gpt-4o-mini"
+os.environ["OPENAI_MODEL_NAME"] = "gpt-5-mini"
 
 class HybridKnowledgeExtractionMethod(BaseKnowledgeExtractionMethod):
     """Hybrid 2-task knowledge extraction method using CrewAI."""
@@ -286,4 +286,4 @@ class HybridKnowledgeExtractionMethod(BaseKnowledgeExtractionMethod):
             logger.error(f"trace_data type: {type(trace_data)}")
             if isinstance(trace_data, str):
                 logger.error(f"trace_data content (first 200 chars): {repr(trace_data[:200])}")
-            return {"entities": [], "relations": []} 
+            return {"entities": [], "relations": []} 

agentgraph/methods/baseline/openai_agent.py

@@ -350,7 +350,7 @@ class OpenAIAgentKnowledgeExtractor(BaseKnowledgeExtractionMethod):
     2. Validation Agent -> Validate generated graph only
     """
     
-    def __init__(self, model: str = "gpt-4o-mini", **kwargs):
+    def __init__(self, model: str = "gpt-5-mini", **kwargs):
         super().__init__("openai_agent_method", **kwargs)
         self.model = model
         

agentgraph/methods/baseline/original_method.py

@@ -46,7 +46,7 @@ logging.getLogger("chromadb").setLevel(logging.WARNING)
 verbose_level = 0
 
 # Set environment variables
-os.environ["OPENAI_MODEL_NAME"] = "gpt-4o-mini"
+os.environ["OPENAI_MODEL_NAME"] = "gpt-5-mini"
 
 class OriginalKnowledgeExtractionMethod(BaseKnowledgeExtractionMethod):
     """Original 3-task knowledge extraction method using CrewAI."""
@@ -186,4 +186,4 @@ class OriginalKnowledgeExtractionMethod(BaseKnowledgeExtractionMethod):
             return result.get("kg_data", {"entities": [], "relations": []})
         else:
             # Return empty knowledge graph on failure
-            return {"entities": [], "relations": []} 
+            return {"entities": [], "relations": []} 

agentgraph/methods/baseline/pydantic_method.py

@@ -46,7 +46,7 @@ logging.getLogger("httpx").setLevel(logging.WARNING)
 
 
 async def get_agent_graph_entities(trace_content: str, temperature: float = 0.0) -> AgentRunResult[List[Entity]]:
-    model = os.environ.get("OPENAI_MODEL_NAME", "gpt-4o-mini")
+    model = os.environ.get("OPENAI_MODEL_NAME", "gpt-5-mini")
 
     # Use shared prompt templates
     system_prompt = ENTITY_EXTRACTION_SYSTEM_PROMPT
@@ -65,7 +65,7 @@ async def get_agent_graph_entities(trace_content: str, temperature: float = 0.0)
 async def get_agent_graph_relations(
     trace_content: str, entities: Optional[List[Entity]] = None, temperature: float = 0
 ) -> AgentRunResult[List[Relation]]:
-    model = os.environ.get("OPENAI_MODEL_NAME", "gpt-4o-mini")
+    model = os.environ.get("OPENAI_MODEL_NAME", "gpt-5-mini")
     
     # Use shared prompt templates
     system_prompt = RELATION_EXTRACTION_SYSTEM_PROMPT
@@ -118,7 +118,7 @@ def validate_knowledge_graph(kg: KnowledgeGraph) -> KnowledgeGraph:
 
 
 async def build_agent_graph(entities: List[Entity], relations: List[Relation], temperature: float = 0.0) -> AgentRunResult[KnowledgeGraph]:
-    model = os.environ.get("OPENAI_MODEL_NAME", "gpt-4o-mini")
+    model = os.environ.get("OPENAI_MODEL_NAME", "gpt-5-mini")
     
     # Use shared prompt templates
     system_prompt = GRAPH_BUILDER_SYSTEM_PROMPT
@@ -141,7 +141,7 @@ async def build_agent_graph(entities: List[Entity], relations: List[Relation], t
 # Hybrid method functions
 async def get_hybrid_extraction(trace_content: str, temperature: float = 0.0) -> AgentRunResult[str]:
     """First stage of hybrid method: combined entity and relation extraction (text output)."""
-    model = os.environ.get("OPENAI_MODEL_NAME", "gpt-4o-mini")
+    model = os.environ.get("OPENAI_MODEL_NAME", "gpt-5-mini")
     
     role = "Knowledge Extraction Specialist"
     goal = "Extract comprehensive entities and relationships from agent system data efficiently"
@@ -177,7 +177,7 @@ Your goal is: {goal}
 
 async def get_hybrid_validation(extraction_text: str, temperature: float = 0.0) -> AgentRunResult[KnowledgeGraph]:
     """Second stage of hybrid method: validation and enhancement (matches original)."""
-    model = os.environ.get("OPENAI_MODEL_NAME", "gpt-4o-mini")
+    model = os.environ.get("OPENAI_MODEL_NAME", "gpt-5-mini")
     
     role = "Knowledge Graph Validator and Enhancer"
     goal = "Validate, enhance, and structure extracted knowledge into a comprehensive knowledge graph"
@@ -306,7 +306,7 @@ async def get_agent_graph(trace_content: str, sequential: bool = False, hybrid:
 class PydanticKnowledgeExtractor(BaseKnowledgeExtractionMethod):
     """Direct LLM knowledge extraction method using pydantic_ai with structured output."""
     
-    def __init__(self, model: str = "gpt-4o-mini", sequential: bool = False, hybrid: bool = False, temperature: float = 0.0, **kwargs):
+    def __init__(self, model: str = "gpt-5-mini", sequential: bool = False, hybrid: bool = False, temperature: float = 0.0, **kwargs):
         method_name = "pydantic_ai_method"
         if hybrid:
             method_name = "pydantic_hybrid_method"

agentgraph/methods/baseline/unified_method.py

@@ -47,7 +47,7 @@ logging.getLogger("chromadb").setLevel(logging.WARNING)
 verbose_level = 0
 
 # Set environment variables
-os.environ["OPENAI_MODEL_NAME"] = "gpt-4o-mini"
+os.environ["OPENAI_MODEL_NAME"] = "gpt-5-mini"
 
 class UnifiedKnowledgeExtractionMethod(BaseKnowledgeExtractionMethod):
     """Unified 1-task knowledge extraction method using CrewAI."""
@@ -228,4 +228,4 @@ class UnifiedKnowledgeExtractionMethod(BaseKnowledgeExtractionMethod):
             logger.error(f"trace_data type: {type(trace_data)}")
             if isinstance(trace_data, str):
                 logger.error(f"trace_data content (first 200 chars): {repr(trace_data[:200])}")
-            return {"entities": [], "relations": []} 
+            return {"entities": [], "relations": []} 

agentgraph/methods/cristian/knowledge_graph_agents.py

@@ -412,7 +412,7 @@ Agent(
     role='SQL Query Generator',
     goal='Generate accurate Databricks SQL queries based on business requirements',
     backstory='You are an expert SQL developer specializing in Databricks SQL Warehouse. You understand complex business logic and can translate natural language requirements into efficient SQL queries.',
-    llm='gpt-4o-mini'
+    llm='gpt-5-mini'
 )
 ```
 

agentgraph/methods/experimental/enhanced_adaptive_extractor.py

@@ -42,7 +42,7 @@ logging.basicConfig(level=logging.INFO)
 logger = logging.getLogger(__name__)
 
 # Environment setup
-os.environ["OPENAI_MODEL_NAME"] = "gpt-4o-mini"
+os.environ["OPENAI_MODEL_NAME"] = "gpt-5-mini"
 
 class ContentSection(BaseModel):
     """Represents a section of content for analysis"""
@@ -691,4 +691,4 @@ def create_enhanced_adaptive_crew() -> Crew:
         tasks=[],  # Tasks are created dynamically
         verbose=True,
         process=Process.sequential
-    )
+    )

agentgraph/methods/production/multi_agent_knowledge_extractor.py

@@ -350,14 +350,14 @@ def format_context_documents(context_documents=None):
     return "\n".join(formatted_docs)
 
 
-def extract_knowledge_graph_with_context(input_data, context_documents=None, model="gpt-4o-mini"):
+def extract_knowledge_graph_with_context(input_data, context_documents=None, model="gpt-5-mini"):
     """
     Extract knowledge graph using the multi-agent crew with optional context documents.
     
     Args:
         input_data: The trace data to analyze
         context_documents: Optional list of context documents to enhance extraction
-        model: The OpenAI model name to use for extraction (default: gpt-4o-mini)
+        model: The OpenAI model name to use for extraction (default: gpt-5-mini)
         
     Returns:
         Knowledge graph extraction result
@@ -390,17 +390,17 @@ def extract_knowledge_graph(input_data):
     Returns:
         Knowledge graph extraction result
     """
-    return extract_knowledge_graph_with_context(input_data, context_documents=None, model="gpt-4o-mini")
+    return extract_knowledge_graph_with_context(input_data, context_documents=None, model="gpt-5-mini")
 
 
 # Create a default crew instance for async methods (can be updated by setting model)
-agent_monitoring_crew = create_agent_monitoring_crew("gpt-4o-mini")
+agent_monitoring_crew = create_agent_monitoring_crew("gpt-5-mini")
 
 
 class AgentMonitoringCrewFactory:
     """Factory class to create agent monitoring crews with dynamic models."""
     
-    def __init__(self, model: str = "gpt-4o-mini"):
+    def __init__(self, model: str = "gpt-5-mini"):
         self.model = model
         self._crew = None
     
@@ -429,4 +429,4 @@ class AgentMonitoringCrewFactory:
 
 
 # This will be imported by the extraction factory
-agent_monitoring_crew_factory = AgentMonitoringCrewFactory()
+agent_monitoring_crew_factory = AgentMonitoringCrewFactory()

agentgraph/methods/production/pydantic_multi_agent_knowledge_extractor.py

@@ -163,7 +163,7 @@ def _ensure_event_loop():
 def extract_knowledge_graph_with_context(
     input_data: str,
     context_documents: Optional[List[Dict[str, Any]]] = None,
-    model: str = "gpt-4o-mini",
+    model: str = "gpt-5-mini",
 ):
     """Replicates `extract_knowledge_graph_with_context` but using Pydantic-AI.
 
@@ -220,7 +220,7 @@ def extract_knowledge_graph(input_data: str):
     return extract_knowledge_graph_with_context(
         input_data=input_data,
         context_documents=None,
-        model="gpt-4o-mini",
+        model="gpt-5-mini",
     )
 
 # ---------------------------------------------------------------------------
@@ -230,7 +230,7 @@ def extract_knowledge_graph(input_data: str):
 class PydanticAICrewFactory:
     """Factory that mimics the AgentMonitoringCrewFactory interface but powered by Pydantic-AI."""
 
-    def __init__(self, model: str = "gpt-4o-mini"):
+    def __init__(self, model: str = "gpt-5-mini"):
         self.model = model
 
     def set_model(self, model: str):
@@ -267,4 +267,4 @@ PydanticAgentMonitoringFactory = PydanticAICrewFactory
 
 # Simple progress printing helper
 def _log(step: str):
-    print(f"[PydanticAI] {step}") 
+    print(f"[PydanticAI] {step}") 

agentgraph/methods/production/task_prompts.py

@@ -421,7 +421,7 @@ ENTITY_EXTRACTION_INSTRUCTION_PROMPT = """
         role='SQL Query Generator',
         goal='Generate accurate Databricks SQL queries based on business requirements',
         backstory='You are an expert SQL developer specializing in Databricks SQL Warehouse. You understand complex business logic and can translate natural language requirements into efficient SQL queries.',
-        llm='gpt-4o-mini'
+        llm='gpt-5-mini'
     )
     ```
     

agentgraph/reconstruction/example_rag_usage.py

@@ -60,7 +60,7 @@ def example_usage():
     reconstructed_relations = reconstruct_prompts_from_knowledge_graph_rag(
         knowledge_graph=knowledge_graph,
         original_trace=original_trace,
-        llm_config={"model": "gpt-4o-mini", "temperature": 0.1}
+        llm_config={"model": "gpt-5-mini", "temperature": 0.1}
     )
     
     for relation in reconstructed_relations:
@@ -76,7 +76,7 @@ def example_usage():
     reconstructor = RagPromptReconstructor(
         knowledge_graph=knowledge_graph,
         original_trace=original_trace,
-        llm_config={"model": "gpt-4o-mini", "temperature": 0.1}
+        llm_config={"model": "gpt-5-mini", "temperature": 0.1}
     )
     
     # Reconstruct a specific relation
@@ -97,4 +97,4 @@ def example_usage():
     print(f"Reconstruction metadata: {enriched_kg.get('reconstruction_metadata', {})}")
 
 if __name__ == "__main__":
-    example_usage()
+    example_usage()

agentgraph/reconstruction/rag_prompt_reconstructor.py

@@ -70,7 +70,7 @@ class RagPromptReconstructor:
             return LLM(**llm_config)
         
         return LLM(
-            model="gpt-4o-mini",
+            model="gpt-5-mini",
             temperature=0.1,
         )
     
@@ -423,4 +423,4 @@ def enrich_knowledge_graph_with_prompts_rag(knowledge_graph: Dict[str, Any],
         "original_trace_length": len(original_trace)
     }
     
-    return enhanced_kg
+    return enhanced_kg

agentgraph/testing/knowledge_graph_tester.py

@@ -81,9 +81,9 @@ def load_litellm_config(model: str = "gpt-5-mini", api_key: str = None):
 def run_knowledge_graph_tests(
     testing_data: Dict[str, Any],
     perturbation_types: List[str],
-    model: str = "gpt-4o-mini",
+    model: str = "gpt-5-mini",
     max_relations: int = None,
-    judge_model: str = "gpt-4o-mini",
+    judge_model: str = "gpt-5-mini",
     openai_api_key: str = None,
     progress_callback: Optional[Callable[[int, int, str], None]] = None,
     **kwargs
@@ -132,4 +132,4 @@ def run_knowledge_graph_tests(
 
 
 if __name__ == '__main__':
-    sys.exit(0)
+    sys.exit(0)

agentgraph/testing/perturbation_types/jailbreak.py

@@ -244,9 +244,9 @@ def test_relation_jailbreak(
 
 def run_jailbreak_tests(
     testing_data: Dict[str, Any],
-    model: str = "gpt-4o-mini",
+    model: str = "gpt-5-mini",
     max_relations: int = None,
-    judge_model: str = "gpt-4o-mini",
+    judge_model: str = "gpt-5-mini",
     openai_api_key: str = None,
     progress_callback: Optional[Callable[[int, int, str], None]] = None,
     **kwargs
@@ -351,4 +351,4 @@ def run_jailbreak_tests(
     
     logger.info(f"Jailbreak testing completed: {results['summary']['successful_tests']}/{results['summary']['total_tested']} successful")
     
-    return results 
+    return results 

backend/dependencies.py

@@ -56,7 +56,7 @@ def get_knowledge_graph_tester(kg_path: Optional[str] = None) -> Any:
             knowledge_graph_path = kg_path or DEFAULT_KNOWLEDGE_GRAPH
             knowledge_graph_tester = KnowledgeGraphTester(
                 knowledge_graph_path=knowledge_graph_path,
-                model="gpt-4o"  # Default model
+                model="gpt-5-mini"  # Default model
             )
             logger.info(f"Initialized KnowledgeGraphTester with {knowledge_graph_path}")
         except Exception as e:
@@ -90,4 +90,4 @@ def get_prompt_reconstructor() -> Any:
                 detail=f"Error initializing reconstructor: {str(e)}"
             )
 
-    return prompt_reconstructor 
+    return prompt_reconstructor 

backend/routers/traces.py

@@ -985,7 +985,7 @@ async def get_enhanced_trace_statistics(trace_id: str, db: Session = Depends(get
                 "input_cost_usd": cost_info.get("input_cost_usd", 0.0),
                 "output_cost_usd": cost_info.get("output_cost_usd", 0.0),
                 "avg_cost_per_call_usd": cost_info.get("avg_cost_per_call_usd", 0.0),
-                "model_used": cost_info.get("model_used", "gpt-4o-mini"),
+                "model_used": cost_info.get("model_used", "gpt-5-mini"),
                 "pricing_source": cost_info.get("pricing_source", "fallback"),
                 "cost_efficiency_tokens_per_dollar": cost_info.get("cost_efficiency_tokens_per_dollar", 0),
                 "model_metadata": cost_info.get("model_metadata"),
@@ -1072,7 +1072,7 @@ class ProcessTraceRequest(BaseModel):
     splitter_type: str = "agent_semantic"
     force_regenerate: bool = True
     method_name: str = "production"
-    model: str = "gpt-4o-mini"
+    model: str = "gpt-5-mini"
     chunking_config: Optional[ChunkingConfig] = None
 
 @router.post("/{trace_id}/process")
@@ -1133,4 +1133,4 @@ async def process_trace(
         raise HTTPException(
             status_code=500, 
             detail=f"Error starting trace processing: {str(e)}"
-        )
+        )

backend/services/cost_calculation_service.py

@@ -39,6 +39,21 @@ class CostCalculationService:
     def _get_fallback_pricing_data(self) -> Dict[str, Any]:
         """Return fallback pricing data if GitHub fetch fails."""
         return {
+            "gpt-5-mini": {
+                "input_cost_per_token": 0.00000015,
+                "output_cost_per_token": 0.0000006,
+                "max_tokens": 128000,
+                "max_input_tokens": 128000,
+                "max_output_tokens": 16384,
+                "litellm_provider": "openai",
+                "mode": "chat",
+                "supports_function_calling": True,
+                "supports_vision": True,
+                "supports_response_schema": True,
+                "supports_prompt_caching": False,
+                "supports_system_messages": True,
+                "supports_tool_choice": True
+            },
             "gpt-4o-mini": {
                 "input_cost_per_token": 0.00000015,
                 "output_cost_per_token": 0.0000006,
@@ -134,7 +149,7 @@ class CostCalculationService:
     def _normalize_model_name(self, model_name: str) -> str:
         """Normalize model name to match pricing keys."""
         if not model_name:
-            return "gpt-4o-mini"  # Default fallback
+            return "gpt-5-mini"  # Default fallback
         
         model_lower = model_name.lower()
         
@@ -142,7 +157,9 @@ class CostCalculationService:
         model_lower = re.sub(r'^(openai/|anthropic/|gpt-|claude-)', '', model_lower)
         
         # Handle GPT models
-        if "gpt-4o-mini" in model_lower:
+        if "gpt-5-mini" in model_lower:
+            return "gpt-5-mini"
+        elif "gpt-4o-mini" in model_lower:
             return "gpt-4o-mini"
         elif "gpt-4o" in model_lower:
             return "gpt-4o"
@@ -162,7 +179,7 @@ class CostCalculationService:
             return "claude-3-opus-20240229"
         
         # Default fallback
-        return "gpt-4o-mini"
+        return "gpt-5-mini"
     
     def calculate_cost(
         self, 
@@ -201,7 +218,7 @@ class CostCalculationService:
             # Fallback to default model if not found
             if not model_pricing:
                 fallback_data = self._get_fallback_pricing_data()
-                model_pricing = fallback_data.get(normalized_model, fallback_data["gpt-4o-mini"])
+                model_pricing = fallback_data.get(normalized_model, fallback_data["gpt-5-mini"])
             
             # Extract pricing information
             input_cost_per_token = model_pricing.get("input_cost_per_token", 0.00000015)
@@ -270,9 +287,9 @@ class CostCalculationService:
             total_completion_tokens = token_analytics.get("total_completion_tokens", 0)
             prompt_calls = prompt_analytics.get("prompt_calls_detected", 0)
             
-            # For now, assume gpt-4o-mini as default model since we don't store model info in trace
+            # For now, assume gpt-5-mini as default model since we don't store model info in trace
             # In future versions, this could be enhanced to detect model from trace content
-            default_model = "gpt-4o-mini"
+            default_model = "gpt-5-mini"
             
             cost_info = self.calculate_cost(default_model, total_prompt_tokens, total_completion_tokens)
             
@@ -295,4 +312,4 @@ class CostCalculationService:
             return {"error": str(e)}
 
 # Global instance
-cost_service = CostCalculationService() 
+cost_service = CostCalculationService() 

backend/services/processing_service.py

@@ -18,7 +18,7 @@ class PipelineError(Exception):
     """Exception raised for errors in the pipeline processing."""
     pass
 
-async def process_trace_task(trace_id: str, session: Session, task_id: str = None, splitter_type: str = "agent_semantic", force_regenerate: bool = False, method_name: str = "production", model: str = "gpt-4o-mini", chunking_config = None) -> Dict[str, Any]:
+async def process_trace_task(trace_id: str, session: Session, task_id: str = None, splitter_type: str = "agent_semantic", force_regenerate: bool = False, method_name: str = "production", model: str = "gpt-5-mini", chunking_config = None) -> Dict[str, Any]:
     """
     Process a single trace and return the merged knowledge graph with proper timeout handling.
     """
@@ -73,7 +73,7 @@ async def process_trace_task(trace_id: str, session: Session, task_id: str = Non
         
         chunking_service = ChunkingService(
             default_batch_size=3,
-            default_model="gpt-4o-mini"
+            default_model="gpt-5-mini"
         )
         
         check_timeout()
@@ -272,4 +272,4 @@ async def process_trace_task(trace_id: str, session: Session, task_id: str = Non
         except Exception as update_error:
             logger.error(f"Failed to update trace status: {update_error}")
             
-        raise PipelineError(error_message) 
+        raise PipelineError(error_message) 

backend/services/test_service.py

@@ -84,7 +84,7 @@ class TestService:
     def run_perturbation_test(
         tester,
         knowledge_graph: str,
-        model: str = "gpt-4o",
+        model: str = "gpt-5-mini",
         perturbation_type: str = "jailbreak",
         relation_type: str = "",
         max_jailbreaks: int = 5,
@@ -383,4 +383,4 @@ class TestService:
             return {
                 "exists": False,
                 "error": str(e)
-            } 
+            } 

backend/services/testing_service.py

@@ -407,7 +407,7 @@ async def perturb_knowledge_graph_task(kg_id: str, task_id: str) -> bool:
             test_results = run_knowledge_graph_tests(
                 testing_data=testing_data,
                 perturbation_types=["jailbreak", "counterfactual_bias"],
-                model="gpt-4o-mini",
+                model="gpt-5-mini",
                 progress_callback=progress_callback,
             )
             update_task_status(task_id, "RUNNING", "Tests completed, saving results", 80)
@@ -438,7 +438,7 @@ async def perturb_knowledge_graph_task(kg_id: str, task_id: str) -> bool:
                         test_result=relation_result,
                         perturbation_score=relation_result.get("perturbation_score"),
                         test_metadata={
-                            "model": "gpt-4o-mini",
+                            "model": "gpt-5-mini",
                             'test_timestamp': datetime.now(timezone.utc).isoformat(),
                         }
                     )
@@ -459,4 +459,4 @@ async def perturb_knowledge_graph_task(kg_id: str, task_id: str) -> bool:
         logger.error(error_message)
         logger.error(traceback.format_exc())
         update_task_status(task_id, "FAILED", error_message)
-        return False 
+        return False 

frontend/src/components/features/traces/TraceKnowledgeGraphView.tsx

@@ -437,7 +437,7 @@ export function TraceKnowledgeGraphView({
     chunkingConfig?: { min_chunk_size?: number; max_chunk_size?: number }
   ) => {
     const finalMethodName = methodName || "production";
-    const finalModel = model || "gpt-4o-mini";
+    const finalModel = model || "gpt-5-mini";
     console.log("TraceKnowledgeGraphView: Using method name:", finalMethodName);
     console.log("TraceKnowledgeGraphView: Using model:", finalModel);
     console.log(

frontend/src/components/features/traces/TraceOverviewSection.tsx

@@ -317,7 +317,7 @@ export function TraceOverviewSection({
                         {/* Left side: Model and Cost */}
                         <div className="flex-1">
                           <p className="text-lg font-bold text-green-800 mb-1">
-                            {enhancedStats?.cost?.model_used || "gpt-4o-mini"}
+                            {enhancedStats?.cost?.model_used || "gpt-5-mini"}
                           </p>
                           <p className="text-base font-semibold text-green-600">
                             Avg:{" "}

frontend/src/components/shared/FloatingActionWidget.tsx

@@ -268,7 +268,7 @@ export function FloatingActionWidget() {
         splitterType,
         true, // force_regenerate = true to allow generating new graphs even if existing ones exist
         finalMethodName,
-        modelName || "gpt-4o-mini",
+        modelName || "gpt-5-mini",
         chunkingConfig
       );
 

frontend/src/components/shared/modals/TraceDetailsModal.tsx

@@ -260,7 +260,7 @@ This action cannot be undone.`;
         splitterType,
         true,
         finalMethodName,
-        modelName || "gpt-4o-mini",
+        modelName || "gpt-5-mini",
         chunkingConfig
       );
 

frontend/src/lib/api.ts

@@ -180,7 +180,7 @@ export const api = {
       splitterType: string = "agent_semantic",
       forceRegenerate: boolean = true,
       methodName: string = "production",
-      model: string = "gpt-4o-mini",
+      model: string = "gpt-5-mini",
       chunkingConfig?: { min_chunk_size?: number; max_chunk_size?: number }
     ) => {
       const requestBody = {

frontend/src/lib/models.ts

@@ -117,7 +117,7 @@ export const getModelById = (modelId: string): ModelConfig | undefined => {
   return allModels.find((model) => model.id === modelId);
 };
 
-export const getDefaultModel = (): string => "gpt-4o-mini";
+export const getDefaultModel = (): string => "gpt-5-mini";
 
 export function getCurrentModelName(): string {
   try {

utils/config.py

@@ -8,7 +8,7 @@ load_dotenv(dotenv_path=env_path)
 
 # OpenAI Configuration
 OPENAI_API_KEY = os.getenv("OPENAI_API_KEY")
-OPENAI_MODEL_NAME = os.getenv("OPENAI_MODEL_NAME", "gpt-4o-mini")
+OPENAI_MODEL_NAME = os.getenv("OPENAI_MODEL_NAME", "gpt-5-mini")
 
 AZURE_API_KEY = os.getenv("AZURE_API_KEY")
 AZURE_API_BASE = os.getenv("AZURE_API_BASE")
@@ -44,4 +44,4 @@ def validate_config():
         print(f"Please set them in the .env file or as environment variables")
         return False
     
-    return True 
+    return True