docs fix for the swarmns api fix

multiple_image_processing.py

@@ -1,7 +1,4 @@
 from swarms import Agent
-from swarms.prompts.logistics import (
-    Quality_Control_Agent_Prompt,
-)
 
 
 # Image for analysis
@@ -12,8 +9,8 @@ quality_control_agent = Agent(
     agent_name="Quality Control Agent",
     agent_description="A quality control agent that analyzes images and provides a detailed report on the quality of the product in the image.",
     model_name="claude-3-5-sonnet-20240620",
-    system_prompt=Quality_Control_Agent_Prompt,
+    # system_prompt=Quality_Control_Agent_Prompt,
-    multi_modal=True,
+    # multi_modal=True,
     max_loops=1,
     output_type="str-all-except-first",
     summarize_multiple_images=True,
@@ -22,7 +19,7 @@ quality_control_agent = Agent(
 
 response = quality_control_agent.run(
     task="what is in the image?",
-    imgs=[factory_image, factory_image],
+    imgs=[factory_image, "burning_image.jpg"],
 )
 
 print(response)

swarms/structs/agent.py

@@ -985,7 +985,7 @@ class Agent:
 
             self.short_memory.add(role=self.user_name, content=task)
 
-            if self.plan_enabled or self.planning_prompt is not None:
+            if self.plan_enabled is True:
                 self.plan(task)
 
             # Set the loop count
@@ -1360,10 +1360,15 @@ class Agent:
             # Get the current conversation history
             history = self.short_memory.get_str()
 
+            plan_prompt = f"Create a comprehensive step-by-step plan to complete the following task: \n\n {task}"
+
             # Construct the planning prompt by combining history, planning prompt, and task
-            planning_prompt = (
+            if exists(self.planning_prompt):
-                f"{history}\n\n{self.planning_prompt}\n\nTask: {task}"
+                planning_prompt = f"{history}\n\n{self.planning_prompt}\n\nTask: {task}"
-            )
+            else:
+                planning_prompt = (
+                    f"{history}\n\n{plan_prompt}\n\nTask: {task}"
+                )
 
             # Generate the plan using the LLM
             plan = self.llm.run(task=planning_prompt, *args, **kwargs)
@@ -1371,9 +1376,6 @@ class Agent:
             # Store the generated plan in short-term memory
             self.short_memory.add(role=self.agent_name, content=plan)
 
-            logger.info(
-                f"Successfully created plan for task: {task[:50]}..."
-            )
             return None
 
         except Exception as error:
@@ -2501,6 +2503,7 @@ class Agent:
         task: Optional[Union[str, Any]] = None,
         img: Optional[str] = None,
         imgs: Optional[List[str]] = None,
+        correct_answer: Optional[str] = None,
         *args,
         **kwargs,
     ) -> Any:
@@ -2534,6 +2537,14 @@ class Agent:
                 output = self.run_multiple_images(
                     task=task, imgs=imgs, *args, **kwargs
                 )
+            elif exists(correct_answer):
+                output = self.continuous_run_with_answer(
+                    task=task,
+                    img=img,
+                    correct_answer=correct_answer,
+                    *args,
+                    **kwargs,
+                )
             else:
                 output = self._run(
                     task=task,
@@ -2909,7 +2920,7 @@ class Agent:
         self, task: str, imgs: List[str], *args, **kwargs
     ):
         """
-        Run the agent with multiple images.
+        Run the agent with multiple images using concurrent processing.
 
         Args:
             task (str): The task to be performed on each image.
@@ -2932,12 +2943,33 @@ class Agent:
         Raises:
             Exception: If an error occurs while processing any of the images.
         """
-        outputs = []
+        # Calculate number of workers as 95% of available CPU cores
-        for img in imgs:
+        cpu_count = os.cpu_count()
-            output = self.run(task=task, img=img, *args, **kwargs)
+        max_workers = max(1, int(cpu_count * 0.95))
-            outputs.append(output)
+
+        # Use ThreadPoolExecutor for concurrent processing
+        with ThreadPoolExecutor(max_workers=max_workers) as executor:
+            # Submit all image processing tasks
+            future_to_img = {
+                executor.submit(
+                    self.run, task=task, img=img, *args, **kwargs
+                ): img
+                for img in imgs
+            }
+
+            # Collect results in order
+            outputs = []
+            for future in future_to_img:
+                try:
+                    output = future.result()
+                    outputs.append(output)
+                except Exception as e:
+                    logger.error(f"Error processing image: {e}")
+                    outputs.append(
+                        None
+                    )  # or raise the exception based on your preference
 
-        # Combine the outputs into a single string
+        # Combine the outputs into a single string if summarization is enabled
         if self.summarize_multiple_images is True:
             output = "\n".join(outputs)
 
@@ -2959,3 +2991,58 @@ class Agent:
             outputs = self.run(task=prompt, *args, **kwargs)
 
         return outputs
+
+    def continuous_run_with_answer(
+        self,
+        task: str,
+        img: Optional[str] = None,
+        correct_answer: str = None,
+        max_attempts: int = 10,
+    ):
+        """
+        Run the agent with the task until the correct answer is provided.
+
+        Args:
+            task (str): The task to be performed
+            correct_answer (str): The correct answer that must be found in the response
+            max_attempts (int): Maximum number of attempts before giving up (default: 10)
+
+        Returns:
+            str: The response containing the correct answer
+
+        Raises:
+            Exception: If max_attempts is reached without finding the correct answer
+        """
+        attempts = 0
+
+        while attempts < max_attempts:
+            attempts += 1
+
+            if self.verbose:
+                logger.info(
+                    f"Attempt {attempts}/{max_attempts} to find correct answer"
+                )
+
+            response = self._run(task=task, img=img)
+
+            # Check if the correct answer is in the response (case-insensitive)
+            if correct_answer.lower() in response.lower():
+                if self.verbose:
+                    logger.info(
+                        f"Correct answer found on attempt {attempts}"
+                    )
+                return response
+            else:
+                # Add feedback to help guide the agent
+                feedback = "Your previous response was incorrect. Think carefully about the question and ensure your response directly addresses what was asked."
+                self.short_memory.add(role="User", content=feedback)
+
+                if self.verbose:
+                    logger.info(
+                        f"Correct answer not found. Expected: '{correct_answer}'"
+                    )
+
+        # If we reach here, we've exceeded max_attempts
+        raise Exception(
+            f"Failed to find correct answer '{correct_answer}' after {max_attempts} attempts"
+        )

vision_tools.py

@@ -0,0 +1,68 @@
+from swarms.structs import Agent
+from swarms.prompts.logistics import (
+    Quality_Control_Agent_Prompt,
+)
+
+
+# Image for analysis
+factory_image = "image.jpg"
+
+
+def security_analysis(danger_level: str) -> str:
+    """
+    Analyzes the security danger level and returns an appropriate response.
+
+    Args:
+        danger_level (str, optional): The level of danger to analyze.
+            Can be "low", "medium", "high", or None. Defaults to None.
+
+    Returns:
+        str: A string describing the danger level assessment.
+            - "No danger level provided" if danger_level is None
+            - "No danger" if danger_level is "low"
+            - "Medium danger" if danger_level is "medium"
+            - "High danger" if danger_level is "high"
+            - "Unknown danger level" for any other value
+    """
+    if danger_level is None:
+        return "No danger level provided"
+
+    if danger_level == "low":
+        return "No danger"
+
+    if danger_level == "medium":
+        return "Medium danger"
+
+    if danger_level == "high":
+        return "High danger"
+
+    return "Unknown danger level"
+
+
+custom_system_prompt = f"""
+{Quality_Control_Agent_Prompt}
+
+You have access to tools that can help you with your analysis. When you need to perform a security analysis, you MUST use the security_analysis function with an appropriate danger level (low, medium, or high) based on your observations.
+
+Always use the available tools when they are relevant to the task. If you determine there is any level of danger or security concern, call the security_analysis function with the appropriate danger level.
+"""
+
+# Quality control agent
+quality_control_agent = Agent(
+    agent_name="Quality Control Agent",
+    agent_description="A quality control agent that analyzes images and provides a detailed report on the quality of the product in the image.",
+    # model_name="anthropic/claude-3-opus-20240229",
+    model_name="gpt-4o-mini",
+    system_prompt=custom_system_prompt,
+    multi_modal=True,
+    max_loops=1,
+    output_type="str-all-except-first",
+    # tools_list_dictionary=[schema],
+    tools=[security_analysis],
+)
+
+
+response = quality_control_agent.run(
+    task="Analyze the image and then perform a security analysis. Based on what you see in the image, determine if there is a low, medium, or high danger level and call the security_analysis function with that danger level",
+    img=factory_image,
+)