quantalogic 0.53.0__py3-none-any.whl → 0.55.0__py3-none-any.whl

quantalogic/__init__.py

@@ -1,6 +1,7 @@
 """QuantaLogic package initialization."""
 
 import warnings
+from importlib.metadata import version as get_version
 
 # Suppress specific warnings related to Pydantic's V2 configuration changes
 warnings.filterwarnings(
@@ -10,6 +11,12 @@ warnings.filterwarnings(
     message=".*config keys have changed in V2:.*|.*'fields' config key is removed in V2.*",
 )
 
+try:
+    __version__: str = get_version("quantalogic")
+except Exception as e:
+    __version__ = "unknown"
+    print(f"Unable to retrieve version: {e}")
+
 # Import public API
 from .agent import Agent  # noqa: E402
 from .console_print_events import console_print_events  # noqa: E402

quantalogic/flow/flow.py

@@ -7,11 +7,12 @@
 #     "pydantic>=2.0.0",            # Data validation and settings
 #     "anyio>=4.0.0",               # Async utilities
 #     "jinja2>=3.1.0",              # Templating engine
-#     "instructor[litellm]>=0.5.0"  # Structured LLM output with litellm integration
+#     "instructor"  # Structured LLM output with litellm integration
 # ]
 # ///
 
 import asyncio
+import inspect  # Added for accurate parameter detection
 from dataclasses import dataclass
 from enum import Enum
 from pathlib import Path
@@ -52,17 +53,24 @@ class WorkflowEvent:
 WorkflowObserver = Callable[[WorkflowEvent], None]
 
 
-# Define a class for sub-workflow nodes
+# Define a class for sub-workflow nodes with updated inputs handling
 class SubWorkflowNode:
-    def __init__(self, sub_workflow: "Workflow", inputs: Dict[str, str], output: str):
-        """Initialize a sub-workflow node."""
+    def __init__(self, sub_workflow: "Workflow", inputs: Dict[str, Any], output: str):
+        """Initialize a sub-workflow node with flexible inputs mapping."""
         self.sub_workflow = sub_workflow
-        self.inputs = inputs
+        self.inputs = inputs  # Maps sub_key to main_key, callable, or value
         self.output = output
 
-    async def __call__(self, engine: "WorkflowEngine", **kwargs):
-        """Execute the sub-workflow with the engine's context."""
-        sub_context = {sub_key: kwargs[main_key] for main_key, sub_key in self.inputs.items()}
+    async def __call__(self, engine: "WorkflowEngine"):
+        """Execute the sub-workflow with the engine's context using inputs mapping."""
+        sub_context = {}
+        for sub_key, mapping in self.inputs.items():
+            if callable(mapping):
+                sub_context[sub_key] = mapping(engine.context)
+            elif isinstance(mapping, str):
+                sub_context[sub_key] = engine.context.get(mapping)
+            else:
+                sub_context[sub_key] = mapping  # Direct value
         sub_engine = self.sub_workflow.build(parent_engine=engine)
         result = await sub_engine.run(sub_context)
         return result.get(self.output)
@@ -132,7 +140,22 @@ class WorkflowEngine:
                 )
                 break
 
-            inputs = {k: self.context[k] for k in self.workflow.node_inputs[current_node] if k in self.context}
+            # Prepare inputs with mappings
+            input_mappings = self.workflow.node_input_mappings.get(current_node, {})
+            inputs = {}
+            # Add all mapped inputs
+            for key, mapping in input_mappings.items():
+                if callable(mapping):
+                    inputs[key] = mapping(self.context)
+                elif isinstance(mapping, str):
+                    inputs[key] = self.context.get(mapping)
+                else:
+                    inputs[key] = mapping  # Direct value
+            # For parameters in node_inputs that are not mapped, get from context
+            for param in self.workflow.node_inputs[current_node]:
+                if param not in inputs:
+                    inputs[param] = self.context.get(param)
+
             result = None
             exception = None
 
@@ -149,7 +172,7 @@ class WorkflowEngine:
 
             try:
                 if isinstance(node_func, SubWorkflowNode):
-                    result = await node_func(self, **inputs)
+                    result = await node_func(self)  # Sub-workflow handles its own inputs
                     usage = None  # Sub-workflow usage is handled by its own nodes
                 else:
                     result = await node_func(**inputs)
@@ -222,8 +245,9 @@ class Workflow:
         self.node_inputs: Dict[str, List[str]] = {}
         self.node_outputs: Dict[str, Optional[str]] = {}
         self.transitions: Dict[str, List[Tuple[str, Optional[Callable]]]] = {}
+        self.node_input_mappings: Dict[str, Dict[str, Any]] = {}  # Store input mappings for nodes
         self.current_node = None
-        self._observers: List[WorkflowObserver] = []  # Store observers for later propagation
+        self._observers: List[WorkflowObserver] = []
         self._register_node(start_node)  # Register the start node without setting current_node
         self.current_node = start_node  # Set current_node explicitly after registration
 
@@ -236,9 +260,12 @@ class Workflow:
         self.node_inputs[name] = inputs
         self.node_outputs[name] = output
 
-    def node(self, name: str):
-        """Add a node to the workflow chain and set it as the current node."""
+    def node(self, name: str, inputs_mapping: Optional[Dict[str, Any]] = None):
+        """Add a node to the workflow chain with an optional inputs mapping."""
         self._register_node(name)
+        if inputs_mapping:
+            self.node_input_mappings[name] = inputs_mapping
+            logger.debug(f"Added inputs mapping for node {name}: {inputs_mapping}")
         self.current_node = name
         return self
 
@@ -270,6 +297,30 @@ class Workflow:
         self.current_node = next_node
         return self
 
+    def branch(self, branches: List[Tuple[str, Optional[Callable]]]) -> "Workflow":
+        """Add multiple conditional branches from the current node."""
+        if not self.current_node:
+            logger.warning("No current node set for branching")
+            return self
+        for next_node, condition in branches:
+            if next_node not in self.nodes:
+                self._register_node(next_node)
+            self.transitions.setdefault(self.current_node, []).append((next_node, condition))
+            logger.debug(f"Added branch from {self.current_node} to {next_node} with condition {condition}")
+        return self
+
+    def converge(self, convergence_node: str) -> "Workflow":
+        """Set a convergence point for all previous branches."""
+        if convergence_node not in self.nodes:
+            self._register_node(convergence_node)
+        # Find all leaf nodes (nodes with no outgoing transitions) and point them to convergence_node
+        for node in self.nodes:
+            if (node not in self.transitions or not self.transitions[node]) and node != convergence_node:
+                self.transitions.setdefault(node, []).append((convergence_node, None))
+                logger.debug(f"Added convergence from {node} to {convergence_node}")
+        self.current_node = convergence_node
+        return self
+
     def parallel(self, *nodes: str):
         """Add parallel nodes to execute concurrently."""
         if self.current_node:
@@ -285,13 +336,14 @@ class Workflow:
             logger.debug(f"Added observer to workflow: {observer}")
         return self  # Support chaining
 
-    def add_sub_workflow(self, name: str, sub_workflow: "Workflow", inputs: Dict[str, str], output: str):
-        """Add a sub-workflow as a node."""
+    def add_sub_workflow(self, name: str, sub_workflow: "Workflow", inputs: Dict[str, Any], output: str):
+        """Add a sub-workflow as a node with flexible inputs mapping."""
         sub_node = SubWorkflowNode(sub_workflow, inputs, output)
         self.nodes[name] = sub_node
-        self.node_inputs[name] = list(inputs.keys())
+        self.node_inputs[name] = []  # Inputs handled internally by SubWorkflowNode
         self.node_outputs[name] = output
         self.current_node = name
+        logger.debug(f"Added sub-workflow {name} with inputs {inputs} and output {output}")
         return self
 
     def build(self, parent_engine: Optional["WorkflowEngine"] = None) -> WorkflowEngine:
@@ -308,32 +360,37 @@ class Nodes:
     @classmethod
     def define(cls, output: Optional[str] = None):
         """Decorator for defining simple workflow nodes."""
-
         def decorator(func: Callable) -> Callable:
             async def wrapped_func(**kwargs):
                 try:
-                    result = await func(**kwargs)
+                    if asyncio.iscoroutinefunction(func):
+                        result = await func(**kwargs)
+                    else:
+                        result = func(**kwargs)
                     logger.debug(f"Node {func.__name__} executed with result: {result}")
                     return result
                 except Exception as e:
                     logger.error(f"Error in node {func.__name__}: {e}")
                     raise
 
-            inputs = list(func.__annotations__.keys())
+            # Get parameter names from function signature
+            sig = inspect.signature(func)
+            inputs = [param.name for param in sig.parameters.values()]
             logger.debug(f"Registering node {func.__name__} with inputs {inputs} and output {output}")
             cls.NODE_REGISTRY[func.__name__] = (wrapped_func, inputs, output)
             return wrapped_func
-
         return decorator
 
     @classmethod
     def validate_node(cls, output: str):
         """Decorator for nodes that validate inputs."""
-
         def decorator(func: Callable) -> Callable:
             async def wrapped_func(**kwargs):
                 try:
-                    result = await func(**kwargs)
+                    if asyncio.iscoroutinefunction(func):
+                        result = await func(**kwargs)
+                    else:
+                        result = func(**kwargs)
                     if not isinstance(result, str):
                         raise ValueError(f"Validation node {func.__name__} must return a string")
                     logger.info(f"Validation result from {func.__name__}: {result}")
@@ -342,11 +399,41 @@ class Nodes:
                     logger.error(f"Validation error in {func.__name__}: {e}")
                     raise
 
-            inputs = list(func.__annotations__.keys())
+            # Get parameter names from function signature
+            sig = inspect.signature(func)
+            inputs = [param.name for param in sig.parameters.values()]
             logger.debug(f"Registering node {func.__name__} with inputs {inputs} and output {output}")
             cls.NODE_REGISTRY[func.__name__] = (wrapped_func, inputs, output)
             return wrapped_func
+        return decorator
 
+    @classmethod
+    def transform_node(cls, output: str, transformer: Callable[[Any], Any]):
+        """Decorator for nodes that transform their inputs."""
+        def decorator(func: Callable) -> Callable:
+            async def wrapped_func(**kwargs):
+                try:
+                    # Apply transformer to the first input value
+                    input_key = list(kwargs.keys())[0] if kwargs else None
+                    if input_key:
+                        transformed_input = transformer(kwargs[input_key])
+                        kwargs[input_key] = transformed_input
+                    if asyncio.iscoroutinefunction(func):
+                        result = await func(**kwargs)
+                    else:
+                        result = func(**kwargs)
+                    logger.debug(f"Transformed node {func.__name__} executed with result: {result}")
+                    return result
+                except Exception as e:
+                    logger.error(f"Error in transform node {func.__name__}: {e}")
+                    raise
+
+            # Get parameter names from function signature
+            sig = inspect.signature(func)
+            inputs = [param.name for param in sig.parameters.values()]
+            logger.debug(f"Registering node {func.__name__} with inputs {inputs} and output {output}")
+            cls.NODE_REGISTRY[func.__name__] = (wrapped_func, inputs, output)
+            return wrapped_func
         return decorator
 
     @staticmethod
@@ -367,20 +454,19 @@ class Nodes:
             raise
 
     @staticmethod
-    def _render_prompt(template: str, prompt_file: Optional[str], context: Dict[str, Any]) -> str:
-        """Render a prompt from either a template string or an external file."""
-        if prompt_file:
-            return Nodes._load_prompt_from_file(prompt_file, context)
+    def _render_template(template: str, template_file: Optional[str], context: Dict[str, Any]) -> str:
+        """Render a Jinja2 template from either a string or an external file."""
+        if template_file:
+            return Nodes._load_prompt_from_file(template_file, context)
         try:
             return Template(template).render(**context)
         except Exception as e:
-            logger.error(f"Error rendering prompt template: {e}")
+            logger.error(f"Error rendering template: {e}")
             raise
 
     @classmethod
     def llm_node(
         cls,
-        model: str,
         system_prompt: str,
         output: str,
         prompt_template: str = "",
@@ -392,34 +478,52 @@ class Nodes:
         frequency_penalty: float = 0.0,
         **kwargs,
     ):
-        """Decorator for creating LLM nodes with plain text output, supporting external prompt files."""
-
+        """Decorator for creating LLM nodes with plain text output, supporting dynamic parameters via input mappings."""
         def decorator(func: Callable) -> Callable:
-            async def wrapped_func(**kwargs):
-                prompt = cls._render_prompt(prompt_template, prompt_file, kwargs)
+            async def wrapped_func(model: str, **func_kwargs):
+                # Extract parameters from func_kwargs if provided, else use defaults
+                system_prompt_to_use = func_kwargs.pop("system_prompt", system_prompt)
+                prompt_template_to_use = func_kwargs.pop("prompt_template", prompt_template)
+                prompt_file_to_use = func_kwargs.pop("prompt_file", prompt_file)
+                temperature_to_use = func_kwargs.pop("temperature", temperature)
+                max_tokens_to_use = func_kwargs.pop("max_tokens", max_tokens)
+                top_p_to_use = func_kwargs.pop("top_p", top_p)
+                presence_penalty_to_use = func_kwargs.pop("presence_penalty", presence_penalty)
+                frequency_penalty_to_use = func_kwargs.pop("frequency_penalty", frequency_penalty)
+
+                # Use only signature parameters for template rendering
+                sig = inspect.signature(func)
+                template_vars = {k: v for k, v in func_kwargs.items() if k in sig.parameters}
+                prompt = cls._render_template(prompt_template_to_use, prompt_file_to_use, template_vars)
                 messages = [
-                    {"role": "system", "content": system_prompt},
+                    {"role": "system", "content": system_prompt_to_use},
                     {"role": "user", "content": prompt},
                 ]
+                
+                # Log the model and a preview of the prompt
+                truncated_prompt = prompt[:200] + "..." if len(prompt) > 200 else prompt
+                logger.info(f"LLM node {func.__name__} using model: {model}")
+                logger.debug(f"System prompt: {system_prompt_to_use[:100]}...")
+                logger.debug(f"User prompt preview: {truncated_prompt}")
+                
                 try:
                     response = await acompletion(
                         model=model,
                         messages=messages,
-                        temperature=temperature,
-                        max_tokens=max_tokens,
-                        top_p=top_p,
-                        presence_penalty=presence_penalty,
-                        frequency_penalty=frequency_penalty,
+                        temperature=temperature_to_use,
+                        max_tokens=max_tokens_to_use,
+                        top_p=top_p_to_use,
+                        presence_penalty=presence_penalty_to_use,
+                        frequency_penalty=frequency_penalty_to_use,
                         drop_params=True,
                         **kwargs,
                     )
                     content = response.choices[0].message.content.strip()
-                    # Attach usage metadata to the function
                     wrapped_func.usage = {
                         "prompt_tokens": response.usage.prompt_tokens,
                         "completion_tokens": response.usage.completion_tokens,
                         "total_tokens": response.usage.total_tokens,
-                        "cost": getattr(response, "cost", None),  # Include cost if available
+                        "cost": getattr(response, "cost", None),
                     }
                     logger.debug(f"LLM output from {func.__name__}: {content[:50]}...")
                     return content
@@ -427,17 +531,17 @@ class Nodes:
                     logger.error(f"Error in LLM node {func.__name__}: {e}")
                     raise
 
-            inputs = list(func.__annotations__.keys())
+            # Get parameter names from function signature and add 'model'
+            sig = inspect.signature(func)
+            inputs = ['model'] + [param.name for param in sig.parameters.values()]
             logger.debug(f"Registering node {func.__name__} with inputs {inputs} and output {output}")
             cls.NODE_REGISTRY[func.__name__] = (wrapped_func, inputs, output)
             return wrapped_func
-
         return decorator
 
     @classmethod
     def structured_llm_node(
         cls,
-        model: str,
         system_prompt: str,
         output: str,
         response_model: Type[BaseModel],
@@ -450,7 +554,7 @@ class Nodes:
         frequency_penalty: float = 0.0,
         **kwargs,
     ):
-        """Decorator for creating LLM nodes with structured output using instructor, supporting external prompt files."""
+        """Decorator for creating LLM nodes with structured output, supporting dynamic parameters via input mappings."""
         try:
             client = instructor.from_litellm(acompletion)
         except ImportError:
@@ -458,32 +562,51 @@ class Nodes:
             raise ImportError("Instructor is required for structured_llm_node")
 
         def decorator(func: Callable) -> Callable:
-            async def wrapped_func(**kwargs):
-                prompt = cls._render_prompt(prompt_template, prompt_file, kwargs)
+            async def wrapped_func(model: str, **func_kwargs):
+                # Extract parameters from func_kwargs if provided, else use defaults
+                system_prompt_to_use = func_kwargs.pop("system_prompt", system_prompt)
+                prompt_template_to_use = func_kwargs.pop("prompt_template", prompt_template)
+                prompt_file_to_use = func_kwargs.pop("prompt_file", prompt_file)
+                temperature_to_use = func_kwargs.pop("temperature", temperature)
+                max_tokens_to_use = func_kwargs.pop("max_tokens", max_tokens)
+                top_p_to_use = func_kwargs.pop("top_p", top_p)
+                presence_penalty_to_use = func_kwargs.pop("presence_penalty", presence_penalty)
+                frequency_penalty_to_use = func_kwargs.pop("frequency_penalty", frequency_penalty)
+
+                # Use only signature parameters for template rendering
+                sig = inspect.signature(func)
+                template_vars = {k: v for k, v in func_kwargs.items() if k in sig.parameters}
+                prompt = cls._render_template(prompt_template_to_use, prompt_file_to_use, template_vars)
                 messages = [
-                    {"role": "system", "content": system_prompt},
+                    {"role": "system", "content": system_prompt_to_use},
                     {"role": "user", "content": prompt},
                 ]
+                
+                # Log the model and a preview of the prompt
+                truncated_prompt = prompt[:200] + "..." if len(prompt) > 200 else prompt
+                logger.info(f"Structured LLM node {func.__name__} using model: {model}")
+                logger.debug(f"System prompt: {system_prompt_to_use[:100]}...")
+                logger.debug(f"User prompt preview: {truncated_prompt}")
+                logger.debug(f"Expected response model: {response_model.__name__}")
+                
                 try:
-                    # Use instructor with completion to get both structured output and raw response
                     structured_response, raw_response = await client.chat.completions.create_with_completion(
                         model=model,
                         messages=messages,
                         response_model=response_model,
-                        temperature=temperature,
-                        max_tokens=max_tokens,
-                        top_p=top_p,
-                        presence_penalty=presence_penalty,
-                        frequency_penalty=frequency_penalty,
+                        temperature=temperature_to_use,
+                        max_tokens=max_tokens_to_use,
+                        top_p=top_p_to_use,
+                        presence_penalty=presence_penalty_to_use,
+                        frequency_penalty=frequency_penalty_to_use,
                         drop_params=True,
                         **kwargs,
                     )
-                    # Attach usage metadata to the function
                     wrapped_func.usage = {
                         "prompt_tokens": raw_response.usage.prompt_tokens,
                         "completion_tokens": raw_response.usage.completion_tokens,
                         "total_tokens": raw_response.usage.total_tokens,
-                        "cost": getattr(raw_response, "cost", None),  # Include cost if available
+                        "cost": getattr(raw_response, "cost", None),
                     }
                     logger.debug(f"Structured output from {func.__name__}: {structured_response}")
                     return structured_response
@@ -494,15 +617,60 @@ class Nodes:
                     logger.error(f"Error in structured LLM node {func.__name__}: {e}")
                     raise
 
-            inputs = list(func.__annotations__.keys())
+            # Get parameter names from function signature and add 'model'
+            sig = inspect.signature(func)
+            inputs = ['model'] + [param.name for param in sig.parameters.values()]
             logger.debug(f"Registering node {func.__name__} with inputs {inputs} and output {output}")
             cls.NODE_REGISTRY[func.__name__] = (wrapped_func, inputs, output)
             return wrapped_func
+        return decorator
+
+    @classmethod
+    def template_node(
+        cls,
+        output: str,
+        template: str = "",
+        template_file: Optional[str] = None,
+    ):
+        """Decorator for creating nodes that apply a Jinja2 template to inputs, supporting dynamic parameters."""
+        def decorator(func: Callable) -> Callable:
+            async def wrapped_func(**func_kwargs):
+                # Extract template parameters from func_kwargs if provided, else use defaults
+                template_to_use = func_kwargs.pop("template", template)
+                template_file_to_use = func_kwargs.pop("template_file", template_file)
+
+                # Use only signature parameters (excluding rendered_content) for template rendering
+                sig = inspect.signature(func)
+                expected_params = [p.name for p in sig.parameters.values() if p.name != 'rendered_content']
+                template_vars = {k: v for k, v in func_kwargs.items() if k in expected_params}
+                rendered_content = cls._render_template(template_to_use, template_file_to_use, template_vars)
+
+                # Filter func_kwargs for the function call
+                filtered_kwargs = {k: v for k, v in func_kwargs.items() if k in expected_params}
+
+                try:
+                    if asyncio.iscoroutinefunction(func):
+                        result = await func(rendered_content=rendered_content, **filtered_kwargs)
+                    else:
+                        result = func(rendered_content=rendered_content, **filtered_kwargs)
+                    logger.debug(f"Template node {func.__name__} rendered: {rendered_content[:50]}...")
+                    return result
+                except Exception as e:
+                    logger.error(f"Error in template node {func.__name__}: {e}")
+                    raise
 
+            # Get parameter names from function signature and add 'rendered_content' if not present
+            sig = inspect.signature(func)
+            inputs = [param.name for param in sig.parameters.values()]
+            if 'rendered_content' not in inputs:
+                inputs.insert(0, 'rendered_content')
+            logger.debug(f"Registering node {func.__name__} with inputs {inputs} and output {output}")
+            cls.NODE_REGISTRY[func.__name__] = (wrapped_func, inputs, output)
+            return wrapped_func
         return decorator
 
 
-# Example workflow with observer integration and updated structured node
+# Example workflow with observer integration, updated nodes, input mappings, and dynamic parameters
 async def example_workflow():
     # Define Pydantic model for structured output
     class OrderDetails(BaseModel):
@@ -534,7 +702,6 @@ async def example_workflow():
                 if usage.get("cost") is not None:
                     self.total_cost += usage["cost"]
                 self.node_usages[event.node_name] = usage
-            # Print summary at workflow completion
             if event.event_type == WorkflowEventType.WORKFLOW_COMPLETED:
                 print(f"Total prompt tokens: {self.total_prompt_tokens}")
                 print(f"Total completion tokens: {self.total_completion_tokens}")
@@ -548,15 +715,12 @@ async def example_workflow():
         return "Order validated" if order.get("items") else "Invalid order"
 
     @Nodes.structured_llm_node(
-        model="gemini/gemini-2.0-flash",
-        system_prompt="You are an inventory checker. Respond with a JSON object containing 'order_id', 'items', and 'in_stock' (boolean).",
+        system_prompt="You are an inventory checker. Respond with a JSON object containing 'order_id', 'items_in_stock', and 'items_out_of_stock'.",
         output="inventory_status",
         response_model=OrderDetails,
         prompt_template="Check if the following items are in stock: {{ items }}. Return the result in JSON format with 'order_id' set to '123'.",
     )
     async def check_inventory(items: List[str]) -> OrderDetails:
-        # This is a placeholder function that would normally call an LLM
-        # The actual implementation is handled by the structured_llm_node decorator
         return OrderDetails(order_id="123", items_in_stock=["item1"], items_out_of_stock=[])
 
     @Nodes.define(output="payment_status")
@@ -579,41 +743,61 @@ async def example_workflow():
     async def notify_customer_out_of_stock(inventory_status: OrderDetails) -> str:
         return "Customer notified of out-of-stock"
 
+    @Nodes.transform_node(output="transformed_items", transformer=lambda x: [item.upper() for item in x])
+    async def transform_items(items: List[str]) -> List[str]:
+        return items
+
+    @Nodes.template_node(
+        output="formatted_message",
+        template="Order contains: {{ items | join(', ') }}",
+    )
+    async def format_order_message(rendered_content: str, items: List[str]) -> str:
+        return rendered_content
+
     # Sub-workflow for payment and shipping
     payment_shipping_sub_wf = Workflow("process_payment").sequence("process_payment", "arrange_shipping")
 
     # Instantiate token usage observer
     token_observer = TokenUsageObserver()
 
-    # Main workflow incorporating the sub-workflow
+    # Main workflow with dynamic parameter overrides
     workflow = (
         Workflow("validate_order")
-        .add_observer(progress_monitor)  # Add progress observer
-        .add_observer(token_observer)  # Add token usage observer
+        .add_observer(progress_monitor)
+        .add_observer(token_observer)
+        .node("validate_order", inputs_mapping={"order": "customer_order"})
+        .node("transform_items")
+        .node("format_order_message", inputs_mapping={
+            "items": "items",
+            "template": "Custom order: {{ items | join(', ') }}"  # Dynamic override
+        })
+        .node("check_inventory", inputs_mapping={
+            "model": lambda ctx: "gemini/gemini-2.0-flash",
+            "items": "transformed_items",
+            "temperature": 0.5,  # Dynamic override
+            "max_tokens": 1000   # Dynamic override
+        })
         .add_sub_workflow(
-            "payment_shipping", payment_shipping_sub_wf, inputs={"order": "order"}, output="shipping_confirmation"
-        )
-        .sequence("validate_order", "check_inventory")
-        .then(
             "payment_shipping",
-            condition=lambda ctx: len(ctx.get("inventory_status").items_out_of_stock) == 0 if ctx.get("inventory_status") else False,
-        )
-        .then(
-            "notify_customer_out_of_stock",
-            condition=lambda ctx: len(ctx.get("inventory_status").items_out_of_stock) > 0 if ctx.get("inventory_status") else True,
+            payment_shipping_sub_wf,
+            inputs={"order": lambda ctx: {"items": ctx["items"]}},
+            output="shipping_confirmation"
         )
-        .parallel("update_order_status", "send_confirmation_email")
-        .node("update_order_status")
-        .node("send_confirmation_email")
-        .node("notify_customer_out_of_stock")
+        .branch([
+            ("payment_shipping", lambda ctx: len(ctx.get("inventory_status").items_out_of_stock) == 0 if ctx.get("inventory_status") else False),
+            ("notify_customer_out_of_stock", lambda ctx: len(ctx.get("inventory_status").items_out_of_stock) > 0 if ctx.get("inventory_status") else True)
+        ])
+        .converge("update_order_status")
+        .sequence("update_order_status", "send_confirmation_email")
     )
 
     # Execute workflow
-    initial_context = {"order": {"items": ["item1", "item2"]}, "items": ["item1", "item2"]}
+    initial_context = {"customer_order": {"items": ["item1", "item2"]}, "items": ["item1", "item2"]}
     engine = workflow.build()
     result = await engine.run(initial_context)
     logger.info(f"Workflow result: {result}")
 
 
 if __name__ == "__main__":
+    logger.info("Initializing Quantalogic Flow Package")
     asyncio.run(example_workflow())