Update app.py

app.py

@@ -1,60 +1,149 @@
 import gradio as gr
-import arxiv
-from transformers import pipeline
+import requests
+import os
+from fpdf import FPDF
+import uuid
+import re
+import matplotlib.pyplot as plt
+import tempfile
 
-# Load summarization model
-summarizer = pipeline("summarization", model="sshleifer/distilbart-cnn-12-6")
+# Load your Nebius API key and Folder ID from environment variables
+NEBIUS_API_KEY = os.getenv("NEBIUS_API_KEY") or "YOUR_NEBIUS_API_KEY"
+FOLDER_ID = os.getenv("FOLDER_ID") or "YOUR_FOLDER_ID"
 
-# Search and summarize papers
-def search_and_summarize(topic, sort_by_option):
+def generate_meal(preferences, ingredients, time):
+    prompt = f"""
+You're a smart kitchen agent.
+Given:
+- Dietary preferences: {preferences}
+- Ingredients available: {ingredients}
+- Time available: {time} minutes
+Tasks:
+1. Suggest one meal idea
+2. Provide step-by-step recipe instructions
+3. List missing ingredients (Shopping List)
+4. Estimate nutrition (calories, protein, carbs, fat)
+Output in this format:
+Meal Name: ...
+Steps:
+1. ...
+2. ...
+Shopping List:
+- ...
+Nutrition:
+- Calories: ... kcal
+- Protein: ... g
+- Carbs: ... g
+- Fat: ... g
+"""
+    headers = {
+        "Authorization": f"Api-Key {NEBIUS_API_KEY}",
+        "Content-Type": "application/json"
+    }
+    data = {
+        "modelUri": f"gpt://{FOLDER_ID}/yandexgpt-lite",
+        "completionOptions": {"stream": False, "temperature": 0.7, "maxTokens": 700},
+        "messages": [{"role": "user", "text": prompt.strip()}]
+    }
+    response = requests.post(
+        "https://llm.api.cloud.yandex.net/foundationModels/v1/completion",
+        headers=headers, json=data
+    )
     try:
-        num_papers = 3  # fixed value
-        sort_mapping = {
-            "Relevance": arxiv.SortCriterion.Relevance,
-            "Most Recent": arxiv.SortCriterion.SubmittedDate
-        }
-
-        search = arxiv.Search(
-            query=topic,
-            max_results=num_papers,
-            sort_by=sort_mapping.get(sort_by_option, arxiv.SortCriterion.Relevance)
-        )
-
-        results = []
-        for result in search.results():
-            summary = summarizer(result.summary[:1000], max_length=120, min_length=30, do_sample=False)[0]['summary_text']
-            authors = ", ".join([author.name for author in result.authors])
-            published_date = result.published.date().strftime("%Y-%m-%d")
-            result_block = (
-                f"📘 *{result.title}*\n\n"
-                f"👩‍🔬 Authors: {authors}\n"
-                f"📅 Published: {published_date}\n\n"
-                f"📝 Summary: {summary}\n\n"
-                f"🔗 [Read More]({result.pdf_url})"
-            )
-            results.append(result_block)
-
-        return "\n\n---\n\n".join(results) if results else "No results found."
-
+        text = response.json()["result"]["alternatives"][0]["message"]["text"]
     except Exception as e:
-        return f"⚠️ An error occurred: {e}"
+        return f"❌ Error: Could not fetch recipe. {e}"
+    return text
+
+def extract_nutrition(recipe_text):
+    # More robust pattern allowing newlines and flexible spacing
+    pattern = (
+        r"Calories:\s*([\d\.]+)\s*kcal.*?"
+        r"Protein:\s*([\d\.]+)\s*g.*?"
+        r"Carbs:\s*([\d\.]+)\s*g.*?"
+        r"Fat:\s*([\d\.]+)\s*g"
+    )
+    match = re.search(pattern, recipe_text.replace("\n", " "))
+    if match:
+        calories, protein, carbs, fat = map(float, match.groups())
+        return {"Calories": calories, "Protein": protein, "Carbs": carbs, "Fat": fat}
+    return None
+
+def plot_nutrition_chart(nutrition):
+    fig, ax = plt.subplots()
+    nutrients = list(nutrition.keys())
+    values = list(nutrition.values())
+    colors = ['#ff9999','#66b3ff','#99ff99','#ffcc99']
+    ax.pie(values, labels=nutrients, autopct='%1.1f%%', startangle=140, colors=colors)
+    ax.axis('equal')
+    plt.tight_layout()
+
+    temp_file = tempfile.NamedTemporaryFile(suffix=".png", delete=False)
+    plt.savefig(temp_file.name)
+    plt.close(fig)
+    return temp_file.name
 
-# Gradio UI
-with gr.Blocks(theme=gr.themes.Base()) as demo:
-    gr.Markdown("# 🤖 AI Research Assistant\nSummarize academic research papers using Hugging Face models + Arxiv!")
+def handle_generate(preferences, ingredients, time):
+    recipe = generate_meal(preferences, ingredients, time)
+    nutrition = extract_nutrition(recipe)
+    chart_path = None
+    if nutrition:
+        chart_path = plot_nutrition_chart(nutrition)
+    return recipe, chart_path
+
+def save_pdf(recipe_text):
+    pdf = FPDF()
+    pdf.add_page()
+    pdf.set_font("Arial", size=12)
+    for line in recipe_text.split('\n'):
+        pdf.multi_cell(0, 10, line)
+    temp_file = tempfile.NamedTemporaryFile(suffix=".pdf", delete=False)
+    pdf.output(temp_file.name)
+    return temp_file.name
+
+# Optional: placeholder function for speech-to-text (you can replace with real STT)
+def dummy_speech_to_text(audio):
+    # This just returns a fixed string for demo. Replace with actual STT model call if you want.
+    if audio is None:
+        return ""
+    return "tomato, onion, garlic"
+
+with gr.Blocks(css="""
+    body { background-color: #111; color: #eee; font-family: 'Segoe UI', sans-serif; }
+    .gradio-container { max-width: 900px; margin: auto; }
+    .input-label { font-weight: 600; margin-bottom: 5px; }
+    .footer { font-size: 0.8rem; color: #666; padding: 10px; text-align: center; }
+    @media (max-width: 600px) {
+        .gradio-container { padding: 10px; }
+    }
+""") as demo:
+    gr.Markdown("# 👨‍🍳 AgentChef: AI Recipe Planner & Smart Kitchen Assistant")
 
     with gr.Row():
-        topic = gr.Textbox(label="🔍 Enter your research topic", placeholder="e.g. diffusion models in AI")
-        sort_by = gr.Dropdown(choices=["Relevance", "Most Recent"], value="Relevance", label="Sort by")
+        with gr.Column(scale=1, min_width=200):
+            preferences = gr.Textbox(label="🥗 Dietary Preferences", placeholder="e.g. vegetarian, keto")
+            ingredients = gr.Textbox(label="🧂 Ingredients You Have", placeholder="e.g. rice, tomato, onion", lines=3)
+            
+            mic = gr.Audio(source="microphone", type="filepath", label="🎤 Speak Ingredients")
+            transcribed = gr.Textbox(label="📝 Transcribed Ingredients (from mic)", interactive=False)
+
+            mic.change(dummy_speech_to_text, inputs=mic, outputs=transcribed)
+            # Add a button to copy transcribed text to ingredients textbox if needed
+            copy_btn = gr.Button("Copy Transcription to Ingredients")
+            copy_btn.click(lambda txt: txt, inputs=transcribed, outputs=ingredients)
+            
+            time = gr.Slider(5, 60, value=20, step=5, label="⏱️ Time Available (minutes)")
+            generate_btn = gr.Button("🍽️ Generate Recipe")
 
-    search_btn = gr.Button("Search 🔎")
-    output = gr.Markdown()
+        with gr.Column(scale=1, min_width=300):
+            recipe_output = gr.Textbox(label="📝 Recipe Output", lines=15, interactive=False)
+            nutrition_chart = gr.Image(label="📊 Nutrition Breakdown", interactive=False)
+            pdf_btn = gr.Button("📄 Export as PDF")
 
-    # Show loading message
-    def show_loading():
-        return "⏳ Loading, please wait..."
+    generate_btn.click(handle_generate,
+                       inputs=[preferences, ingredients, time],
+                       outputs=[recipe_output, nutrition_chart])
 
-    search_btn.click(fn=show_loading, inputs=[], outputs=output, queue=False)
-    search_btn.click(fn=search_and_summarize, inputs=[topic, sort_by], outputs=output)
+    pdf_btn.click(save_pdf, inputs=[recipe_output], outputs=gr.File(label="📥 Download Recipe PDF"))
 
 demo.launch()