feat: 🎸 generate random sound agent tool

.gitignore

@@ -1 +1,3 @@
-venv
+venv
+__pycache__
+.gradio

Gradio_UI.py

@@ -154,8 +154,10 @@ def stream_to_gradio(
             yield message
 
     final_answer = step_log  # Last log is the run's final_answer
-    final_answer = handle_agent_output_types(final_answer)
 
+    if not isinstance(final_answer, (AgentText, AgentImage, AgentAudio)):
+        final_answer = handle_agent_output_types(final_answer)
+    
     if isinstance(final_answer, AgentText):
         yield gr.ChatMessage(
             role="assistant",

app.py

@@ -1,62 +1,72 @@
-from smolagents import CodeAgent,DuckDuckGoSearchTool, HfApiModel,load_tool,tool
-import datetime
-import requests
-import pytz
+from smolagents import CodeAgent,HfApiModel,tool
+import torch
 import yaml
 from tools.final_answer import FinalAnswerTool
+import numpy as np
+import random
+from scipy.signal import square, sawtooth
 
 from Gradio_UI import GradioUI
 
-# Below is an example of a tool that does nothing. Amaze us with your creativity !
 @tool
-def my_custom_tool(arg1:str, arg2:int)-> str: #it's import to specify the return type
-    #Keep this format for the description / args / args description but feel free to modify the tool
-    """A tool that does nothing yet 
-    Args:
-        arg1: the first argument
-        arg2: the second argument
-    """
-    return "What magic will you build ?"
+def generate_random_sound(duration: float = 1.0) -> torch.Tensor:
+    """Generates a random sound with varying frequency and waveform and returns a torch tensor.
 
-@tool
-def get_current_time_in_timezone(timezone: str) -> str:
-    """A tool that fetches the current local time in a specified timezone.
     Args:
-        timezone: A string representing a valid timezone (e.g., 'America/New_York').
+        duration: Length of the sound in seconds (default: 1.0)
     """
-    try:
-        # Create timezone object
-        tz = pytz.timezone(timezone)
-        # Get current time in that timezone
-        local_time = datetime.datetime.now(tz).strftime("%Y-%m-%d %H:%M:%S")
-        return f"The current local time in {timezone} is: {local_time}"
-    except Exception as e:
-        return f"Error fetching time for timezone '{timezone}': {str(e)}"
+    # It seems the playback is about 3x slower, so we'll adjust the duration
+    adjusted_duration = duration / 3
+    
+    # Sample rate
+    sample_rate = 44100
+    num_samples = int(sample_rate * adjusted_duration)
+    t = np.linspace(0, adjusted_duration, num_samples, endpoint=False)
+
+    # Random frequency between 100 and 10000 Hz
+    frequency = random.uniform(100, 10000)
+
+    # Random waveform selection
+    waveform = random.choice(['sine', 'square', 'sawtooth', 'triangle'])
 
+    if waveform == 'sine':
+        signal = np.sin(2 * np.pi * frequency * t)
+    elif waveform == 'square':
+        signal = square(2 * np.pi * frequency * t)
+    elif waveform == 'sawtooth':
+        signal = sawtooth(2 * np.pi * frequency * t)
+    else:  # triangle
+        signal = sawtooth(2 * np.pi * frequency * t, width=0.5)  # Triangle is a symmetric sawtooth
+
+    # Apply fade out
+    fade = np.exp(-3 * t / adjusted_duration)  # Normalize fade over duration
+    signal = signal * fade
+
+    # Normalize to prevent clipping
+    signal = signal / np.max(np.abs(signal))
+
+    # Convert to torch tensor
+    tensor_signal = torch.from_numpy(signal.astype(np.float32))
+    
+    return tensor_signal
 
 final_answer = FinalAnswerTool()
-web_search = DuckDuckGoSearchTool()
 
-# If the agent does not answer, the model is overloaded, please use another model or the following Hugging Face Endpoint that also contains qwen2.5 coder:
-# model_id='https://pflgm2locj2t89co.us-east-1.aws.endpoints.huggingface.cloud' 
+model_id='https://pflgm2locj2t89co.us-east-1.aws.endpoints.huggingface.cloud' 
 
 model = HfApiModel(
 max_tokens=2096,
 temperature=0.5,
-model_id='Qwen/Qwen2.5-Coder-32B-Instruct',# it is possible that this model may be overloaded
+model_id=model_id,
 custom_role_conversions=None,
 )
 
-
-# Import tool from Hub
-image_generation_tool = load_tool("agents-course/text-to-image", trust_remote_code=True)
-
 with open("prompts.yaml", 'r') as stream:
     prompt_templates = yaml.safe_load(stream)
     
 agent = CodeAgent(
     model=model,
-    tools=[final_answer, web_search], ## add your tools here (don't remove final answer)
+    tools=[final_answer, generate_random_sound],
     max_steps=6,
     verbosity_level=1,
     grammar=None,

requirements.txt

@@ -3,3 +3,6 @@ smolagents
 requests
 duckduckgo_search
 pandas
+torch
+scipy
+smolagents[audio]