Updates and fixes

Author: cxbxmxcx

logs/logs.txt

@@ -1,3 +1,11 @@
 Running on local URL:  http://127.0.0.1:7860
 
 To create a public link, set `share=True` in `launch()`.
+action: create_image_with_fal -> {'name': 'create_image_with_fal', 'group': 'fal_actions', 'pointer': <function create_image_with_fal at 0x0000020270B72DD0>, 'agent_action': {'type': 'function', 'function': {'name': 'create_image_with_fal', 'description': '\n    Generates an image using the FAL model service and downloads it to a local directory.\n\n    Args:\n        prompt (str): The text prompt to generate the image.\n        model (str, optional): The FAL model identifier to use for image generation. Default is "fal-ai/flux-pro".\n        image_size (str, optional): The desired image size. Default is "landscape_4_3".\n        num_inference_steps (int, optional): The number of inference steps for image generation. Default is 28.\n        guidance_scale (float, optional): The guidance scale for image generation. Default is 3.5.\n        seed: The seed used to as the randomized input into the model\n\n    Returns:\n        str: The local filename of the downloaded image if successful.\n\n    ', 'parameters': {'type': 'object', 'properties': {'prompt': {'type': 'string', 'description': 'prompt'}, 'model': {'type': 'string', 'description': 'model'}, 'image_size': {'type': 'string', 'description': 'image_size'}, 'num_inference_steps': {'type': 'string', 'description': 'num_inference_steps'}, 'guidance_scale': {'type': 'string', 'description': 'guidance_scale'}, 'seed': {'type': 'string', 'description': 'seed'}}, 'required': ['prompt']}}}, 'prompt_template': None}
+action: create_image_with_fal -> {'prompt': 'A fantastical scene featuring a majestic castle with towers reaching up to the clouds. The castle is surrounded by a wide moat filled with strange, otherworldly creatures. A drawbridge spans the moat, leading to the open castle gates. Various peasants, farmers, merchants, and knights are walking through the gates and on the drawbridge, going about their daily activities. The scene is captured in a wide, low-angle shot with a cinematic camera, emphasizing the grandeur and scale of the castle and its surroundings.', 'num_inference_steps': '28', 'guidance_scale': '20.0', 'seed': '42'}
+https://v2.fal.media/files/01bcaa1c469c4ce388e804a7d65eb88b_e8fdda34d60c49a08d52e459e437f09e.jpg
+action: create_image_with_fal(args={"prompt": "A fantastical scene featuring a majestic castle with towers reaching up to the clouds. The castle is surrounded by a wide moat filled with strange, otherworldly creatures. A drawbridge spans the moat, leading to the open castle gates. Various peasants, farmers, merchants, and knights are walking through the gates and on the drawbridge, going about their daily activities. The scene is captured in a wide, low-angle shot with a cinematic camera, emphasizing the grandeur and scale of the castle and its surroundings.", "num_inference_steps": "28", "guidance_scale": "20.0", "seed": "42"}) -> A_fantastical_scene__fal-ai_flux-pro_20.0_28_304f0252.png
+action: create_image -> {'name': 'create_image', 'group': 'image_actions', 'pointer': <function create_image at 0x0000020270C6BD00>, 'agent_action': {'type': 'function', 'function': {'name': 'create_image', 'description': '\n    Generate an image based on the provided prompt using the DALL-E model.\n    Args:\n        prompt (str): The prompt used to generate the image.\n        model (str, optional): The model used to generate the image. Defaults to "dall-e-3".\n        size (str, optional): The size of the image in pixels. Defaults to "1024x1024". Options are 1024x1024, 1024x1792 or 1792x1024.\n        quality (str, optional): The quality of the generation. Defaults to "standard". Options are "standard" or "hd" for high definition.\n        n (int, optional): The number of generations. Defaults to 1. Always 1 for now.\n    Returns:\n        str: The path to the generated image file.\n    ', 'parameters': {'type': 'object', 'properties': {'prompt': {'type': 'string', 'description': 'prompt'}, 'model': {'type': 'string', 'description': 'model'}, 'size': {'type': 'string', 'description': 'size'}, 'quality': {'type': 'string', 'description': 'quality'}, 'n': {'type': 'string', 'description': 'n'}}, 'required': ['prompt']}}}, 'prompt_template': None}
+action: create_image -> {'prompt': 'A fantastical scene featuring a majestic castle with towers reaching up to the clouds. The castle is surrounded by a wide moat filled with strange, otherworldly creatures. A drawbridge spans the moat, leading to the open castle gates. Various peasants, farmers, merchants, and knights are walking through the gates and on the drawbridge, going about their daily activities. The scene is captured in a wide, low-angle shot with a cinematic camera, emphasizing the grandeur and scale of the castle and its surroundings.', 'size': '1024x1024'}
+action: create_image(args={"prompt": "A fantastical scene featuring a majestic castle with towers reaching up to the clouds. The castle is surrounded by a wide moat filled with strange, otherworldly creatures. A drawbridge spans the moat, leading to the open castle gates. Various peasants, farmers, merchants, and knights are walking through the gates and on the drawbridge, going about their daily activities. The scene is captured in a wide, low-angle shot with a cinematic camera, emphasizing the grandeur and scale of the castle and its surroundings.", "size": "1024x1024"}) -> Afantasticalscenefeaturingamajesticcastlewithtower.png
+assistant > 

main.py

@@ -34,7 +34,8 @@
 def create_new_thread():
     global current_thread
     current_thread = api.create_thread()
-    return "New thread created."
+    logger.reset_logs()
+    return []
 
 
 create_new_thread()
@@ -162,16 +163,20 @@ def stream_worker(assistant_id, thread_id, event_handler):
     # history[-1][1] = wrap_latex_with_markdown(history[-1][1])
     yield history
 
-    files = extract_file_paths(history[-1][1])
-    for file in files:
-        file_path = get_file_path(file)
-        if os.path.exists(file_path):
-            history.append((None, (file_path,)))
-        yield history
+    # files = extract_file_paths(history[-1][1])
+    # for file in files:
+    #     file_path = get_file_path(file)
+    #     if os.path.exists(file_path):
+    #         history.append((None, (file_path,)))
+    #     yield history
 
     # Final flush of images
     while len(eh.images) > 0:
-        history.append((None, (eh.images.pop(),)))
+        file = eh.images.pop()
+        file_path = get_file_path(file)
+        if os.path.exists(file_path):
+            history.append((None, file_path))
+            history.append((None, (file_path,)))
         yield history
 
     initial_thread.join()
@@ -297,7 +302,7 @@ def main_interface():
                                 "Create New Thread", size="lg"
                             )
 
-                new_thread_button.click(create_new_thread, [], None)
+                new_thread_button.click(create_new_thread, [], [chatbot])
 
                 chat_msg = chat_input.submit(
                     ask_assistant,

playground/assistant_actions/fal_actions.py

@@ -14,7 +14,7 @@ def create_image_with_fal(
     model="fal-ai/flux-pro",
     image_size="landscape_4_3",
     num_inference_steps=28,
-    guidance_scale=3.5,
+    guidance_scale=20.0,
     seed=222,
 ):
     """

playground/assistant_actions/image_actions.py

@@ -6,6 +6,7 @@
 
 from playground.actions_manager import agent_action
 from playground.global_values import GlobalValues
+import re
 
 load_dotenv()
 
@@ -37,7 +38,7 @@ def create_image(prompt, model="dall-e-3", size="1024x1024", quality="standard",
     # Download the image
     image_response = requests.get(image_url)
     if image_response.status_code == 200:
-        local_filename = f"{prompt.replace(' ', '_').replace(',','')[:50]}.png"
+        local_filename = re.sub(r"[^\w\-.]", "", prompt)[:50] + ".png"
         local_path = os.path.join(
             GlobalValues.ASSISTANTS_WORKING_FOLDER, local_filename
         )

playground/assistants_utils.py

@@ -112,6 +112,9 @@ def on_image_file_done(self, image_file) -> None:
         print(f"File saved as {image_file}")
         self._images += [image_file]
 
+    def on_image_file(self, image_file) -> None:
+        self._images += [image_file]
+
     def on_tool_call_created(self, tool_call):
         if tool_call.type == "code_interpreter":
             print("# >>> Code Interpreter", flush=True)
@@ -159,6 +162,8 @@ def handle_requires_action(
                                     for c in el.content:
                                         if hasattr(c, "image_file"):
                                             self.on_image_file_done(c.image_file)
+                        elif ".png" in output:
+                            self.on_image_file(output)
 
                         tool_outputs.append(
                             {"tool_call_id": tool.id, "output": str(output)}