simplify launch

app.py

@@ -0,0 +1,237 @@
+import io
+import time
+import gradio as gr
+import base64
+import openai
+from openai import OpenAI
+
+def describe_food_in_image(image, api_key, vision_model, progress=gr.Progress()):
+    start_time = time.time()
+    progress(0, desc="Initializing image analysis...")
+    sn_dev_client = OpenAI(
+        base_url="https://api.sambanova.ai/v1",
+        api_key=api_key
+    )
+    
+    progress(0.2, desc="Converting image to base64...")
+    # Convert the image to base64
+    buffered = io.BytesIO()
+    image.save(buffered, format="PNG")
+    base64_image = base64.b64encode(buffered.getvalue()).decode('utf-8')
+    
+    progress(0.3, desc="Preparing API request...")
+    # Prepare the message with the image
+    messages = [
+        {
+            "role": "user",
+            "content": [
+                {
+                    "type": "text",
+                    "text": "What type of food is shown in this image? Please provide a brief description."
+                },
+                {
+                    "type": "image_url",
+                    "image_url": {
+                        "url": f"data:image/png;base64,{base64_image}"
+                    }
+                }
+            ]
+        }
+    ]
+
+    progress(0.4, desc="Analyzing image...")
+    # Make the API call
+    try:
+        response = sn_dev_client.chat.completions.create(
+            model=vision_model,
+            messages=messages,
+            max_tokens=150
+        )
+        progress(1.0, desc="Image analysis complete")
+        output = response.choices[0].message.content
+        # Return the model's description
+        thinking_time = time.time() - start_time
+        return output, thinking_time
+    except Exception as e:
+        error_message = f"Error: {str(e)}"
+        raise gr.Error(error_message)
+
+def desc_to_ingredients_recipe(food_description, api_key, language_model, progress=gr.Progress()):
+    # Run the code through the SN API (and profile)
+    start_time = time.time()
+    progress(0.6, desc="Initializing recipe generation...")
+    # Sambanova API base URL
+    SN_API_BASE = "https://api.sambanova.ai/v1"
+    sn_client = openai.OpenAI(api_key=api_key, base_url=SN_API_BASE)
+    progress(0.7, desc="Preparing recipe prompt...")
+    # Create the prompt 
+    budget = 10 # between 1-100
+    SYSTEM_PROMPT = f"""
+You are a helpful assistant in normal conversation.
+When given a problem to solve, you are an expert problem-solving assistant. 
+Your task is to provide a detailed, step-by-step solution to a given question with the following budget of {budget}. 
+Follow these instructions carefully:
+1. Read the given question carefully 
+2. Generate a detailed, logical step-by-step solution.
+3. Each step should be formatted as in the example below. 
+4. You are allowed to use at most {budget} steps (starting budget), 
+   keep track of it by counting down from the budget.  
+   STOP GENERATING MORE STEPS when hitting 0. 
+5. Do a self-reflection when you are unsure about how to proceed, 
+   based on the self-reflection and reward, decides whether you need to return 
+   to the previous steps.
+6. After completing the solution steps, reorganize and synthesize the steps 
+   into the final answer as in the format given below and do not mention the budget or steps you took.
+7. Provide a critical, honest and subjective self-evaluation of your reasoning 
+   process at the end. 
+
+Output should be in Markdown format, for example if generating a recipe it would look like this: 
+
+# Recipe
+Hi there! Today you'll learn about how to cook a delicious Thai delicacy, Pad Thai! For this you'll need
+
+## Ingredients 
+For the sauce:
+* 1/2 cup tamarind paste
+* 1/4 cup fish sauce
+* 1/4 cup palm sugar (or brown sugar)
+* 2 tablespoons soy sauce
+* 2 tablespoons rice vinegar
+* 1 tablespoon vegetable oil
+* 2 cloves garlic, minced
+* 1 tablespoon grated fresh ginger
+* 1/4 teaspoon ground white pepper
+* 2 tablespoons water
+* Salt, to taste
+* Fresh lime wedges, for serving
+
+For the noodles:
+* 1 cup rice stick noodles (preferably fresh)
+
+For the protein:
+* 1 cup mixed protein (shrimp, chicken, tofu, or a combination)
+
+For the vegetables:
+* 1 cup bean sprouts
+* 1 cup sliced carrots
+* 1 cup sliced red bell pepper
+* 2 green onions, thinly sliced
+* 1/4 cup chopped peanuts
+* 2 tablespoons chopped fresh cilantro (optional)
+
+## Instructions 
+1. Prepare the sauce: In a blender or food processor, combine tamarind paste, fish sauce, palm sugar, soy sauce, rice vinegar, garlic, ginger, and white pepper. Blend until smooth. Heat the sauce in a saucepan over medium heat, stirring constantly, until it thickens slightly. Remove from heat and stir in vegetable oil and water. Season with salt to taste.
+2. Cook the noodles: Soak the rice stick noodles in hot water for about 5-7 minutes, or according to package instructions. Drain and set aside.
+3. Prepare the protein: Cut the protein into bite-sized pieces and cook according to your preference (e.g., grill, sauté, or boil). Set aside.
+4. Sauté the vegetables: Heat 2 tablespoons of vegetable oil in a large wok or frying pan over medium-high heat. Add the bean sprouts, carrots, and red bell pepper. Cook, stirring constantly, until the vegetables are tender-crisp.
+5. Assemble the Pad Thai: Add the cooked noodles, protein, and sauce to the wok or frying pan. Stir-fry everything together for about 2-3 minutes, until the noodles are well coated with the sauce.
+6. Garnish and serve: Transfer the Pad Thai to a serving platter or individual plates. Sprinkle with green onions, peanuts, and cilantro (if using). Serve with fresh lime wedges on the side.
+
+Tips and Variations:
+* Use a variety of protein sources, such as shrimp, chicken, tofu, or a combination.
+* Add other vegetables, such as mushrooms, zucchini, or baby corn.
+* Use different types of noodles, such as rice vermicelli or egg noodles.
+* Adjust the level of spiciness to your liking by adding more or less chili flakes.
+* Serve with a sprinkle of toasted sesame seeds or chopped scallions for added flavor and texture.
+
+Try it out for yourself, taste it and adjust to your liking. The best part of a good Pad Thai is making it your own!
+"""
+
+    PROMPT = f"Convert this description to a fixed list of ingredients and a list of steps in the ingredients: {food_description}"
+    messages = [{"role": "system", "content": SYSTEM_PROMPT}]
+    messages.append({"role": "user", "content": PROMPT})
+
+    progress(0.8, desc="Generating recipe...")
+    try:
+        completion = sn_client.chat.completions.create(model=language_model, messages=messages)
+        response = completion.choices[0].message.content
+        thinking_time = time.time() - start_time
+        progress(0.9, desc="Recipe generation complete")
+        return response, thinking_time
+    except Exception as e:
+        error_message = f"Error: {str(e)}"
+        raise gr.Error(error_message)
+
+def process_image(image, api_key, vision_model, language_model, progress=gr.Progress()):
+    if not api_key.strip():
+        raise ValueError("Please enter your SambaNova API key")
+    progress(0, desc="Starting image processing...")
+    
+    progress(0.1, desc="Analyzing image...")
+    # Describe the food in the image
+    description, food_desc_time = describe_food_in_image(image, api_key, vision_model, progress)
+    
+    progress(0.5, desc="Generating recipe...")
+    # Generate recipe from description
+    recipe, recipe_gen_time = desc_to_ingredients_recipe(description, api_key, language_model, progress)
+    
+    progress(1.0, desc="Process complete")
+    return description, recipe, f"Time to response: {food_desc_time + recipe_gen_time:.2f} seconds"
+
+# Define the Gradio interface using Blocks
+with gr.Blocks() as iface:
+    gr.Markdown(
+        """
+        # Image to Recipe using SambaNova APIs 🚀
+        
+        **Ever go to a restaurant or see some food on TV and think "I could make this at home!". Well now just upload or take a picture of your food and get a recipe to try!**
+
+        *To use this, follow the instructions below:*
+
+        1. Navigate to <a href="https://cloud.sambanova.ai">https://cloud.sambanova.ai</a>, login and copy your API Key
+        2. Paste it in the SambaNova API Key box
+        3. (optional) Select a different vision or language model
+        4. Click the camera to use your device's camera, or upload button and select an image file, or simply paste from clipboard
+        5. It'll automatically start processing, wait for a few seconds for the LVM and LLM on SambaNova to run
+        6. Read through and try your receipe 😋
+        """, container=True
+    )
+    
+    with gr.Row():
+        with gr.Column(scale=1):
+            api_key_input = gr.Textbox(
+                type="password", 
+                label="SambaNova API Key", 
+                placeholder="Enter your API key here",
+                container=True
+            )
+            image_input = gr.Image(
+                type="pil", 
+                label="Food Image",
+                container=True
+            )
+            vision_model = gr.Dropdown(
+                choices=['Llama-3.2-11B-Vision-Instruct', 'Llama-3.2-90B-Vision-Instruct'],
+                value='Llama-3.2-11B-Vision-Instruct',  
+                label="Select a vision model"
+            )
+            language_model = gr.Dropdown(
+                choices=["Meta-Llama-3.1-405B-Instruct","Meta-Llama-3.1-70B-Instruct", "Meta-Llama-3.1-8B-Instruct"],
+                value='Meta-Llama-3.1-8B-Instruct',  
+                label="Select a language model"
+            )
+            
+        with gr.Column(scale=1):
+            food_description = gr.Textbox(
+                label="Food Description",
+                lines=3,
+                interactive=False
+            )
+            processing_time = gr.Textbox(
+                label="Processing Time",
+                interactive=False
+            )
+            recipe_output = gr.Markdown(
+                label="Recipe", value="Recipe", container=True
+            )
+    
+    # Handle the submit button click
+    image_input.change(
+        fn=process_image,
+        inputs=[image_input, api_key_input, vision_model, language_model],
+        outputs=[food_description, recipe_output, processing_time],
+        api_name="process_image"
+    )
+
+# Launch the app
+iface.launch()