Update main.py
Browse files
main.py
CHANGED
@@ -1,39 +1,22 @@
|
|
1 |
import os
|
2 |
-
import io
|
3 |
-
from io import BytesIO
|
4 |
from fastapi import FastAPI, UploadFile, File, Form
|
5 |
-
from fastapi.responses import JSONResponse, HTMLResponse
|
6 |
from fastapi.middleware.cors import CORSMiddleware
|
7 |
-
from fastapi.
|
8 |
-
from huggingface_hub import InferenceClient
|
9 |
from PyPDF2 import PdfReader
|
10 |
from docx import Document
|
11 |
from PIL import Image
|
12 |
-
|
13 |
-
|
14 |
-
|
15 |
-
# 1) Authentification Hugging Face
|
16 |
-
# ββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ
|
17 |
-
HF_TOKEN = os.getenv("HF_TOKEN") or os.getenv("HUGGINGFACE_HUB_TOKEN")
|
18 |
-
if not HF_TOKEN:
|
19 |
-
raise RuntimeError(
|
20 |
-
"Variable d'environnement HF_TOKEN absente ; crΓ©ez un jeton Β« Read Β» "
|
21 |
-
"sur https://huggingface.co/settings/tokens et exportez-le (voir .env)."
|
22 |
-
)
|
23 |
-
|
24 |
-
login(token=HF_TOKEN) # Authentifie tout le process
|
25 |
-
|
26 |
-
# ββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ
|
27 |
-
# 2) Configuration FastAPI
|
28 |
-
# ββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ
|
29 |
-
PORT = int(os.getenv("PORT", 7860))
|
30 |
-
|
31 |
-
app = FastAPI(
|
32 |
-
title="AI Web App API",
|
33 |
-
description="Backend API for AI-powered web application",
|
34 |
-
version="1.0.0"
|
35 |
-
)
|
36 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
37 |
app.add_middleware(
|
38 |
CORSMiddleware,
|
39 |
allow_origins=["*"],
|
@@ -42,42 +25,142 @@ app.add_middleware(
|
|
42 |
allow_headers=["*"],
|
43 |
)
|
44 |
|
45 |
-
|
46 |
-
|
|
|
|
|
47 |
|
48 |
-
#
|
49 |
-
summary_client = InferenceClient("facebook/bart-large-cnn")
|
50 |
-
qa_client = InferenceClient("deepset/roberta-base-squad2")
|
51 |
-
image_caption_client = InferenceClient("nlpconnect/vit-gpt2-image-captioning")
|
52 |
-
|
53 |
-
# ββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ
|
54 |
-
# 3) Utils : extraction texte, routes API (inchangΓ©s ou presque)
|
55 |
-
# ββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ
|
56 |
def extract_text_from_pdf(content: bytes) -> str:
|
|
|
57 |
reader = PdfReader(io.BytesIO(content))
|
58 |
-
|
|
|
|
|
|
|
59 |
|
60 |
def extract_text_from_docx(content: bytes) -> str:
|
|
|
61 |
doc = Document(io.BytesIO(content))
|
62 |
-
|
|
|
|
|
63 |
|
64 |
def process_uploaded_file(file: UploadFile) -> str:
|
65 |
content = file.file.read()
|
66 |
-
|
67 |
-
|
|
|
68 |
return extract_text_from_pdf(content)
|
69 |
-
|
70 |
return extract_text_from_docx(content)
|
71 |
-
|
72 |
return content.decode("utf-8").strip()
|
73 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
74 |
|
75 |
-
|
76 |
|
77 |
-
|
78 |
-
|
79 |
-
return {"status": "healthy", "version": "1.0.0", "hf_token_set": True}
|
80 |
|
|
|
81 |
if __name__ == "__main__":
|
82 |
import uvicorn
|
83 |
-
uvicorn.run(app, host="0.0.0.0", port=
|
|
|
1 |
import os
|
|
|
|
|
2 |
from fastapi import FastAPI, UploadFile, File, Form
|
|
|
3 |
from fastapi.middleware.cors import CORSMiddleware
|
4 |
+
from fastapi.responses import JSONResponse, HTMLResponse
|
5 |
+
from huggingface_hub import InferenceClient
|
6 |
from PyPDF2 import PdfReader
|
7 |
from docx import Document
|
8 |
from PIL import Image
|
9 |
+
import io
|
10 |
+
from io import BytesIO
|
11 |
+
import requests
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
12 |
|
13 |
+
# Remplace ce token par le tien de manière sécurisée (variable d'environnement recommandée en production)
|
14 |
+
HUGGINGFACE_TOKEN = os.getenv("HF_TOKEN")
|
15 |
+
|
16 |
+
# Initialisation de l'app FastAPI
|
17 |
+
app = FastAPI()
|
18 |
+
|
19 |
+
# Autoriser les requΓͺtes Cross-Origin
|
20 |
app.add_middleware(
|
21 |
CORSMiddleware,
|
22 |
allow_origins=["*"],
|
|
|
25 |
allow_headers=["*"],
|
26 |
)
|
27 |
|
28 |
+
# Initialisation des clients Hugging Face avec authentification
|
29 |
+
summary_client = InferenceClient(model="facebook/bart-large-cnn", token=HUGGINGFACE_TOKEN)
|
30 |
+
qa_client = InferenceClient(model="deepset/roberta-base-squad2", token=HUGGINGFACE_TOKEN)
|
31 |
+
image_caption_client = InferenceClient(model="nlpconnect/vit-gpt2-image-captioning", token=HUGGINGFACE_TOKEN)
|
32 |
|
33 |
+
# Extraction du texte des fichiers
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
34 |
def extract_text_from_pdf(content: bytes) -> str:
|
35 |
+
text = ""
|
36 |
reader = PdfReader(io.BytesIO(content))
|
37 |
+
for page in reader.pages:
|
38 |
+
if page.extract_text():
|
39 |
+
text += page.extract_text() + "\n"
|
40 |
+
return text.strip()
|
41 |
|
42 |
def extract_text_from_docx(content: bytes) -> str:
|
43 |
+
text = ""
|
44 |
doc = Document(io.BytesIO(content))
|
45 |
+
for para in doc.paragraphs:
|
46 |
+
text += para.text + "\n"
|
47 |
+
return text.strip()
|
48 |
|
49 |
def process_uploaded_file(file: UploadFile) -> str:
|
50 |
content = file.file.read()
|
51 |
+
extension = file.filename.split('.')[-1].lower()
|
52 |
+
|
53 |
+
if extension == "pdf":
|
54 |
return extract_text_from_pdf(content)
|
55 |
+
elif extension == "docx":
|
56 |
return extract_text_from_docx(content)
|
57 |
+
elif extension == "txt":
|
58 |
return content.decode("utf-8").strip()
|
59 |
+
else:
|
60 |
+
raise ValueError("Type de fichier non supportΓ©")
|
61 |
+
|
62 |
+
# Point d'entrΓ©e HTML
|
63 |
+
@app.get("/", response_class=HTMLResponse)
|
64 |
+
async def serve_homepage():
|
65 |
+
with open("index.html", "r", encoding="utf-8") as f:
|
66 |
+
return HTMLResponse(content=f.read(), status_code=200)
|
67 |
+
|
68 |
+
# RΓ©sumΓ©
|
69 |
+
@app.post("/analyze")
|
70 |
+
async def analyze_file(file: UploadFile = File(...)):
|
71 |
+
try:
|
72 |
+
text = process_uploaded_file(file)
|
73 |
+
|
74 |
+
if len(text) < 20:
|
75 |
+
return {"summary": "Document trop court pour Γͺtre rΓ©sumΓ©."}
|
76 |
+
|
77 |
+
summary = summary_client.summarization(text[:3000])
|
78 |
+
return {"summary": summary}
|
79 |
+
|
80 |
+
except Exception as e:
|
81 |
+
return JSONResponse(status_code=500, content={"error": f"Erreur lors de l'analyse: {str(e)}"})
|
82 |
+
|
83 |
+
# Question-RΓ©ponse
|
84 |
+
@app.post("/ask")
|
85 |
+
async def ask_question(file: UploadFile = File(...), question: str = Form(...)):
|
86 |
+
try:
|
87 |
+
# Determine if the file is an image
|
88 |
+
content_type = file.content_type
|
89 |
+
if content_type.startswith("image/"):
|
90 |
+
image_bytes = await file.read()
|
91 |
+
image_pil = Image.open(io.BytesIO(image_bytes)).convert("RGB")
|
92 |
+
image_pil.thumbnail((1024, 1024))
|
93 |
+
|
94 |
+
img_byte_arr = BytesIO()
|
95 |
+
image_pil.save(img_byte_arr, format='JPEG')
|
96 |
+
img_byte_arr = img_byte_arr.getvalue()
|
97 |
+
|
98 |
+
# Generate image description
|
99 |
+
result = image_caption_client.image_to_text(img_byte_arr)
|
100 |
+
if isinstance(result, dict):
|
101 |
+
context = result.get("generated_text") or result.get("caption") or ""
|
102 |
+
elif isinstance(result, list) and len(result) > 0:
|
103 |
+
context = result[0].get("generated_text", "")
|
104 |
+
elif isinstance(result, str):
|
105 |
+
context = result
|
106 |
+
else:
|
107 |
+
context = ""
|
108 |
+
|
109 |
+
else:
|
110 |
+
# Not an image, process as document
|
111 |
+
text = process_uploaded_file(file)
|
112 |
+
if len(text) < 20:
|
113 |
+
return {"answer": "Document trop court pour rΓ©pondre Γ la question."}
|
114 |
+
context = text[:3000]
|
115 |
+
|
116 |
+
if not context:
|
117 |
+
return {"answer": "Aucune information disponible pour rΓ©pondre Γ la question."}
|
118 |
+
|
119 |
+
result = qa_client.question_answering(question=question, context=context)
|
120 |
+
return {"answer": result.get("answer", "Aucune rΓ©ponse trouvΓ©e.")}
|
121 |
+
|
122 |
+
except Exception as e:
|
123 |
+
return JSONResponse(status_code=500, content={"error": f"Erreur lors de la recherche de rΓ©ponse: {str(e)}"})
|
124 |
+
|
125 |
+
# InterprΓ©tation d'Image
|
126 |
+
@app.post("/interpret_image")
|
127 |
+
async def interpret_image(image: UploadFile = File(...)):
|
128 |
+
try:
|
129 |
+
# Lire l'image
|
130 |
+
image_bytes = await image.read()
|
131 |
+
|
132 |
+
# Ouvrir l'image avec PIL
|
133 |
+
image_pil = Image.open(io.BytesIO(image_bytes))
|
134 |
+
image_pil = image_pil.convert("RGB")
|
135 |
+
image_pil.thumbnail((1024, 1024))
|
136 |
+
|
137 |
+
# Convertir en bytes (JPEG)
|
138 |
+
img_byte_arr = BytesIO()
|
139 |
+
image_pil.save(img_byte_arr, format='JPEG')
|
140 |
+
img_byte_arr = img_byte_arr.getvalue()
|
141 |
+
|
142 |
+
# Appeler le modèle
|
143 |
+
result = image_caption_client.image_to_text(img_byte_arr)
|
144 |
+
|
145 |
+
# π Affichage du rΓ©sultat brut pour dΓ©bogage
|
146 |
+
print("Résultat brut du modèle image-to-text:", result)
|
147 |
+
|
148 |
+
# Extraire la description si disponible
|
149 |
+
if isinstance(result, dict):
|
150 |
+
description = result.get("generated_text") or result.get("caption") or "Description non trouvΓ©e."
|
151 |
+
elif isinstance(result, list) and len(result) > 0:
|
152 |
+
description = result[0].get("generated_text", "Description non trouvΓ©e.")
|
153 |
+
elif isinstance(result, str):
|
154 |
+
description = result
|
155 |
+
else:
|
156 |
+
description = "Description non trouvΓ©e."
|
157 |
|
158 |
+
return {"description": description}
|
159 |
|
160 |
+
except Exception as e:
|
161 |
+
return JSONResponse(status_code=500, content={"error": f"Erreur lors de l'interprΓ©tation de l'image: {str(e)}"})
|
|
|
162 |
|
163 |
+
# DΓ©marrage local
|
164 |
if __name__ == "__main__":
|
165 |
import uvicorn
|
166 |
+
uvicorn.run("main:app", host="0.0.0.0", port=7860, reload=True)
|