Update app/main.py
Browse files- app/main.py +22 -19
app/main.py
CHANGED
|
@@ -8,6 +8,7 @@ import re
|
|
| 8 |
import random
|
| 9 |
from datetime import datetime
|
| 10 |
from jinja2 import Template
|
|
|
|
| 11 |
import openai
|
| 12 |
|
| 13 |
# —— 环境变量读取 —— #
|
|
@@ -21,34 +22,36 @@ def get_access_token(api_key, secret_key):
|
|
| 21 |
resp = requests.post(
|
| 22 |
"https://aip.baidubce.com/oauth/2.0/token",
|
| 23 |
params={
|
| 24 |
-
"grant_type":"client_credentials",
|
| 25 |
"client_id": api_key,
|
| 26 |
"client_secret": secret_key
|
| 27 |
}
|
| 28 |
)
|
|
|
|
| 29 |
return resp.json().get("access_token")
|
| 30 |
|
| 31 |
def ocr_image(image_bytes, token):
|
|
|
|
| 32 |
img_b64 = base64.b64encode(image_bytes).decode()
|
| 33 |
-
|
| 34 |
-
|
| 35 |
-
|
| 36 |
-
|
| 37 |
-
)
|
|
|
|
| 38 |
return resp.json().get("words_result", [])
|
| 39 |
|
| 40 |
# —— 文本高亮 —— #
|
| 41 |
def highlight_brackets(text):
|
| 42 |
-
# 中括号绿色;小括号红色
|
| 43 |
text = re.sub(r'\[([^\[\]]+)\]', r'<span class="highlight-bracket-green">\1</span>', text)
|
| 44 |
text = re.sub(r'\(([^\(\)]+)\)', r'<span class="highlight-bracket">\1</span>', text)
|
| 45 |
return text.replace("\n", "<br>")
|
| 46 |
|
| 47 |
# —— 主处理函数 —— #
|
| 48 |
-
def process(
|
| 49 |
# 1. OCR
|
| 50 |
token = get_access_token(BAIDU_API_KEY, BAIDU_SECRET_KEY)
|
| 51 |
-
words = ocr_image(
|
| 52 |
essay_text = "\n".join([w["words"] for w in words])
|
| 53 |
|
| 54 |
# 2. 格式化原文
|
|
@@ -59,7 +62,7 @@ def process(image):
|
|
| 59 |
)
|
| 60 |
fm = openai.ChatCompletion.create(
|
| 61 |
model="gpt-4o-mini",
|
| 62 |
-
messages=[{"role":"user","content":fmt_prompt}]
|
| 63 |
)
|
| 64 |
revised = fm.choices[0].message.content
|
| 65 |
|
|
@@ -70,7 +73,7 @@ def process(image):
|
|
| 70 |
)
|
| 71 |
cm = openai.ChatCompletion.create(
|
| 72 |
model="gpt-4o-mini",
|
| 73 |
-
messages=[{"role":"user","content":corr_prompt}]
|
| 74 |
)
|
| 75 |
corrected = cm.choices[0].message.content
|
| 76 |
|
|
@@ -81,7 +84,7 @@ def process(image):
|
|
| 81 |
)
|
| 82 |
rm = openai.ChatCompletion.create(
|
| 83 |
model="gpt-4o-mini",
|
| 84 |
-
messages=[{"role":"user","content":review_prompt}]
|
| 85 |
)
|
| 86 |
review = rm.choices[0].message.content
|
| 87 |
|
|
@@ -92,7 +95,7 @@ def process(image):
|
|
| 92 |
)
|
| 93 |
rr = openai.ChatCompletion.create(
|
| 94 |
model="gpt-4o-mini",
|
| 95 |
-
messages=[{"role":"user","content":rate_prompt}]
|
| 96 |
)
|
| 97 |
rating = rr.choices[0].message.content
|
| 98 |
|
|
@@ -102,7 +105,7 @@ def process(image):
|
|
| 102 |
)
|
| 103 |
wm = openai.ChatCompletion.create(
|
| 104 |
model="gpt-4o-mini",
|
| 105 |
-
messages=[{"role":"user","content":rewrite_prompt}]
|
| 106 |
)
|
| 107 |
perfect = wm.choices[0].message.content
|
| 108 |
|
|
@@ -125,19 +128,19 @@ def process(image):
|
|
| 125 |
full_html = tpl.render(code=code, content=html_content)
|
| 126 |
|
| 127 |
# 写文件
|
| 128 |
-
|
| 129 |
-
|
| 130 |
-
|
| 131 |
with open(html_path, "w", encoding="utf-8") as f:
|
| 132 |
f.write(full_html)
|
| 133 |
-
pdfkit.from_string(full_html, pdf_path, options={"enable-local-file-access":""})
|
| 134 |
|
| 135 |
return full_html, html_path, pdf_path
|
| 136 |
|
| 137 |
# —— Gradio 接口 —— #
|
| 138 |
with gr.Blocks(title="英语作文批改") as demo:
|
| 139 |
gr.Markdown("## 上传英语作文照片,等待批改完成后下载 HTML 或 PDF")
|
| 140 |
-
image_in = gr.
|
| 141 |
output_html = gr.HTML()
|
| 142 |
btn = gr.Button("开始批改")
|
| 143 |
file_html = gr.File(label="下载 HTML")
|
|
|
|
| 8 |
import random
|
| 9 |
from datetime import datetime
|
| 10 |
from jinja2 import Template
|
| 11 |
+
import urllib.parse
|
| 12 |
import openai
|
| 13 |
|
| 14 |
# —— 环境变量读取 —— #
|
|
|
|
| 22 |
resp = requests.post(
|
| 23 |
"https://aip.baidubce.com/oauth/2.0/token",
|
| 24 |
params={
|
| 25 |
+
"grant_type": "client_credentials",
|
| 26 |
"client_id": api_key,
|
| 27 |
"client_secret": secret_key
|
| 28 |
}
|
| 29 |
)
|
| 30 |
+
resp.raise_for_status()
|
| 31 |
return resp.json().get("access_token")
|
| 32 |
|
| 33 |
def ocr_image(image_bytes, token):
|
| 34 |
+
# Base64 encode and URL-encode the image
|
| 35 |
img_b64 = base64.b64encode(image_bytes).decode()
|
| 36 |
+
img_encoded = urllib.parse.quote_plus(img_b64)
|
| 37 |
+
url = f"https://aip.baidubce.com/rest/2.0/ocr/v1/handwriting?access_token={token}"
|
| 38 |
+
headers = {"Content-Type": "application/x-www-form-urlencoded"}
|
| 39 |
+
data = f"image={img_encoded}&language_type=ENG"
|
| 40 |
+
resp = requests.post(url, headers=headers, data=data)
|
| 41 |
+
resp.raise_for_status()
|
| 42 |
return resp.json().get("words_result", [])
|
| 43 |
|
| 44 |
# —— 文本高亮 —— #
|
| 45 |
def highlight_brackets(text):
|
|
|
|
| 46 |
text = re.sub(r'\[([^\[\]]+)\]', r'<span class="highlight-bracket-green">\1</span>', text)
|
| 47 |
text = re.sub(r'\(([^\(\)]+)\)', r'<span class="highlight-bracket">\1</span>', text)
|
| 48 |
return text.replace("\n", "<br>")
|
| 49 |
|
| 50 |
# —— 主处理函数 —— #
|
| 51 |
+
def process(image_bytes: bytes):
|
| 52 |
# 1. OCR
|
| 53 |
token = get_access_token(BAIDU_API_KEY, BAIDU_SECRET_KEY)
|
| 54 |
+
words = ocr_image(image_bytes, token)
|
| 55 |
essay_text = "\n".join([w["words"] for w in words])
|
| 56 |
|
| 57 |
# 2. 格式化原文
|
|
|
|
| 62 |
)
|
| 63 |
fm = openai.ChatCompletion.create(
|
| 64 |
model="gpt-4o-mini",
|
| 65 |
+
messages=[{"role": "user", "content": fmt_prompt}]
|
| 66 |
)
|
| 67 |
revised = fm.choices[0].message.content
|
| 68 |
|
|
|
|
| 73 |
)
|
| 74 |
cm = openai.ChatCompletion.create(
|
| 75 |
model="gpt-4o-mini",
|
| 76 |
+
messages=[{"role": "user", "content": corr_prompt}]
|
| 77 |
)
|
| 78 |
corrected = cm.choices[0].message.content
|
| 79 |
|
|
|
|
| 84 |
)
|
| 85 |
rm = openai.ChatCompletion.create(
|
| 86 |
model="gpt-4o-mini",
|
| 87 |
+
messages=[{"role": "user", "content": review_prompt}]
|
| 88 |
)
|
| 89 |
review = rm.choices[0].message.content
|
| 90 |
|
|
|
|
| 95 |
)
|
| 96 |
rr = openai.ChatCompletion.create(
|
| 97 |
model="gpt-4o-mini",
|
| 98 |
+
messages=[{"role": "user", "content": rate_prompt}]
|
| 99 |
)
|
| 100 |
rating = rr.choices[0].message.content
|
| 101 |
|
|
|
|
| 105 |
)
|
| 106 |
wm = openai.ChatCompletion.create(
|
| 107 |
model="gpt-4o-mini",
|
| 108 |
+
messages=[{"role": "user", "content": rewrite_prompt}]
|
| 109 |
)
|
| 110 |
perfect = wm.choices[0].message.content
|
| 111 |
|
|
|
|
| 128 |
full_html = tpl.render(code=code, content=html_content)
|
| 129 |
|
| 130 |
# 写文件
|
| 131 |
+
os.makedirs("app/output", exist_ok=True)
|
| 132 |
+
html_path = f"app/output/{code}.html"
|
| 133 |
+
pdf_path = f"app/output/{code}.pdf"
|
| 134 |
with open(html_path, "w", encoding="utf-8") as f:
|
| 135 |
f.write(full_html)
|
| 136 |
+
pdfkit.from_string(full_html, pdf_path, options={"enable-local-file-access": ""})
|
| 137 |
|
| 138 |
return full_html, html_path, pdf_path
|
| 139 |
|
| 140 |
# —— Gradio 接口 —— #
|
| 141 |
with gr.Blocks(title="英语作文批改") as demo:
|
| 142 |
gr.Markdown("## 上传英语作文照片,等待批改完成后下载 HTML 或 PDF")
|
| 143 |
+
image_in = gr.Image(type="bytes", label="上传照片")
|
| 144 |
output_html = gr.HTML()
|
| 145 |
btn = gr.Button("开始批改")
|
| 146 |
file_html = gr.File(label="下载 HTML")
|