Update app/main.py
Browse files- app/main.py +15 -25
app/main.py
CHANGED
@@ -17,15 +17,14 @@ from openai import OpenAI
|
|
17 |
# —— 环境变量读取 —— #
|
18 |
BAIDU_API_KEY = os.getenv("BAIDU_API_KEY", "")
|
19 |
BAIDU_SECRET_KEY = os.getenv("BAIDU_SECRET_KEY", "")
|
20 |
-
OPENAI_API_KEY = os.getenv("OPENAI_API_KEY", "")
|
21 |
|
22 |
-
# —— 指定代理站
|
23 |
-
openai.api_key =
|
24 |
-
openai.api_base = "https://
|
25 |
|
26 |
# —— 初始化 v1 客户端 —— #
|
27 |
-
client = OpenAI(api_key=
|
28 |
-
#
|
29 |
client.api_base = openai.api_base
|
30 |
|
31 |
# —— Baidu OCR —— #
|
@@ -43,14 +42,11 @@ def get_access_token(api_key, secret_key):
|
|
43 |
|
44 |
def ocr_image(image_bytes: bytes, token: str):
|
45 |
img_b64 = base64.b64encode(image_bytes).decode()
|
46 |
-
img_encoded = quote_plus(img_b64)
|
47 |
resp = requests.post(
|
48 |
f"https://aip.baidubce.com/rest/2.0/ocr/v1/handwriting?access_token={token}",
|
49 |
headers={"Content-Type": "application/x-www-form-urlencoded"},
|
50 |
-
data={
|
51 |
-
"image": img_encoded,
|
52 |
-
"language_type": "ENG"
|
53 |
-
}
|
54 |
)
|
55 |
resp.raise_for_status()
|
56 |
return resp.json().get("words_result", [])
|
@@ -63,17 +59,15 @@ def highlight_brackets(text: str) -> str:
|
|
63 |
|
64 |
# —— 主处理函数 —— #
|
65 |
def process(image_pil):
|
66 |
-
# —— PIL 转 bytes —— #
|
67 |
buf = BytesIO()
|
68 |
image_pil.save(buf, format="PNG")
|
69 |
image_bytes = buf.getvalue()
|
70 |
|
71 |
-
# —— OCR 提取原文 —— #
|
72 |
token = get_access_token(BAIDU_API_KEY, BAIDU_SECRET_KEY)
|
73 |
words = ocr_image(image_bytes, token)
|
74 |
essay_text = "\n".join([w["words"] for w in words])
|
75 |
|
76 |
-
#
|
77 |
fmt_prompt = (
|
78 |
"请帮我整理下面的英语作文文本格式,只整理英文正文部分,"
|
79 |
"保证原汁原味(明显错误空格换行、乱码、非常用字符改正),"
|
@@ -85,7 +79,7 @@ def process(image_pil):
|
|
85 |
)
|
86 |
revised = fm_resp.choices[0].message.content
|
87 |
|
88 |
-
#
|
89 |
corr_prompt = (
|
90 |
"请帮我把下面的英语作文的语法错误改正,输出改正后的文章,"
|
91 |
"原文错误用()括起来,修改部分用[]括起来:\n\n" + revised
|
@@ -96,7 +90,7 @@ def process(image_pil):
|
|
96 |
)
|
97 |
corrected = cm_resp.choices[0].message.content
|
98 |
|
99 |
-
#
|
100 |
review_prompt = (
|
101 |
"下面是一份已经批改过的英语作文,请根据批注给出逐条批改意见:\n\n" + corrected
|
102 |
)
|
@@ -106,7 +100,7 @@ def process(image_pil):
|
|
106 |
)
|
107 |
review = rm_resp.choices[0].message.content
|
108 |
|
109 |
-
#
|
110 |
rate_prompt = (
|
111 |
"请按照 IELTS/CEFR 写作评价体系,从语言通顺度、连贯度、词汇与语法三维度打分并给出原因:\n\n" + revised
|
112 |
)
|
@@ -116,7 +110,7 @@ def process(image_pil):
|
|
116 |
)
|
117 |
rating = rr_resp.choices[0].message.content
|
118 |
|
119 |
-
#
|
120 |
rewrite_prompt = (
|
121 |
"请使用优秀表达重写下面这篇作文,加粗可供学习的部分:\n\n" + revised
|
122 |
)
|
@@ -126,7 +120,7 @@ def process(image_pil):
|
|
126 |
)
|
127 |
perfect = wm_resp.choices[0].message.content
|
128 |
|
129 |
-
#
|
130 |
code = f"{random.randint(0,9999):04}-{datetime.now().strftime('%Y%m%d%H%M%S')}"
|
131 |
tpl_path = os.path.join("app", "templates", "base.html")
|
132 |
with open(tpl_path, encoding="utf-8") as f:
|
@@ -141,7 +135,7 @@ def process(image_pil):
|
|
141 |
)
|
142 |
full_html = tpl.render(code=code, content=html_content)
|
143 |
|
144 |
-
#
|
145 |
output_dir = os.path.join("app", "output")
|
146 |
os.makedirs(output_dir, exist_ok=True)
|
147 |
html_path = os.path.join(output_dir, f"{code}.html")
|
@@ -149,11 +143,7 @@ def process(image_pil):
|
|
149 |
|
150 |
with open(html_path, "w", encoding="utf-8") as f:
|
151 |
f.write(full_html)
|
152 |
-
|
153 |
-
pdfkit.from_string(
|
154 |
-
full_html, pdf_path,
|
155 |
-
options={"enable-local-file-access": ""}
|
156 |
-
)
|
157 |
|
158 |
return full_html, html_path, pdf_path
|
159 |
|
|
|
17 |
# —— 环境变量读取 —— #
|
18 |
BAIDU_API_KEY = os.getenv("BAIDU_API_KEY", "")
|
19 |
BAIDU_SECRET_KEY = os.getenv("BAIDU_SECRET_KEY", "")
|
|
|
20 |
|
21 |
+
# —— 指定代理站 & API Key —— #
|
22 |
+
openai.api_key = "Magic-123456"
|
23 |
+
openai.api_base = "https://freeapi.oaiopen.com/v1"
|
24 |
|
25 |
# —— 初始化 v1 客户端 —— #
|
26 |
+
client = OpenAI(api_key=openai.api_key)
|
27 |
+
# 同步 base_url
|
28 |
client.api_base = openai.api_base
|
29 |
|
30 |
# —— Baidu OCR —— #
|
|
|
42 |
|
43 |
def ocr_image(image_bytes: bytes, token: str):
|
44 |
img_b64 = base64.b64encode(image_bytes).decode()
|
45 |
+
img_encoded = quote_plus(img_b64)
|
46 |
resp = requests.post(
|
47 |
f"https://aip.baidubce.com/rest/2.0/ocr/v1/handwriting?access_token={token}",
|
48 |
headers={"Content-Type": "application/x-www-form-urlencoded"},
|
49 |
+
data={"image": img_encoded, "language_type": "ENG"}
|
|
|
|
|
|
|
50 |
)
|
51 |
resp.raise_for_status()
|
52 |
return resp.json().get("words_result", [])
|
|
|
59 |
|
60 |
# —— 主处理函数 —— #
|
61 |
def process(image_pil):
|
|
|
62 |
buf = BytesIO()
|
63 |
image_pil.save(buf, format="PNG")
|
64 |
image_bytes = buf.getvalue()
|
65 |
|
|
|
66 |
token = get_access_token(BAIDU_API_KEY, BAIDU_SECRET_KEY)
|
67 |
words = ocr_image(image_bytes, token)
|
68 |
essay_text = "\n".join([w["words"] for w in words])
|
69 |
|
70 |
+
# 1. 格式化原文
|
71 |
fmt_prompt = (
|
72 |
"请帮我整理下面的英语作文文本格式,只整理英文正文部分,"
|
73 |
"保证原汁原味(明显错误空格换行、乱码、非常用字符改正),"
|
|
|
79 |
)
|
80 |
revised = fm_resp.choices[0].message.content
|
81 |
|
82 |
+
# 2. 批改
|
83 |
corr_prompt = (
|
84 |
"请帮我把下面的英语作文的语法错误改正,输出改正后的文章,"
|
85 |
"原文错误用()括起来,修改部分用[]括起来:\n\n" + revised
|
|
|
90 |
)
|
91 |
corrected = cm_resp.choices[0].message.content
|
92 |
|
93 |
+
# 3. 批改意见
|
94 |
review_prompt = (
|
95 |
"下面是一份已经批改过的英语作文,请根据批注给出逐条批改意见:\n\n" + corrected
|
96 |
)
|
|
|
100 |
)
|
101 |
review = rm_resp.choices[0].message.content
|
102 |
|
103 |
+
# 4. 评分
|
104 |
rate_prompt = (
|
105 |
"请按照 IELTS/CEFR 写作评价体系,从语言通顺度、连贯度、词汇与语法三维度打分并给出原因:\n\n" + revised
|
106 |
)
|
|
|
110 |
)
|
111 |
rating = rr_resp.choices[0].message.content
|
112 |
|
113 |
+
# 5. 优秀范文
|
114 |
rewrite_prompt = (
|
115 |
"请使用优秀表达重写下面这篇作文,加粗可供学习的部分:\n\n" + revised
|
116 |
)
|
|
|
120 |
)
|
121 |
perfect = wm_resp.choices[0].message.content
|
122 |
|
123 |
+
# 6. 渲染 HTML
|
124 |
code = f"{random.randint(0,9999):04}-{datetime.now().strftime('%Y%m%d%H%M%S')}"
|
125 |
tpl_path = os.path.join("app", "templates", "base.html")
|
126 |
with open(tpl_path, encoding="utf-8") as f:
|
|
|
135 |
)
|
136 |
full_html = tpl.render(code=code, content=html_content)
|
137 |
|
138 |
+
# 7. 写文件并返回
|
139 |
output_dir = os.path.join("app", "output")
|
140 |
os.makedirs(output_dir, exist_ok=True)
|
141 |
html_path = os.path.join(output_dir, f"{code}.html")
|
|
|
143 |
|
144 |
with open(html_path, "w", encoding="utf-8") as f:
|
145 |
f.write(full_html)
|
146 |
+
pdfkit.from_string(full_html, pdf_path, options={"enable-local-file-access": ""})
|
|
|
|
|
|
|
|
|
147 |
|
148 |
return full_html, html_path, pdf_path
|
149 |
|