add pdf doc
Browse files
app.py
ADDED
@@ -0,0 +1,199 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import gradio as gr
|
2 |
+
import os
|
3 |
+
import tempfile
|
4 |
+
import dwani
|
5 |
+
import logging
|
6 |
+
|
7 |
+
# Set up logging
|
8 |
+
logging.basicConfig(level=logging.INFO)
|
9 |
+
logger = logging.getLogger(__name__)
|
10 |
+
|
11 |
+
# Configure dwani API settings
|
12 |
+
dwani.api_key = os.getenv("DWANI_API_KEY")
|
13 |
+
dwani.api_base = os.getenv("DWANI_API_BASE_URL")
|
14 |
+
|
15 |
+
# Log API configuration for debugging
|
16 |
+
logger.debug("DWANI_API_KEY: %s", "Set" if dwani.api_key else "Not set")
|
17 |
+
logger.debug("DWANI_API_BASE_URL: %s", dwani.api_base)
|
18 |
+
|
19 |
+
# Language options for dropdowns
|
20 |
+
language_options = [
|
21 |
+
("English", "eng_Latn"),
|
22 |
+
("Kannada", "kan_Knda"),
|
23 |
+
("Hindi", "hin_Deva")
|
24 |
+
]
|
25 |
+
|
26 |
+
# Create lists for Gradio dropdowns (display names only)
|
27 |
+
language_names = [lang[0] for lang in language_options]
|
28 |
+
|
29 |
+
# Map display names to language codes
|
30 |
+
lang_code_map = {lang[0]: lang[1] for lang in language_options}
|
31 |
+
|
32 |
+
|
33 |
+
def parse_page_numbers(pages_str):
|
34 |
+
"""
|
35 |
+
Parse a string of comma-separated page numbers/ranges into a sorted list of unique integers.
|
36 |
+
Example inputs:
|
37 |
+
"1,3,5"
|
38 |
+
"1-3,5"
|
39 |
+
"""
|
40 |
+
pages = set()
|
41 |
+
for part in pages_str.split(","):
|
42 |
+
part = part.strip()
|
43 |
+
if "-" in part:
|
44 |
+
try:
|
45 |
+
start, end = map(int, part.split("-"))
|
46 |
+
if start > end or start < 1:
|
47 |
+
continue
|
48 |
+
pages.update(range(start, end + 1))
|
49 |
+
except ValueError:
|
50 |
+
continue
|
51 |
+
else:
|
52 |
+
try:
|
53 |
+
page = int(part)
|
54 |
+
if page >= 1:
|
55 |
+
pages.add(page)
|
56 |
+
except ValueError:
|
57 |
+
continue
|
58 |
+
return sorted(pages)
|
59 |
+
|
60 |
+
|
61 |
+
def results_to_markdown(results):
|
62 |
+
"""
|
63 |
+
Convert the results dictionary into a Markdown formatted string.
|
64 |
+
"""
|
65 |
+
md_lines = []
|
66 |
+
for page, content in results.items():
|
67 |
+
md_lines.append(f"## {page}")
|
68 |
+
if "error" in content:
|
69 |
+
md_lines.append(f"**Error:** {content['error']}")
|
70 |
+
else:
|
71 |
+
md_lines.append(f"**Original Text:**\n\n``````")
|
72 |
+
md_lines.append(f"**Response:**\n\n{content.get('Response', '')}")
|
73 |
+
md_lines.append(f"**Processed Page:** {content.get('Processed Page', '')}")
|
74 |
+
md_lines.append(f"**Translated Response:**\n\n{content.get('Translated Response', '')}")
|
75 |
+
md_lines.append("\n---\n")
|
76 |
+
return "\n".join(md_lines)
|
77 |
+
|
78 |
+
|
79 |
+
def process_pdf(pdf_file, pages_str, prompt, src_lang, tgt_lang):
|
80 |
+
logger.debug("Received inputs - PDF: %s, Pages: %s, Prompt: %s, Source Lang: %s, Target Lang: %s",
|
81 |
+
pdf_file, pages_str, prompt, src_lang, tgt_lang)
|
82 |
+
|
83 |
+
# Validate inputs
|
84 |
+
if not pdf_file:
|
85 |
+
logger.error("No PDF file provided")
|
86 |
+
return "Error: Please upload a PDF file", None
|
87 |
+
|
88 |
+
if not prompt.strip():
|
89 |
+
logger.error("Prompt is empty")
|
90 |
+
return "Error: Please provide a non-empty prompt", None
|
91 |
+
|
92 |
+
pages = parse_page_numbers(pages_str)
|
93 |
+
if not pages:
|
94 |
+
logger.error("Invalid or empty page numbers input: %s", pages_str)
|
95 |
+
return "Error: Please provide valid page numbers (e.g., 1,3,5 or 1-3)", None
|
96 |
+
|
97 |
+
# Get language codes
|
98 |
+
src_lang_code = lang_code_map.get(src_lang)
|
99 |
+
tgt_lang_code = lang_code_map.get(tgt_lang)
|
100 |
+
|
101 |
+
if not src_lang_code or not tgt_lang_code:
|
102 |
+
logger.error("Invalid language selection - Source: %s, Target: %s", src_lang, tgt_lang)
|
103 |
+
return "Error: Invalid source or target language selection", None
|
104 |
+
|
105 |
+
# Get file path from Gradio File object
|
106 |
+
file_path = pdf_file.name if hasattr(pdf_file, 'name') else pdf_file
|
107 |
+
|
108 |
+
logger.debug("Calling API with file: %s, pages: %s, prompt: %s, src_lang: %s, tgt_lang: %s",
|
109 |
+
file_path, pages, prompt, src_lang_code, tgt_lang_code)
|
110 |
+
|
111 |
+
system_prompt = "Do not return any asterisk"
|
112 |
+
|
113 |
+
results = {}
|
114 |
+
for page_number in pages:
|
115 |
+
try:
|
116 |
+
result = dwani.Documents.run_doc_query(
|
117 |
+
file_path=file_path,
|
118 |
+
prompt=f"{prompt} {system_prompt}",
|
119 |
+
page_number=page_number,
|
120 |
+
src_lang=src_lang_code,
|
121 |
+
tgt_lang=tgt_lang_code
|
122 |
+
)
|
123 |
+
logger.debug("API response for page %d: %s", page_number, result)
|
124 |
+
results[f"Page {page_number}"] = {
|
125 |
+
"Original Text": result.get("original_text", "N/A"),
|
126 |
+
"Response": result.get("response", "N/A"),
|
127 |
+
"Processed Page": result.get("processed_page", "N/A"),
|
128 |
+
"Translated Response": result.get("translated_response", "N/A")
|
129 |
+
}
|
130 |
+
except dwani.exceptions.DhwaniAPIError as e:
|
131 |
+
logger.error("Dhwani API error on page %d: %s", page_number, str(e))
|
132 |
+
results[f"Page {page_number}"] = {"error": f"API error: {str(e)}"}
|
133 |
+
except Exception as e:
|
134 |
+
logger.error("Unexpected error on page %d: %s", page_number, str(e))
|
135 |
+
results[f"Page {page_number}"] = {"error": f"Unexpected error: {str(e)}"}
|
136 |
+
|
137 |
+
# Convert results to markdown text
|
138 |
+
markdown_text = results_to_markdown(results)
|
139 |
+
|
140 |
+
# Save markdown to a temporary file for download
|
141 |
+
temp_md_file = tempfile.NamedTemporaryFile(delete=False, suffix=".md", mode='w', encoding='utf-8')
|
142 |
+
temp_md_file.write(markdown_text)
|
143 |
+
temp_md_file.close()
|
144 |
+
|
145 |
+
# Return markdown text and file path for download
|
146 |
+
return markdown_text, temp_md_file.name
|
147 |
+
|
148 |
+
|
149 |
+
# Define Gradio interface
|
150 |
+
with gr.Blocks(title="PDF Custom Prompt Processor with Multi-Page Support") as demo:
|
151 |
+
gr.Markdown("# PDF Custom Prompt Processor")
|
152 |
+
gr.Markdown("Upload a PDF, specify page numbers (comma-separated or ranges), enter a prompt, and select source and target languages.")
|
153 |
+
|
154 |
+
with gr.Row():
|
155 |
+
with gr.Column():
|
156 |
+
pdf_input = gr.File(label="Upload PDF", file_types=[".pdf"])
|
157 |
+
pages_input = gr.Textbox(
|
158 |
+
label="Page Numbers",
|
159 |
+
placeholder="e.g., 1,3,5 or 1-3",
|
160 |
+
value="1",
|
161 |
+
lines=1
|
162 |
+
)
|
163 |
+
prompt = gr.Textbox(
|
164 |
+
label="Custom Prompt",
|
165 |
+
placeholder="e.g., List the key points",
|
166 |
+
value="List the key points",
|
167 |
+
lines=3
|
168 |
+
)
|
169 |
+
src_lang_input = gr.Dropdown(
|
170 |
+
label="Source Language",
|
171 |
+
choices=language_names,
|
172 |
+
value="English"
|
173 |
+
)
|
174 |
+
tgt_lang_input = gr.Dropdown(
|
175 |
+
label="Target Language",
|
176 |
+
choices=language_names,
|
177 |
+
value="Kannada"
|
178 |
+
)
|
179 |
+
submit_btn = gr.Button("Process")
|
180 |
+
|
181 |
+
with gr.Column():
|
182 |
+
output_md = gr.Markdown(label="Response (Markdown)")
|
183 |
+
download_md = gr.File(label="Download Markdown File")
|
184 |
+
|
185 |
+
submit_btn.click(
|
186 |
+
fn=process_pdf,
|
187 |
+
inputs=[pdf_input, pages_input, prompt, src_lang_input, tgt_lang_input],
|
188 |
+
outputs=[output_md, download_md]
|
189 |
+
)
|
190 |
+
|
191 |
+
|
192 |
+
# Launch the interface
|
193 |
+
if __name__ == "__main__":
|
194 |
+
if not dwani.api_key or not dwani.api_base:
|
195 |
+
logger.error("API key or base URL not set. Please set DWANI_API_KEY and DWANI_API_BASE_URL environment variables.")
|
196 |
+
print("Error: Please set DWANI_API_KEY and DWANI_API_BASE_URL environment variables.")
|
197 |
+
else:
|
198 |
+
logger.debug("Starting Gradio interface...")
|
199 |
+
demo.launch()
|