Implement a class AgentOrchestrator that executes tools in a loop with error handling. Truncate conversation history if it exceeds 10 messages. Handle errors with fallbacks, and support defined step and failure limits.

The AgentOrchestrator class should encapsulate the orchestration logic. It maintains a list called history that stores the last 10 messages. The run method takes a list of tools and optional limits for steps and failures. Inside run, a loop iterates over the steps up to the step limit. For each step, it selects the next tool, calls its execute method inside a try/except block, and appends the result to history. If an exception occurs, a fallback tool (e.g., a generic "fallback" tool) is invoked. After each iteration, the history is trimmed to the last 10 entries. The loop stops if the failure count reaches the failure limit or if all steps are completed. Complexity is O(s) time where s is the number of steps, and O(h) space for the history (h ≤ 10). A minimal skeleton:

class AgentOrchestrator:
    def __init__(self, tools, step_limit=10, failure_limit=3):
        self.tools = tools
        self.step_limit = step_limit
        self.failure_limit = failure_limit
        self.history = []

    def run(self, initial_input):
        failures = 0
        for step in range(self.step_limit):
            tool = self.tools[step % len(self.tools)]
            try:
                result = tool.execute(initial_input)
                self.history.append(result)
                failures = 0
            except Exception as e:
                failures += 1
                self.history.append(f"Error: {e}")
                if failures >= self.failure_limit:
                    break
            self.history = self.history[-10:]
        return self.history

This design keeps the history bounded, handles errors gracefully, and respects the configured limits.