Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
@@ -2,22 +2,23 @@ import base64
|
|
2 |
from io import BytesIO
|
3 |
import json
|
4 |
import os
|
|
|
5 |
from openai import OpenAI
|
6 |
from dotenv import load_dotenv
|
7 |
from typhoon_ocr import prepare_ocr_messages
|
8 |
import gradio as gr
|
9 |
from PIL import Image
|
10 |
|
11 |
-
# โหลด environment variables
|
12 |
load_dotenv()
|
13 |
|
14 |
-
# ตั้งค่า OpenAI
|
15 |
openai = OpenAI(
|
16 |
base_url=os.getenv("TYPHOON_BASE_URL"),
|
17 |
api_key=os.getenv("TYPHOON_API_KEY")
|
18 |
)
|
19 |
|
20 |
-
# ตั้งค่า Theme (
|
21 |
theme = gr.themes.Soft(
|
22 |
primary_hue=gr.themes.Color(
|
23 |
c50="#f7f7fd",
|
@@ -40,61 +41,122 @@ theme = gr.themes.Soft(
|
|
40 |
OUTPUT_FILE = "ocr_results.txt"
|
41 |
|
42 |
def save_ocr_result(text):
|
43 |
-
"""บันทึกผลลัพธ์ OCR แบบต่อเนื่องในไฟล์เดียว
|
44 |
with open(OUTPUT_FILE, "a", encoding="utf-8") as f:
|
45 |
f.write(text + "\n\n")
|
46 |
return OUTPUT_FILE
|
47 |
|
48 |
def clear_output_file():
|
49 |
-
"""ล้างไฟล์ผลลัพธ์เก่า
|
50 |
with open(OUTPUT_FILE, "w", encoding="utf-8") as f:
|
51 |
f.write("")
|
52 |
|
53 |
-
def
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
54 |
if pdf_or_image_file is None:
|
55 |
return None, "No file uploaded"
|
56 |
|
57 |
orig_filename = pdf_or_image_file.name
|
|
|
|
|
58 |
|
59 |
try:
|
60 |
-
#
|
61 |
-
|
62 |
-
|
63 |
-
|
64 |
-
|
65 |
-
|
66 |
-
|
67 |
-
|
68 |
-
|
69 |
-
|
70 |
-
|
71 |
-
|
72 |
-
|
73 |
-
|
74 |
-
|
75 |
-
|
76 |
-
|
77 |
-
|
78 |
-
|
79 |
-
|
80 |
-
|
81 |
-
|
82 |
-
|
83 |
-
|
84 |
-
|
85 |
-
|
86 |
-
|
87 |
-
|
88 |
-
|
89 |
-
|
90 |
-
|
91 |
-
|
92 |
-
|
93 |
-
|
94 |
-
|
95 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
96 |
|
97 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
98 |
|
99 |
except Exception as e:
|
100 |
return None, f"Error processing file: {str(e)}", None
|
@@ -148,9 +210,6 @@ with gr.Blocks(theme=theme) as demo:
|
|
148 |
font-size: 12px;
|
149 |
}
|
150 |
"""
|
151 |
-
# เลือกเพจ
|
152 |
-
page_number = gr.Number(label="📄 Page Number (for PDFs only)", value=1, minimum=1, step=1)
|
153 |
-
|
154 |
# ปุ่มรัน
|
155 |
run_button = gr.Button("🚀 Run")
|
156 |
|
@@ -167,11 +226,11 @@ with gr.Blocks(theme=theme) as demo:
|
|
167 |
# เชื่อมต่อ UI กับฟังก์ชัน
|
168 |
run_button.click(
|
169 |
fn=process_pdf,
|
170 |
-
inputs=[pdf_input, task_dropdown
|
171 |
outputs=[image_output, markdown_output, download_button]
|
172 |
)
|
173 |
|
174 |
-
#
|
175 |
clear_output_file()
|
176 |
|
177 |
# รันแอป
|
|
|
2 |
from io import BytesIO
|
3 |
import json
|
4 |
import os
|
5 |
+
import PyPDF2 # เพิ่มไลบรารีสำหรับอ่าน PDF
|
6 |
from openai import OpenAI
|
7 |
from dotenv import load_dotenv
|
8 |
from typhoon_ocr import prepare_ocr_messages
|
9 |
import gradio as gr
|
10 |
from PIL import Image
|
11 |
|
12 |
+
# โหลด environment variables
|
13 |
load_dotenv()
|
14 |
|
15 |
+
# ตั้งค่า OpenAI API
|
16 |
openai = OpenAI(
|
17 |
base_url=os.getenv("TYPHOON_BASE_URL"),
|
18 |
api_key=os.getenv("TYPHOON_API_KEY")
|
19 |
)
|
20 |
|
21 |
+
# ตั้งค่า Theme (เดิม)
|
22 |
theme = gr.themes.Soft(
|
23 |
primary_hue=gr.themes.Color(
|
24 |
c50="#f7f7fd",
|
|
|
41 |
OUTPUT_FILE = "ocr_results.txt"
|
42 |
|
43 |
def save_ocr_result(text):
|
44 |
+
"""บันทึกผลลัพธ์ OCR แบบต่อเนื่องในไฟล์เดียว"""
|
45 |
with open(OUTPUT_FILE, "a", encoding="utf-8") as f:
|
46 |
f.write(text + "\n\n")
|
47 |
return OUTPUT_FILE
|
48 |
|
49 |
def clear_output_file():
|
50 |
+
"""ล้างไฟล์ผลลัพธ์เก่า"""
|
51 |
with open(OUTPUT_FILE, "w", encoding="utf-8") as f:
|
52 |
f.write("")
|
53 |
|
54 |
+
def get_pdf_page_count(pdf_path):
|
55 |
+
"""หาจำนวนหน้าของ PDF"""
|
56 |
+
with open(pdf_path, 'rb') as f:
|
57 |
+
reader = PyPDF2.PdfReader(f)
|
58 |
+
return len(reader.pages)
|
59 |
+
return 0
|
60 |
+
|
61 |
+
def process_pdf(pdf_or_image_file, task_type):
|
62 |
if pdf_or_image_file is None:
|
63 |
return None, "No file uploaded"
|
64 |
|
65 |
orig_filename = pdf_or_image_file.name
|
66 |
+
combined_text = ""
|
67 |
+
image_pil = None # ใช้เก็บภาพหน้าแรกของ PDF
|
68 |
|
69 |
try:
|
70 |
+
# ตรวจสอบว่าเป็น PDF หรือไม่
|
71 |
+
if orig_filename.lower().endswith(".pdf"):
|
72 |
+
total_pages = get_pdf_page_count(orig_filename)
|
73 |
+
|
74 |
+
if total_pages == 0:
|
75 |
+
return None, "ไม่สามารถอ่านจำนวนหน้าของ PDF ได้"
|
76 |
+
|
77 |
+
# ประมวลผลทุกหน้า
|
78 |
+
for page_num in range(1, total_pages + 1):
|
79 |
+
# เตรียมข้อมูลสำหรับ OCR
|
80 |
+
messages = prepare_ocr_messages(
|
81 |
+
pdf_or_image_path=orig_filename,
|
82 |
+
task_type=task_type,
|
83 |
+
target_image_dim=1800,
|
84 |
+
target_text_length=8000,
|
85 |
+
page_num=page_num
|
86 |
+
)
|
87 |
+
|
88 |
+
# ดึงภาพหน้าแรก
|
89 |
+
if page_num == 1:
|
90 |
+
image_url = messages[0]["content"][1]["image_url"]["url"]
|
91 |
+
image_base64 = image_url.replace("data:image/png;base64,", "")
|
92 |
+
image_pil = Image.open(BytesIO(base64.b64decode(image_base64)))
|
93 |
+
|
94 |
+
# ส่งไป API
|
95 |
+
response = openai.chat.completions.create(
|
96 |
+
model=os.getenv("TYPHOON_OCR_MODEL"),
|
97 |
+
messages=messages,
|
98 |
+
max_tokens=16384,
|
99 |
+
extra_body={
|
100 |
+
"repetition_penalty": 1.2,
|
101 |
+
"temperature": 0.1,
|
102 |
+
"top_p": 0.6,
|
103 |
+
},
|
104 |
+
)
|
105 |
+
text_output = response.choices[0].message.content
|
106 |
+
|
107 |
+
# ดึง natural_text
|
108 |
+
try:
|
109 |
+
json_data = json.loads(text_output)
|
110 |
+
markdown_out = json_data.get('natural_text', "").replace("<figure>", "").replace("</figure>", "")
|
111 |
+
except Exception as e:
|
112 |
+
markdown_out = f"⚠️ Could not extract `natural_text` from output.\nError: {str(e)}"
|
113 |
+
|
114 |
+
# รวมผลลัพธ์ทุกหน้า
|
115 |
+
combined_text += f"[Page {page_num}]\n{markdown_out}\n\n"
|
116 |
+
|
117 |
+
# บันทึกผลลัพธ์ทั้งหมดลงไฟล์
|
118 |
+
save_ocr_result(combined_text)
|
119 |
+
return image_pil, combined_text, gr.File.update(value=OUTPUT_FILE)
|
120 |
|
121 |
+
# หากเป็นไฟล์ภาพ
|
122 |
+
else:
|
123 |
+
# ประมวลผลหน้าเดียว
|
124 |
+
messages = prepare_ocr_messages(
|
125 |
+
pdf_or_image_path=orig_filename,
|
126 |
+
task_type=task_type,
|
127 |
+
target_image_dim=1800,
|
128 |
+
target_text_length=8000,
|
129 |
+
page_num=1
|
130 |
+
)
|
131 |
+
|
132 |
+
# ดึงภาพ
|
133 |
+
image_url = messages[0]["content"][1]["image_url"]["url"]
|
134 |
+
image_base64 = image_url.replace("data:image/png;base64,", "")
|
135 |
+
image_pil = Image.open(BytesIO(base64.b64decode(image_base64)))
|
136 |
+
|
137 |
+
# ส่งไป API
|
138 |
+
response = openai.chat.completions.create(
|
139 |
+
model=os.getenv("TYPHOON_OCR_MODEL"),
|
140 |
+
messages=messages,
|
141 |
+
max_tokens=16384,
|
142 |
+
extra_body={
|
143 |
+
"repetition_penalty": 1.2,
|
144 |
+
"temperature": 0.1,
|
145 |
+
"top_p": 0.6,
|
146 |
+
},
|
147 |
+
)
|
148 |
+
text_output = response.choices[0].message.content
|
149 |
+
|
150 |
+
# ดึง natural_text
|
151 |
+
try:
|
152 |
+
json_data = json.loads(text_output)
|
153 |
+
markdown_out = json_data.get('natural_text', "").replace("<figure>", "").replace("</figure>", "")
|
154 |
+
except Exception as e:
|
155 |
+
markdown_out = f"⚠️ Could not extract `natural_text` from output.\nError: {str(e)}"
|
156 |
+
|
157 |
+
# บันทึกผลลัพธ์ลงไฟล์
|
158 |
+
save_ocr_result(markdown_out)
|
159 |
+
return image_pil, markdown_out, gr.File.update(value=OUTPUT_FILE)
|
160 |
|
161 |
except Exception as e:
|
162 |
return None, f"Error processing file: {str(e)}", None
|
|
|
210 |
font-size: 12px;
|
211 |
}
|
212 |
"""
|
|
|
|
|
|
|
213 |
# ปุ่มรัน
|
214 |
run_button = gr.Button("🚀 Run")
|
215 |
|
|
|
226 |
# เชื่อมต่อ UI กับฟังก์ชัน
|
227 |
run_button.click(
|
228 |
fn=process_pdf,
|
229 |
+
inputs=[pdf_input, task_dropdown],
|
230 |
outputs=[image_output, markdown_output, download_button]
|
231 |
)
|
232 |
|
233 |
+
# เริ่มต้นใหม่ (ล้างไฟล์ผลลัพธ์เก่า)
|
234 |
clear_output_file()
|
235 |
|
236 |
# รันแอป
|