Update app.py

app.py

@@ -1,8 +1,8 @@
-import requests
-import json
 import os
+import requests
+import gradio as gr
+from bs4 import BeautifulSoup
 
-# Votre analyseur de post
 api_token = os.environ.get("TOKEN")
 API_URL = "https://api-inference.huggingface.co/models/meta-llama/Meta-Llama-3-8B-Instruct"
 headers = {"Authorization": f"Bearer {api_token}"}
@@ -11,10 +11,10 @@ def query(payload):
     response = requests.post(API_URL, headers=headers, json=payload)
     return response.json()
 
-def analyze_sentiment(text):
+def analyze_sentiment(pl7_texts):
     output = query({
-        "inputs": f'''
-system
+        "inputs": f'''<|begin_of_text|>
+<|start_header_id|>system<|end_header_id|>
 You're going to deeply analyze the texts I'm going to give you and you're only going to tell me which category they belong to by answering only the words that correspond to the following categories:
 For posts that talk about chat models/LLM, return "Chatmodel/LLM"
 For posts that talk about image generation models, return "image_generation"
@@ -26,104 +26,46 @@ For posts about tools and libraries, return "tools_libraries"
 For posts containing tutorials and guides, return "tutorials_guides"
 For posts about debugging and problem-solving, return "debugging"
 Respond only with the category name, without any additional explanation or text.
-
-user
-{text}
-
-assistant
-
+<|eot_id|>
+<|start_header_id|>user<|end_header_id|>
+{pl7_texts}
+<|eot_id|>
+<|start_header_id|>assistant<|end_header_id|>
 '''
     })
-
     if isinstance(output, list) and len(output) > 0:
         response = output[0].get('generated_text', '').strip().lower()
-        return response
-
-    return "Erreur: Réponse vide ou non valide"
-
-def fetch_posts(limit=10, max_posts=50):
-    url = "https://huggingface.co/api/posts"
-    params = {'limit': limit}
-    offset = 0
-    all_posts = []
-
-    while len(all_posts) < max_posts:
-        params['offset'] = offset
-        response = requests.get(url, params=params)
-        data = response.json()
-        
-        if not data['socialPosts']:
-            break
-        
-        all_posts.extend(data['socialPosts'])
-        offset += len(data['socialPosts'])
-
-    return all_posts[:max_posts]
-
-def display_post(post):
-    print(f"\n{'=' * 50}")
-    print(f"Contenu:\n{post['rawContent']}")
-    print(f"Likes: {sum(reaction['count'] for reaction in post['reactions'])}")
-    print(f"URL: {post['url']}")
-    print(f"Analyse: {post.get('analysis', 'Analyse non disponible')}")
-    print(f"Catégorie: {post.get('category', 'Non analysé')}")
-    print(f"{'=' * 50}\n")
-
-# Récupérer les posts
-posts = fetch_posts(limit=10, max_posts=30)  # Récupère jusqu'à 30 posts
-
-# Stocker les posts dans des variables différentes
-post_variables = {f'post_{i}': post for i, post in enumerate(posts)}
-
-# Dictionnaire pour compter les catégories
-category_count = {
-    'questions': 0,
-    'Chatmodel/LLM': 0,
-    'image_generation': 0,
-    'fine_tuning': 0,
-    'ethics_bias': 0,
-    'datasets': 0,
-    'tools_libraries': 0,
-    'tutorials_guides': 0,
-    'debugging': 0,
-    'Erreur: Réponse ambiguë': 0,
-    'Erreur: Réponse vide ou non valide': 0
-}
-
-# Analyser chaque post et ajouter les résultats de l'analyse au dictionnaire
-for key in post_variables:
-    category = analyze_sentiment(post_variables[key]['rawContent'])
-    
-    if category == 'questions':
-        category_count['questions'] += 2
-    elif category == 'Chatmodel/LLM':
-        category_count['Chatmodel/LLM'] += 2
-    elif category == 'image_generation':
-        category_count['image_generation'] += 2
-    elif category == 'fine_tuning':
-        category_count['fine_tuning'] += 2
-    elif category == 'ethics_bias':
-        category_count['ethics_bias'] += 2
-    elif category == 'datasets':
-        category_count['datasets'] += 2
-    elif category == 'tools_libraries':
-        category_count['tools_libraries'] += 2
-    elif category == 'tutorials_guides':
-        category_count['tutorials_guides'] += 2
-    elif category == 'debugging':
-        category_count['debugging'] += 2
-    elif category == "Erreur: Réponse ambiguë":
-        category_count["Erreur: Réponse ambiguë"] += 2
-    elif category == "Erreur: Réponse vide ou non valide":
-        category_count["Erreur: Réponse vide ou non valide"] += 2
-
-    post_variables[key]['category'] = category
-
-# Imprimer un post spécifique
-post_to_print = 'post_2'  # Par exemple, le troisième post (index 2)
-display_post(post_variables[post_to_print])
-
-# Afficher le nombre de posts par catégorie
-print("\nNombre de posts par catégorie:")
-for category, count in category_count.items():
-    print(f"{category}: {count}")
+        categories = {
+            'questions': 'Questions',
+            'chatmodel/llm': 'Chat Model/LLM',
+            'image_generation': 'Image Generation',
+            'fine_tuning': 'Fine-tuning',
+            'ethics_bias': 'Ethics and Bias',
+            'datasets': 'Datasets',
+            'tools_libraries': 'Tools and Libraries',
+            'tutorials_guides': 'Tutorials and Guides',
+            'debugging': 'Debugging'
+        }
+        return categories.get(response, f"Error: Ambiguous response - '{response}'")
+    return "Error: No valid response received"
+
+url = 'https://huggingface.co/posts'
+response = requests.get(url)
+
+if response.status_code == 200:
+    soup = BeautifulSoup(response.content, 'html.parser')
+    pl7_elements = soup.find_all(class_='pl-7')
+    pl7_texts = [element.text.strip() for element in pl7_elements]
+
+    for idx, text in enumerate(pl7_texts, start=1):
+        print(f"Text pl-7 {idx}: {text}")
+        sentiment = analyze_sentiment(text)
+        print(f"Sentiment: {sentiment}\n")
+
+    if len(pl7_texts) >= 4:
+        print(f"Content of pl7_text_1: {pl7_texts[0]}")
+        print(f"Content of pl7_text_2: {pl7_texts[1]}")
+    else:
+        print("Not enough pl-7 elements found")
+else:
+    print(f"Error {response.status_code} when retrieving {url}")