Spaces:
Running
Running
Update app.py
Browse files
app.py
CHANGED
@@ -1,18 +1,49 @@
|
|
1 |
-
from flask import Flask, render_template, request, jsonify
|
2 |
import os
|
3 |
import json
|
4 |
from textwrap import dedent
|
5 |
from crewai import Agent, Crew, Process, Task
|
6 |
-
from crewai_tools import SerperDevTool
|
7 |
from crewai import LLM
|
8 |
-
from typing import List
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
9 |
|
10 |
app = Flask(__name__)
|
11 |
|
12 |
-
# Configuration
|
13 |
-
|
14 |
-
|
15 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
16 |
llm = LLM(
|
17 |
model="gemini/gemini-1.5-flash",
|
18 |
temperature=0.7,
|
@@ -20,100 +51,195 @@ llm = LLM(
|
|
20 |
max_tokens=8000,
|
21 |
)
|
22 |
|
23 |
-
#
|
24 |
search_tool = SerperDevTool()
|
25 |
|
26 |
-
|
27 |
-
|
28 |
-
|
29 |
-
|
30 |
-
|
31 |
-
|
32 |
-
|
33 |
-
|
34 |
-
|
35 |
-
|
36 |
-
|
37 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
38 |
|
39 |
-
|
40 |
-
|
41 |
-
|
42 |
-
|
43 |
-
|
44 |
-
|
45 |
-
|
46 |
-
|
47 |
-
|
48 |
-
|
|
|
|
|
49 |
|
50 |
-
def
|
51 |
-
|
52 |
-
|
53 |
-
|
54 |
-
|
55 |
-
if json_start == -1 or json_end == 0:
|
56 |
-
raise ValueError("JSON non trouvé dans le résultat")
|
57 |
|
58 |
-
|
59 |
-
|
60 |
-
|
61 |
-
|
62 |
-
|
63 |
-
def research_task(topic: str) -> Task:
|
64 |
-
return Task(
|
65 |
-
description=dedent(f"""Effectuer une recherche approfondie sur le sujet '{topic}'.
|
66 |
-
Compiler les informations les plus pertinentes et les plus récentes."""),
|
67 |
-
expected_output="Une liste d'informations pertinentes sur le sujet.",
|
68 |
-
agent=researcher
|
69 |
-
)
|
70 |
-
|
71 |
-
def flashcard_creation_task(research_task: Task) -> Task:
|
72 |
-
return Task(
|
73 |
-
description=dedent("""Transformer les informations fournies par le Chercheur
|
74 |
-
en une série de flashcards au format JSON. Chaque flashcard doit avoir une question
|
75 |
-
d'un côté et une réponse concise de l'autre."""),
|
76 |
-
expected_output="Une liste de flashcards au format JSON.",
|
77 |
-
agent=writer,
|
78 |
-
context=[research_task]
|
79 |
-
)
|
80 |
-
|
81 |
-
@app.route('/')
|
82 |
-
def index():
|
83 |
-
return render_template('index.html')
|
84 |
|
85 |
-
|
86 |
-
|
87 |
-
|
88 |
-
|
89 |
-
|
|
|
|
|
|
|
|
|
|
|
90 |
|
91 |
-
|
92 |
-
|
93 |
-
|
94 |
-
|
95 |
|
96 |
-
# Création de la crew
|
97 |
crew = Crew(
|
98 |
-
agents=[researcher, writer],
|
99 |
-
tasks=[
|
100 |
process=Process.sequential,
|
101 |
verbose=True
|
102 |
)
|
103 |
|
104 |
-
# Exécution de la crew
|
105 |
result = crew.kickoff()
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
106 |
|
107 |
-
|
108 |
-
|
109 |
-
|
110 |
-
|
111 |
-
|
112 |
-
else:
|
113 |
-
return jsonify({'error': 'Aucun résultat généré par la crew.'}), 500
|
114 |
|
115 |
except Exception as e:
|
|
|
116 |
return jsonify({'error': str(e)}), 500
|
117 |
|
|
|
|
|
|
|
|
|
|
|
118 |
if __name__ == '__main__':
|
119 |
app.run(debug=True)
|
|
|
1 |
+
from flask import Flask, render_template, request, jsonify, secure_filename
|
2 |
import os
|
3 |
import json
|
4 |
from textwrap import dedent
|
5 |
from crewai import Agent, Crew, Process, Task
|
6 |
+
from crewai_tools import SerperDevTool, PDFSearchTool
|
7 |
from crewai import LLM
|
8 |
+
from typing import List, Dict, Union, Optional
|
9 |
+
import tempfile
|
10 |
+
import logging
|
11 |
+
from datetime import datetime
|
12 |
+
import hashlib
|
13 |
+
|
14 |
+
# Configure logging
|
15 |
+
logging.basicConfig(
|
16 |
+
level=logging.INFO,
|
17 |
+
format='%(asctime)s - %(name)s - %(levelname)s - %(message)s'
|
18 |
+
)
|
19 |
+
logger = logging.getLogger(__name__)
|
20 |
|
21 |
app = Flask(__name__)
|
22 |
|
23 |
+
# Configuration
|
24 |
+
UPLOAD_FOLDER = tempfile.gettempdir()
|
25 |
+
ALLOWED_EXTENSIONS = {'pdf'}
|
26 |
+
MAX_FILE_SIZE = 16 * 1024 * 1024 # 16MB
|
27 |
+
app.config['UPLOAD_FOLDER'] = UPLOAD_FOLDER
|
28 |
+
app.config['MAX_CONTENT_LENGTH'] = MAX_FILE_SIZE
|
29 |
+
|
30 |
+
# Load environment variables
|
31 |
+
def get_env_variable(var_name: str) -> str:
|
32 |
+
"""Safely get environment variable with error handling."""
|
33 |
+
value = os.environ.get(var_name)
|
34 |
+
if value is None:
|
35 |
+
raise ValueError(f"Environment variable {var_name} is not set")
|
36 |
+
return value
|
37 |
+
|
38 |
+
try:
|
39 |
+
# API Keys configuration
|
40 |
+
os.environ["GEMINI_API_KEY"] = get_env_variable("GEMINI_API_KEY")
|
41 |
+
os.environ["SERPER_API_KEY"] = get_env_variable("SERPER_API_KEY")
|
42 |
+
except ValueError as e:
|
43 |
+
logger.error(f"Configuration error: {e}")
|
44 |
+
raise
|
45 |
+
|
46 |
+
# Initialize LLM
|
47 |
llm = LLM(
|
48 |
model="gemini/gemini-1.5-flash",
|
49 |
temperature=0.7,
|
|
|
51 |
max_tokens=8000,
|
52 |
)
|
53 |
|
54 |
+
# Initialize tools
|
55 |
search_tool = SerperDevTool()
|
56 |
|
57 |
+
def create_pdf_tool(pdf_path: Optional[str] = None) -> PDFSearchTool:
|
58 |
+
"""Create a PDFSearchTool with optional PDF path."""
|
59 |
+
config = {
|
60 |
+
'llm': {
|
61 |
+
'provider': 'google',
|
62 |
+
'config': {
|
63 |
+
'model': 'gemini-1.5-flash',
|
64 |
+
},
|
65 |
+
},
|
66 |
+
'embedder': {
|
67 |
+
'provider': 'google',
|
68 |
+
'config': {
|
69 |
+
'model': 'models/embedding-001',
|
70 |
+
'task_type': 'retrieval_document',
|
71 |
+
},
|
72 |
+
},
|
73 |
+
}
|
74 |
+
|
75 |
+
if pdf_path:
|
76 |
+
return PDFSearchTool(pdf=pdf_path, config=config)
|
77 |
+
return PDFSearchTool(config=config)
|
78 |
+
|
79 |
+
def allowed_file(filename: str) -> bool:
|
80 |
+
"""Check if the file extension is allowed."""
|
81 |
+
return '.' in filename and filename.rsplit('.', 1)[1].lower() in ALLOWED_EXTENSIONS
|
82 |
+
|
83 |
+
def generate_safe_filename(filename: str) -> str:
|
84 |
+
"""Generate a safe filename with timestamp and hash."""
|
85 |
+
timestamp = datetime.now().strftime('%Y%m%d_%H%M%S')
|
86 |
+
file_hash = hashlib.md5(filename.encode()).hexdigest()[:10]
|
87 |
+
ext = filename.rsplit('.', 1)[1].lower()
|
88 |
+
return f"upload_{timestamp}_{file_hash}.{ext}"
|
89 |
+
|
90 |
+
class FlashcardGenerator:
|
91 |
+
def __init__(self, topic: str, pdf_path: Optional[str] = None):
|
92 |
+
self.topic = topic
|
93 |
+
self.pdf_path = pdf_path
|
94 |
+
self.researcher = self._create_researcher()
|
95 |
+
self.writer = self._create_writer()
|
96 |
+
|
97 |
+
def _create_researcher(self) -> Agent:
|
98 |
+
"""Create the researcher agent with appropriate tools."""
|
99 |
+
tools = [search_tool]
|
100 |
+
if self.pdf_path:
|
101 |
+
tools.append(create_pdf_tool(self.pdf_path))
|
102 |
+
|
103 |
+
return Agent(
|
104 |
+
role='Chercheur de Sujets',
|
105 |
+
goal=dedent(f"""Trouver les informations les plus pertinentes et précises sur {self.topic}
|
106 |
+
en utilisant l'API SerpApi et en analysant les PDFs fournis."""),
|
107 |
+
backstory=dedent("""Un chercheur expert spécialisé dans la collecte d'informations sur divers sujets.
|
108 |
+
Capable d'utiliser l'API SerpApi pour des recherches précises et d'analyser des documents PDF."""),
|
109 |
+
tools=tools,
|
110 |
+
llm=llm,
|
111 |
+
verbose=True,
|
112 |
+
allow_delegation=False
|
113 |
+
)
|
114 |
|
115 |
+
def _create_writer(self) -> Agent:
|
116 |
+
"""Create the writer agent."""
|
117 |
+
return Agent(
|
118 |
+
role='Rédacteur de Flashcards',
|
119 |
+
goal=dedent("""Créer des flashcards claires et concises en format question-réponse
|
120 |
+
basées sur les informations fournies par le Chercheur."""),
|
121 |
+
backstory=dedent("""Un expert en pédagogie et en création de matériel d'apprentissage.
|
122 |
+
Capable de transformer des informations complexes en flashcards simples et mémorisables."""),
|
123 |
+
llm=llm,
|
124 |
+
verbose=True,
|
125 |
+
allow_delegation=False
|
126 |
+
)
|
127 |
|
128 |
+
def create_research_task(self) -> Task:
|
129 |
+
"""Create the research task."""
|
130 |
+
description = f"""Effectuer une recherche approfondie sur le sujet '{self.topic}'."""
|
131 |
+
if self.pdf_path:
|
132 |
+
description += f" Analyser également le contenu du PDF fourni: {self.pdf_path}"
|
|
|
|
|
133 |
|
134 |
+
return Task(
|
135 |
+
description=dedent(description),
|
136 |
+
expected_output="Une liste d'informations pertinentes sur le sujet.",
|
137 |
+
agent=self.researcher
|
138 |
+
)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
139 |
|
140 |
+
def create_flashcard_task(self, research_task: Task) -> Task:
|
141 |
+
"""Create the flashcard creation task."""
|
142 |
+
return Task(
|
143 |
+
description=dedent("""Transformer les informations fournies par le Chercheur
|
144 |
+
en une série de flashcards au format JSON. Chaque flashcard doit avoir une question
|
145 |
+
d'un côté et une réponse concise de l'autre. Les réponses doivent être claires et informatives."""),
|
146 |
+
expected_output="Une liste de flashcards au format JSON.",
|
147 |
+
agent=self.writer,
|
148 |
+
context=[research_task]
|
149 |
+
)
|
150 |
|
151 |
+
def generate(self) -> List[Dict[str, str]]:
|
152 |
+
"""Generate flashcards using the crew workflow."""
|
153 |
+
research_task = self.create_research_task()
|
154 |
+
flashcard_task = self.create_flashcard_task(research_task)
|
155 |
|
|
|
156 |
crew = Crew(
|
157 |
+
agents=[self.researcher, self.writer],
|
158 |
+
tasks=[research_task, flashcard_task],
|
159 |
process=Process.sequential,
|
160 |
verbose=True
|
161 |
)
|
162 |
|
|
|
163 |
result = crew.kickoff()
|
164 |
+
return self.extract_json_from_result(result.tasks_output[-1].raw)
|
165 |
+
|
166 |
+
@staticmethod
|
167 |
+
def extract_json_from_result(result_text: str) -> List[Dict[str, str]]:
|
168 |
+
"""Extract and validate JSON from the result text."""
|
169 |
+
try:
|
170 |
+
json_start = result_text.find('[')
|
171 |
+
json_end = result_text.rfind(']') + 1
|
172 |
+
if json_start == -1 or json_end == 0:
|
173 |
+
raise ValueError("JSON non trouvé dans le résultat")
|
174 |
+
|
175 |
+
json_str = result_text[json_start:json_end]
|
176 |
+
flashcards = json.loads(json_str)
|
177 |
+
|
178 |
+
# Validate flashcard format
|
179 |
+
for card in flashcards:
|
180 |
+
if not isinstance(card, dict) or 'question' not in card or 'answer' not in card:
|
181 |
+
raise ValueError("Format de flashcard invalide")
|
182 |
+
|
183 |
+
return flashcards
|
184 |
+
except (json.JSONDecodeError, ValueError) as e:
|
185 |
+
logger.error(f"Error extracting JSON: {str(e)}")
|
186 |
+
raise ValueError(f"Erreur lors de l'extraction du JSON : {str(e)}")
|
187 |
+
|
188 |
+
@app.route('/')
|
189 |
+
def index():
|
190 |
+
"""Render the main page."""
|
191 |
+
return render_template('index.html')
|
192 |
+
|
193 |
+
@app.route('/generate', methods=['POST'])
|
194 |
+
def generate_flashcards():
|
195 |
+
"""Handle flashcard generation requests."""
|
196 |
+
try:
|
197 |
+
# Validate topic
|
198 |
+
topic = request.form.get('topic')
|
199 |
+
if not topic:
|
200 |
+
return jsonify({'error': 'Veuillez entrer un sujet.'}), 400
|
201 |
+
|
202 |
+
# Handle file upload
|
203 |
+
pdf_path = None
|
204 |
+
if 'file' in request.files:
|
205 |
+
file = request.files['file']
|
206 |
+
if file and file.filename:
|
207 |
+
if not allowed_file(file.filename):
|
208 |
+
return jsonify({'error': 'Format de fichier non supporté. Veuillez utiliser un PDF.'}), 400
|
209 |
+
|
210 |
+
if file.content_length and file.content_length > MAX_FILE_SIZE:
|
211 |
+
return jsonify({'error': 'Le fichier est trop volumineux. Maximum 16MB.'}), 400
|
212 |
+
|
213 |
+
# Generate safe filename and save file
|
214 |
+
safe_filename = generate_safe_filename(file.filename)
|
215 |
+
pdf_path = os.path.join(app.config['UPLOAD_FOLDER'], safe_filename)
|
216 |
+
file.save(pdf_path)
|
217 |
+
logger.info(f"File saved: {pdf_path}")
|
218 |
+
|
219 |
+
try:
|
220 |
+
# Generate flashcards
|
221 |
+
generator = FlashcardGenerator(topic, pdf_path)
|
222 |
+
flashcards = generator.generate()
|
223 |
+
|
224 |
+
return jsonify({
|
225 |
+
'success': True,
|
226 |
+
'flashcards': flashcards
|
227 |
+
})
|
228 |
|
229 |
+
finally:
|
230 |
+
# Clean up PDF file
|
231 |
+
if pdf_path and os.path.exists(pdf_path):
|
232 |
+
os.remove(pdf_path)
|
233 |
+
logger.info(f"File cleaned up: {pdf_path}")
|
|
|
|
|
234 |
|
235 |
except Exception as e:
|
236 |
+
logger.error(f"Error generating flashcards: {str(e)}")
|
237 |
return jsonify({'error': str(e)}), 500
|
238 |
|
239 |
+
@app.errorhandler(413)
|
240 |
+
def request_entity_too_large(error):
|
241 |
+
"""Handle file size limit exceeded error."""
|
242 |
+
return jsonify({'error': 'Le fichier est trop volumineux. Maximum 16MB.'}), 413
|
243 |
+
|
244 |
if __name__ == '__main__':
|
245 |
app.run(debug=True)
|