Update main.py
Browse files
main.py
CHANGED
@@ -1,22 +1,26 @@
|
|
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 |
-
#
|
14 |
-
|
|
|
|
|
|
|
15 |
|
16 |
-
|
17 |
-
|
|
|
|
|
|
|
18 |
|
19 |
-
# Autoriser les requêtes Cross-Origin
|
20 |
app.add_middleware(
|
21 |
CORSMiddleware,
|
22 |
allow_origins=["*"],
|
@@ -25,142 +29,111 @@ app.add_middleware(
|
|
25 |
allow_headers=["*"],
|
26 |
)
|
27 |
|
28 |
-
#
|
29 |
-
|
30 |
-
|
31 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
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
|
46 |
-
text += para.text + "\n"
|
47 |
-
return text.strip()
|
48 |
|
49 |
def process_uploaded_file(file: UploadFile) -> str:
|
50 |
-
content
|
51 |
-
extension = file.filename.split(
|
52 |
-
|
53 |
if extension == "pdf":
|
54 |
return extract_text_from_pdf(content)
|
55 |
-
|
56 |
return extract_text_from_docx(content)
|
57 |
-
|
58 |
return content.decode("utf-8").strip()
|
59 |
-
|
60 |
-
|
|
|
|
|
|
|
61 |
|
62 |
-
# Point d'entrée HTML
|
63 |
@app.get("/", response_class=HTMLResponse)
|
64 |
-
async def
|
65 |
-
|
66 |
-
|
|
|
|
|
67 |
|
68 |
-
|
69 |
-
|
70 |
-
async def analyze_file(file: UploadFile = File(...)):
|
71 |
try:
|
72 |
text = process_uploaded_file(file)
|
73 |
-
|
74 |
if len(text) < 20:
|
75 |
-
return {"
|
76 |
-
|
77 |
-
|
78 |
-
|
|
|
79 |
|
80 |
-
|
81 |
-
return JSONResponse(status_code=500, content={"error": f"Erreur lors de l'analyse: {str(e)}"})
|
82 |
|
83 |
-
|
84 |
-
|
85 |
-
async def ask_question(file: UploadFile = File(...), question: str = Form(...)):
|
86 |
try:
|
87 |
-
|
88 |
-
|
89 |
-
|
90 |
-
|
91 |
-
|
92 |
-
|
93 |
-
|
94 |
-
|
95 |
-
|
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 |
-
|
111 |
-
|
112 |
-
|
113 |
-
|
114 |
-
context = text[:3000]
|
115 |
|
116 |
-
|
117 |
-
return {"answer": "Aucune information disponible pour répondre à la question."}
|
118 |
-
|
119 |
-
result = qa_client.question_answering(question=question, context=context)
|
120 |
-
return {"answer": result.get("answer", "Aucune réponse trouvée.")}
|
121 |
|
122 |
-
|
123 |
-
|
124 |
-
|
125 |
-
# Interprétation d'Image
|
126 |
-
@app.post("/interpret_image")
|
127 |
-
async def interpret_image(image: UploadFile = File(...)):
|
128 |
try:
|
129 |
-
|
130 |
-
|
131 |
-
|
132 |
-
|
133 |
-
|
134 |
-
|
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 |
-
# 🔍 Affichage du résultat brut pour débogage
|
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 trouvée."
|
151 |
-
elif isinstance(result, list) and len(result) > 0:
|
152 |
-
description = result[0].get("generated_text", "Description non trouvée.")
|
153 |
-
elif isinstance(result, str):
|
154 |
-
description = result
|
155 |
else:
|
156 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
157 |
|
158 |
-
|
|
|
|
|
159 |
|
160 |
-
|
161 |
-
|
|
|
162 |
|
163 |
-
# Démarrage local
|
164 |
if __name__ == "__main__":
|
165 |
import uvicorn
|
166 |
-
uvicorn.run(
|
|
|
1 |
+
import os, io
|
2 |
from fastapi import FastAPI, UploadFile, File, Form
|
3 |
from fastapi.middleware.cors import CORSMiddleware
|
4 |
+
from fastapi.responses import JSONResponse, HTMLResponse, FileResponse
|
5 |
+
from fastapi.staticfiles import StaticFiles
|
6 |
from huggingface_hub import InferenceClient
|
7 |
from PyPDF2 import PdfReader
|
8 |
from docx import Document
|
9 |
from PIL import Image
|
|
|
10 |
from io import BytesIO
|
|
|
11 |
|
12 |
+
# -----------------------------------------------------------------------------
|
13 |
+
# CONFIGURATION
|
14 |
+
# -----------------------------------------------------------------------------
|
15 |
+
HUGGINGFACE_TOKEN = os.getenv("HF_TOKEN") # injected as a secret
|
16 |
+
PORT = int(os.getenv("PORT", 7860)) # HF Spaces provides it
|
17 |
|
18 |
+
app = FastAPI(
|
19 |
+
title="AI‑Powered Web‑App API",
|
20 |
+
description="Backend endpoints for summarisation, captioning and QA",
|
21 |
+
version="1.1.0",
|
22 |
+
)
|
23 |
|
|
|
24 |
app.add_middleware(
|
25 |
CORSMiddleware,
|
26 |
allow_origins=["*"],
|
|
|
29 |
allow_headers=["*"],
|
30 |
)
|
31 |
|
32 |
+
# Optional: serve static assets from /static (images, css, js)
|
33 |
+
app.mount("/static", StaticFiles(directory="static"), name="static")
|
34 |
+
|
35 |
+
# -----------------------------------------------------------------------------
|
36 |
+
# MODEL CLIENTS (remote HuggingFace Inference API)
|
37 |
+
# -----------------------------------------------------------------------------
|
38 |
+
summary_client = InferenceClient("facebook/bart-large-cnn", token=HUGGINGFACE_TOKEN)
|
39 |
+
qa_client = InferenceClient("deepset/roberta-base-squad2", token=HUGGINGFACE_TOKEN)
|
40 |
+
image_caption_client = InferenceClient("nlpconnect/vit-gpt2-image-captioning", token=HUGGINGFACE_TOKEN)
|
41 |
+
|
42 |
+
# -----------------------------------------------------------------------------
|
43 |
+
# UTILITY FUNCTIONS
|
44 |
+
# -----------------------------------------------------------------------------
|
45 |
|
|
|
46 |
def extract_text_from_pdf(content: bytes) -> str:
|
|
|
47 |
reader = PdfReader(io.BytesIO(content))
|
48 |
+
return "\n".join(page.extract_text() or "" for page in reader.pages).strip()
|
|
|
|
|
|
|
49 |
|
50 |
def extract_text_from_docx(content: bytes) -> str:
|
|
|
51 |
doc = Document(io.BytesIO(content))
|
52 |
+
return "\n".join(p.text for p in doc.paragraphs).strip()
|
|
|
|
|
53 |
|
54 |
def process_uploaded_file(file: UploadFile) -> str:
|
55 |
+
content = file.file.read()
|
56 |
+
extension = file.filename.split(".")[-1].lower()
|
|
|
57 |
if extension == "pdf":
|
58 |
return extract_text_from_pdf(content)
|
59 |
+
if extension == "docx":
|
60 |
return extract_text_from_docx(content)
|
61 |
+
if extension == "txt":
|
62 |
return content.decode("utf-8").strip()
|
63 |
+
raise ValueError("Unsupported file type")
|
64 |
+
|
65 |
+
# -----------------------------------------------------------------------------
|
66 |
+
# ROUTES
|
67 |
+
# -----------------------------------------------------------------------------
|
68 |
|
|
|
69 |
@app.get("/", response_class=HTMLResponse)
|
70 |
+
async def serve_index():
|
71 |
+
"""Send the frontend HTML."""
|
72 |
+
return FileResponse("index.html")
|
73 |
+
|
74 |
+
# ---------- Summarisation -----------------------------------------------------
|
75 |
|
76 |
+
@app.post("/api/summarize")
|
77 |
+
async def summarize_document(file: UploadFile = File(...)):
|
|
|
78 |
try:
|
79 |
text = process_uploaded_file(file)
|
|
|
80 |
if len(text) < 20:
|
81 |
+
return {"result": "Document too short to summarise."}
|
82 |
+
summary_text = summary_client.summarization(text[:3000])
|
83 |
+
return {"result": str(summary_text)}
|
84 |
+
except Exception as exc:
|
85 |
+
return JSONResponse(status_code=500, content={"error": f"Analyse failure: {exc}"})
|
86 |
|
87 |
+
# ---------- Image Caption -----------------------------------------------------
|
|
|
88 |
|
89 |
+
@app.post("/api/caption")
|
90 |
+
async def caption_image(file: UploadFile = File(...)):
|
|
|
91 |
try:
|
92 |
+
image_bytes = await file.read()
|
93 |
+
image_pil = Image.open(io.BytesIO(image_bytes)).convert("RGB")
|
94 |
+
image_pil.thumbnail((1024, 1024))
|
95 |
+
buf = BytesIO(); image_pil.save(buf, format="JPEG"); img = buf.getvalue()
|
96 |
+
result = image_caption_client.image_to_text(img)
|
97 |
+
if isinstance(result, dict):
|
98 |
+
caption = result.get("generated_text") or result.get("caption") or "No caption found."
|
99 |
+
elif isinstance(result, list):
|
100 |
+
caption = result[0].get("generated_text", "No caption found.")
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
101 |
else:
|
102 |
+
caption = str(result)
|
103 |
+
return {"result": str(caption)}
|
104 |
+
except Exception as exc:
|
105 |
+
return JSONResponse(status_code=500, content={"error": f"Caption failure: {exc}"})
|
|
|
106 |
|
107 |
+
# ---------- Question Answering ----------------------------------------------
|
|
|
|
|
|
|
|
|
108 |
|
109 |
+
@app.post("/api/qa")
|
110 |
+
async def question_answering(file: UploadFile = File(...), question: str = Form(...)):
|
|
|
|
|
|
|
|
|
111 |
try:
|
112 |
+
if file.content_type.startswith("image/"):
|
113 |
+
image_bytes = await file.read()
|
114 |
+
pil = Image.open(io.BytesIO(image_bytes)).convert("RGB"); pil.thumbnail((1024, 1024))
|
115 |
+
buf = BytesIO(); pil.save(buf, format="JPEG"); img = buf.getvalue()
|
116 |
+
res = image_caption_client.image_to_text(img)
|
117 |
+
context = res.get("generated_text") if isinstance(res, dict) else str(res)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
118 |
else:
|
119 |
+
context = process_uploaded_file(file)[:3000]
|
120 |
+
if not context:
|
121 |
+
return {"result": "No context – cannot answer."}
|
122 |
+
answer = qa_client.question_answering(question=question, context=context)
|
123 |
+
return {"result": str(answer.get("answer", "No answer found."))}
|
124 |
+
except Exception as exc:
|
125 |
+
return JSONResponse(status_code=500, content={"error": f"QA failure: {exc}"})
|
126 |
+
|
127 |
+
# ---------- Health check ------------------------------------------------------
|
128 |
|
129 |
+
@app.get("/api/health")
|
130 |
+
async def health():
|
131 |
+
return {"status": "healthy", "hf_token_set": bool(HUGGINGFACE_TOKEN)}
|
132 |
|
133 |
+
# -----------------------------------------------------------------------------
|
134 |
+
# ENTRYPOINT
|
135 |
+
# -----------------------------------------------------------------------------
|
136 |
|
|
|
137 |
if __name__ == "__main__":
|
138 |
import uvicorn
|
139 |
+
uvicorn.run(app, host="0.0.0.0", port=PORT)
|