Spaces:
Running
Running
Update app.py
Browse files
app.py
CHANGED
@@ -5,47 +5,72 @@ from nltk import download, sent_tokenize
|
|
5 |
import google.generativeai as genai
|
6 |
import os
|
7 |
import re
|
|
|
|
|
|
|
8 |
|
9 |
# Download NLTK data
|
10 |
download('punkt')
|
11 |
download('punkt_tab')
|
12 |
|
13 |
-
# Configure Gemini API using
|
14 |
api_key = os.environ.get("GEMINI_API_KEY")
|
15 |
if not api_key:
|
16 |
-
raise ValueError("GEMINI_API_KEY not found in environment variables. Please set it in
|
|
|
17 |
genai.configure(api_key=api_key)
|
18 |
-
model = genai.GenerativeModel('gemini-2.5')
|
19 |
|
20 |
-
#
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
21 |
PROMPT = """
|
22 |
You are an AI content reviewer. Analyze the provided text for the following:
|
23 |
-
1.
|
24 |
-
2.
|
25 |
-
3.
|
26 |
-
4.
|
27 |
|
28 |
Return the results in the following markdown format:
|
29 |
-
```markdown
|
30 |
# Blog Review Report
|
31 |
|
32 |
## Grammar Corrections
|
33 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
34 |
|
35 |
## Legal Policy Violations
|
36 |
-
- [
|
|
|
|
|
|
|
37 |
|
38 |
## Crude/Abusive Language
|
39 |
- [List instances of crude or abusive language or "None detected"]
|
40 |
|
41 |
## Sensitive Topics
|
42 |
- [List instances of sensitive topics or "None detected"]
|
43 |
-
```
|
44 |
|
45 |
-
For each issue, provide the
|
46 |
"""
|
47 |
|
48 |
-
def fetch_url_content(url):
|
49 |
try:
|
50 |
response = requests.get(url, timeout=10)
|
51 |
response.raise_for_status()
|
@@ -56,50 +81,132 @@ def fetch_url_content(url):
|
|
56 |
except Exception as e:
|
57 |
return f"Error fetching URL: {str(e)}"
|
58 |
|
59 |
-
def review_blog(
|
60 |
-
#
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
61 |
if input_type == "URL":
|
62 |
-
|
|
|
63 |
if input_text.startswith("Error"):
|
64 |
-
return input_text,
|
65 |
|
66 |
# Tokenize input for analysis
|
67 |
sentences = sent_tokenize(input_text)
|
68 |
analysis_text = "\n".join(sentences)
|
69 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
70 |
# Query Gemini with the prompt
|
71 |
try:
|
72 |
-
response = model.generate_content
|
73 |
report = response.text.strip()
|
74 |
# Ensure the response is markdown by removing any code fences
|
75 |
-
report = re.sub(r'
|
76 |
except Exception as e:
|
77 |
report = f"Error analyzing content with Gemini: {str(e)}"
|
|
|
|
|
|
|
|
|
|
|
78 |
|
79 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
80 |
|
81 |
-
# Gradio UI
|
82 |
-
with gr.Blocks(theme=gr.themes.Monochrome()) as demo:
|
83 |
gr.Markdown("# 📝 AI Blog Reviewer")
|
84 |
gr.Markdown("Enter blog text or a URL to review for grammar, legal issues, crude language, and sensitive topics. The report is generated in markdown format.")
|
85 |
|
86 |
-
|
87 |
-
|
88 |
-
|
|
|
|
|
|
|
|
|
89 |
|
90 |
gr.Markdown("### 📄 Review Report")
|
91 |
report_output = gr.Markdown()
|
92 |
download_btn = gr.File(label="Download Report", visible=False)
|
93 |
|
94 |
-
|
95 |
-
|
96 |
-
|
97 |
-
|
98 |
-
|
99 |
-
review_btn.click(
|
100 |
-
review_blog,
|
101 |
-
inputs=[blog_input, input_type],
|
102 |
-
outputs=[report_output, download_btn]
|
103 |
)
|
104 |
|
105 |
demo.launch()
|
|
|
5 |
import google.generativeai as genai
|
6 |
import os
|
7 |
import re
|
8 |
+
import tempfile
|
9 |
+
import asyncio
|
10 |
+
import time
|
11 |
|
12 |
# Download NLTK data
|
13 |
download('punkt')
|
14 |
download('punkt_tab')
|
15 |
|
16 |
+
# Configure Gemini API using environment variable
|
17 |
api_key = os.environ.get("GEMINI_API_KEY")
|
18 |
if not api_key:
|
19 |
+
raise ValueError("GEMINI_API_KEY not found in environment variables. Please set it in your environment.")
|
20 |
+
|
21 |
genai.configure(api_key=api_key)
|
|
|
22 |
|
23 |
+
# Use gemini-1.5-flash for faster and more accessible text analysis
|
24 |
+
try:
|
25 |
+
model = genai.GenerativeModel('gemini-1.5-flash')
|
26 |
+
except Exception as e:
|
27 |
+
# Fallback: List available models if the specified model is not found
|
28 |
+
print(f"Error initializing model: {str(e)}")
|
29 |
+
print("Available models:")
|
30 |
+
for m in genai.list_models():
|
31 |
+
print(m.name)
|
32 |
+
raise ValueError("Failed to initialize gemini-1.5-flash. Check available models above and update the model name.")
|
33 |
+
|
34 |
+
# Prompt for Gemini to analyze text with specified output format
|
35 |
PROMPT = """
|
36 |
You are an AI content reviewer. Analyze the provided text for the following:
|
37 |
+
1. Grammar Issues: Identify and suggest corrections for grammatical errors.
|
38 |
+
2. Legal Policy Violations: Flag content that may violate common legal policies (e.g., copyright infringement, defamation, incitement to violence).
|
39 |
+
3. Crude/Abusive Language: Detect crude, offensive, or abusive language.
|
40 |
+
4. Sensitive Topics: Identify content related to sensitive topics such as racism, gender bias, or other forms of discrimination.
|
41 |
|
42 |
Return the results in the following markdown format:
|
|
|
43 |
# Blog Review Report
|
44 |
|
45 |
## Grammar Corrections
|
46 |
+
1. [Heading of issue]
|
47 |
+
- CONTENT: [Exact line or part of text with the issue]
|
48 |
+
- SUGGESTION: [Suggested correction]
|
49 |
+
- ISSUE: [Description of the issue]
|
50 |
+
|
51 |
+
2. [Heading of next issue]
|
52 |
+
- CONTENT: [Exact line or part of text with the issue]
|
53 |
+
- SUGGESTION: [Suggested correction]
|
54 |
+
- ISSUE: [Description of the issue]
|
55 |
+
|
56 |
+
[Continue numbering for additional issues or state "None detected"]
|
57 |
|
58 |
## Legal Policy Violations
|
59 |
+
- CONTENT: [Exact line or part of text with the issue]
|
60 |
+
SUGGESTION: [Suggested action or correction]
|
61 |
+
ISSUE: [Description of the legal violation]
|
62 |
+
[Or state "None detected"]
|
63 |
|
64 |
## Crude/Abusive Language
|
65 |
- [List instances of crude or abusive language or "None detected"]
|
66 |
|
67 |
## Sensitive Topics
|
68 |
- [List instances of sensitive topics or "None detected"]
|
|
|
69 |
|
70 |
+
For each issue, provide the exact text, a suggested correction or action, and a concise explanation. Be precise and ensure the output strictly follows the specified format.
|
71 |
"""
|
72 |
|
73 |
+
async def fetch_url_content(url):
|
74 |
try:
|
75 |
response = requests.get(url, timeout=10)
|
76 |
response.raise_for_status()
|
|
|
81 |
except Exception as e:
|
82 |
return f"Error fetching URL: {str(e)}"
|
83 |
|
84 |
+
async def review_blog(text_input, url_input, progress=gr.Progress()):
|
85 |
+
# Determine input type based on which field is populated
|
86 |
+
if text_input and not url_input:
|
87 |
+
input_type = "Text"
|
88 |
+
input_text = text_input
|
89 |
+
elif url_input and not text_input:
|
90 |
+
input_type = "URL"
|
91 |
+
input_text = url_input
|
92 |
+
else:
|
93 |
+
return "Review Blog", "Error: Please provide input in either the Text or URL tab, but not both.", gr.update(visible=False)
|
94 |
+
|
95 |
+
# Handle empty input
|
96 |
+
if not input_text:
|
97 |
+
return "Review Blog", "Error: No input provided.", gr.update(visible=False)
|
98 |
+
|
99 |
+
# Handle URL input
|
100 |
if input_type == "URL":
|
101 |
+
progress(0, desc="Fetching content...")
|
102 |
+
input_text = await fetch_url_content(input_text)
|
103 |
if input_text.startswith("Error"):
|
104 |
+
return "Review Blog", input_text, gr.update(visible=False)
|
105 |
|
106 |
# Tokenize input for analysis
|
107 |
sentences = sent_tokenize(input_text)
|
108 |
analysis_text = "\n".join(sentences)
|
109 |
|
110 |
+
# Simulate progress for API call
|
111 |
+
progress(0, desc="Generating report...")
|
112 |
+
start_time = time.time()
|
113 |
+
for i in range(1, 10):
|
114 |
+
await asyncio.sleep(1) # Simulate progress every second
|
115 |
+
progress(i / 10, desc=f"Generating report... ({int(i * 10)}%)")
|
116 |
+
if time.time() - start_time > 30: # Timeout after 30 seconds
|
117 |
+
return "Review Blog", "Error: API request timed out after 30 seconds.", gr.update(visible=False)
|
118 |
+
|
119 |
# Query Gemini with the prompt
|
120 |
try:
|
121 |
+
response = await asyncio.to_thread(model.generate_content, PROMPT + "\n\nText to analyze:\n" + analysis_text)
|
122 |
report = response.text.strip()
|
123 |
# Ensure the response is markdown by removing any code fences
|
124 |
+
report = re.sub(r'^markdown\n|$', '', report, flags=re.MULTILINE)
|
125 |
except Exception as e:
|
126 |
report = f"Error analyzing content with Gemini: {str(e)}"
|
127 |
+
# Fallback: List available models for debugging
|
128 |
+
print("Available models:")
|
129 |
+
for m in genai.list_models():
|
130 |
+
print(m.name)
|
131 |
+
return "Review Blog", report, gr.update(visible=False)
|
132 |
|
133 |
+
# Create a temporary file to store the report
|
134 |
+
try:
|
135 |
+
with tempfile.NamedTemporaryFile(mode='w', suffix='.md', delete=False, encoding='utf-8') as temp_file:
|
136 |
+
temp_file.write(report)
|
137 |
+
temp_file_path = temp_file.name
|
138 |
+
progress(1.0, desc="Report generated!")
|
139 |
+
return "Review Blog", report, gr.update(visible=True, value=temp_file_path)
|
140 |
+
except Exception as e:
|
141 |
+
return "Review Blog", f"Error creating temporary file: {str(e)}", gr.update(visible=False)
|
142 |
+
|
143 |
+
# Custom CSS for hover effect, loading state, and Inter font
|
144 |
+
custom_css = """
|
145 |
+
@import url('https://fonts.googleapis.com/css2?family=Inter:wght@400;500;600;700&display=swap');
|
146 |
+
.gradio-container {
|
147 |
+
font-family: 'Inter', sans-serif !important;
|
148 |
+
}
|
149 |
+
.review-btn {
|
150 |
+
transition: all 0.3s ease;
|
151 |
+
font-weight: 500;
|
152 |
+
background-color: #2c3e50;
|
153 |
+
color: white;
|
154 |
+
border-radius: 8px;
|
155 |
+
padding: 10px 20px;
|
156 |
+
}
|
157 |
+
.review-btn:hover {
|
158 |
+
background-color: #4CAF50;
|
159 |
+
color: white;
|
160 |
+
transform: scale(1.05);
|
161 |
+
}
|
162 |
+
.review-btn:disabled {
|
163 |
+
opacity: 0.7;
|
164 |
+
cursor: not-allowed;
|
165 |
+
}
|
166 |
+
.review-btn:disabled::after {
|
167 |
+
content: ' ⏳';
|
168 |
+
}
|
169 |
+
.tab-nav button {
|
170 |
+
font-family: 'Inter', sans-serif;
|
171 |
+
font-weight: 500;
|
172 |
+
}
|
173 |
+
input, textarea {
|
174 |
+
font-family: 'Inter', sans-serif;
|
175 |
+
}
|
176 |
+
.gr-progress {
|
177 |
+
background-color: #e0e0e0;
|
178 |
+
border-radius: 8px;
|
179 |
+
overflow: hidden;
|
180 |
+
}
|
181 |
+
.gr-progress > div {
|
182 |
+
background-color: #4CAF50;
|
183 |
+
height: 20px;
|
184 |
+
transition: width 0.3s ease;
|
185 |
+
}
|
186 |
+
"""
|
187 |
|
188 |
+
# Gradio UI with Tabs
|
189 |
+
with gr.Blocks(theme=gr.themes.Monochrome(), css=custom_css) as demo:
|
190 |
gr.Markdown("# 📝 AI Blog Reviewer")
|
191 |
gr.Markdown("Enter blog text or a URL to review for grammar, legal issues, crude language, and sensitive topics. The report is generated in markdown format.")
|
192 |
|
193 |
+
with gr.Tabs():
|
194 |
+
with gr.TabItem("Text"):
|
195 |
+
text_input = gr.Textbox(lines=8, label="Blog Content", placeholder="Paste your blog text here...")
|
196 |
+
with gr.TabItem("URL"):
|
197 |
+
url_input = gr.Textbox(lines=1, label="Blog URL", placeholder="Enter the blog URL here...")
|
198 |
+
|
199 |
+
status_button = gr.Button(value="Review Blog", elem_classes=["review-btn"])
|
200 |
|
201 |
gr.Markdown("### 📄 Review Report")
|
202 |
report_output = gr.Markdown()
|
203 |
download_btn = gr.File(label="Download Report", visible=False)
|
204 |
|
205 |
+
# Bind the review button to process inputs
|
206 |
+
status_button.click(
|
207 |
+
fn=review_blog,
|
208 |
+
inputs=[text_input, url_input],
|
209 |
+
outputs=[status_button, report_output, download_btn]
|
|
|
|
|
|
|
|
|
210 |
)
|
211 |
|
212 |
demo.launch()
|