Update app.py
Browse files
app.py
CHANGED
|
@@ -1,114 +1,4 @@
|
|
| 1 |
-
|
| 2 |
-
if not source_url or not target_url or not anchor_text:
|
| 3 |
-
return "β Please provide Source URL, Target URL, and Anchor Text."
|
| 4 |
-
|
| 5 |
-
# Auto-correct swapped inputs
|
| 6 |
-
warn = ""
|
| 7 |
-
if looks_like_url(anchor_text) and not looks_like_url(target_url):
|
| 8 |
-
anchor_text, target_url = target_url, anchor_text
|
| 9 |
-
warn = "βΉοΈ Detected swapped inputs. I used the URL as Target URL and the text as Anchor.\n\n"
|
| 10 |
-
|
| 11 |
-
source_url = normalize_url(source_url)
|
| 12 |
-
target_url = normalize_url(target_url)
|
| 13 |
-
|
| 14 |
-
try:
|
| 15 |
-
# Get source blocks for later use
|
| 16 |
-
source_blocks = get_text_blocks(source_url)
|
| 17 |
-
|
| 18 |
-
# Only get 1 result, not multiple!
|
| 19 |
-
results = suggest_insertions(source_url, target_url, anchor_text, top_k=1, suggest_alternative=suggest_alternative_anchor)
|
| 20 |
-
res = results[0]
|
| 21 |
-
except Exception as e:
|
| 22 |
-
return f"β Error processing the page: {str(e)}"
|
| 23 |
-
|
| 24 |
-
if "error" in res:
|
| 25 |
-
return f"β {res['error']}"
|
| 26 |
-
|
| 27 |
-
# Detect language from the original sentence
|
| 28 |
-
original_sentence = res['best_sentence_original']
|
| 29 |
-
detected_lang = detect_language(original_sentence)
|
| 30 |
-
language_name = get_language_name(detected_lang)
|
| 31 |
-
print(f"Detected language: {language_name} ({detected_lang})")
|
| 32 |
-
|
| 33 |
-
# Process original anchor suggestion
|
| 34 |
-
draft_html = res["best_sentence_with_anchor"]
|
| 35 |
-
|
| 36 |
-
# Check if anchor was already present in the article
|
| 37 |
-
anchor_was_present = res.get("anchor_was_present", False)
|
| 38 |
-
keyword_in_article = res.get("keyword_in_article", False)
|
| 39 |
-
|
| 40 |
-
# If anchor is present in the article (even if not in the best sentence)
|
| 41 |
-
if keyword_in_article:
|
| 42 |
-
# Anchor exists somewhere in article
|
| 43 |
-
if anchor_was_present:
|
| 44 |
-
# Anchor is in the suggested sentence - just show where to add the link
|
| 45 |
-
final_output = to_plain_text(draft_html) if plain_text else draft_html
|
| 46 |
-
result = warn + f"β
**Anchor text '{anchor_text}' found in article!**\n\n"
|
| 47 |
-
result += f"π Add link here:\n\n"
|
| 48 |
-
result += f"{final_output}"
|
| 49 |
-
else:
|
| 50 |
-
# Anchor is in article but not in this sentence
|
| 51 |
-
if smart_rewrite:
|
| 52 |
-
g = gpt_rewrite(draft_html, anchor_text, target_url, style="neutral", language=language_name)
|
| 53 |
-
final_html = g["sentence_html"]
|
| 54 |
-
else:
|
| 55 |
-
final_html = draft_html
|
| 56 |
-
|
| 57 |
-
polished = gpt_validate_and_polish(final_html, anchor_text, target_url, language=language_name)
|
| 58 |
-
final_html = polished.get("sentence_html", final_html)
|
| 59 |
-
final_output = to_plain_text(final_html) if plain_text else final_html
|
| 60 |
-
|
| 61 |
-
result = warn + f"β
**Anchor text '{anchor_text}' found in article!**\n\n"
|
| 62 |
-
result += f"π Add link here:\n\n"
|
| 63 |
-
result += f"{final_output}"
|
| 64 |
-
else:
|
| 65 |
-
# Anchor doesn't exist in article at all - need to add it
|
| 66 |
-
if smart_rewrite:
|
| 67 |
-
g = gpt_rewrite(draft_html, anchor_text, target_url, style="neutral", language=language_name)
|
| 68 |
-
final_html = g["sentence_html"]
|
| 69 |
-
else:
|
| 70 |
-
final_html = draft_html
|
| 71 |
-
|
| 72 |
-
polished = gpt_validate_and_polish(final_html, anchor_text, target_url, language=language_name)
|
| 73 |
-
final_html = polished.get("sentence_html", final_html)
|
| 74 |
-
final_output = to_plain_text(final_html) if plain_text else final_html
|
| 75 |
-
|
| 76 |
-
result = warn + f"β οΈ **Anchor text '{anchor_text}' not found in article**\n\n"
|
| 77 |
-
result += f"π Result 1 - Suggested placement:\n\n"
|
| 78 |
-
result += f"Change this sentence: {original_sentence}\n\n"
|
| 79 |
-
result += f"With this one: {final_output}"
|
| 80 |
-
|
| 81 |
-
# Show alternative if requested and available
|
| 82 |
-
if suggest_alternative_anchor and res.get("alternative_anchor"):
|
| 83 |
-
alt_anchor = res["alternative_anchor"]
|
| 84 |
-
alt_content = res.get("alternative_sentence", "") # This now contains position info + content
|
| 85 |
-
|
| 86 |
-
if alt_content:
|
| 87 |
-
# The content already has the link included from GPT
|
| 88 |
-
# Parse the position text and actual content
|
| 89 |
-
if "[Insert after paragraph" in alt_content:
|
| 90 |
-
parts = alt_content.split("\n\n", 1)
|
| 91 |
-
position_and_sentence = parts[0] if len(parts) > 0 else ""
|
| 92 |
-
actual_content = parts[1] if len(parts) > 1 else alt_content
|
| 93 |
-
else:
|
| 94 |
-
position_and_sentence = ""
|
| 95 |
-
actual_content = alt_content
|
| 96 |
-
|
| 97 |
-
alt_output = to_plain_text(actual_content) if plain_text else actual_content
|
| 98 |
-
|
| 99 |
-
# Add alternative as Result 2
|
| 100 |
-
result += f"\n\n{'='*50}\n\n"
|
| 101 |
-
result += f"π Result 2 - Suggested new anchor with placement:\n"
|
| 102 |
-
result += f"π‘ Using keyword: '{alt_anchor}'\n"
|
| 103 |
-
|
| 104 |
-
if position_and_sentence:
|
| 105 |
-
result += f"\n{position_and_sentence}\n\n"
|
| 106 |
-
|
| 107 |
-
result += f"Add this: {alt_output}"
|
| 108 |
-
|
| 109 |
-
return result:
|
| 110 |
-
# No paragraph info available
|
| 111 |
-
result += f"\n{alt_output}"import os, re, json, requests, urllib.parse
|
| 112 |
import torch, torch.nn.functional as F
|
| 113 |
from bs4 import BeautifulSoup
|
| 114 |
from transformers import AutoTokenizer, AutoModel
|
|
@@ -417,11 +307,11 @@ def find_alternative_anchor(blocks, target_url, original_anchor):
|
|
| 417 |
chosen_keyword = result.get("chosen_keyword", keywords[0] if keywords else original_anchor)
|
| 418 |
new_content = result.get("new_content", "")
|
| 419 |
insert_after = result.get("insert_after_paragraph", 0)
|
| 420 |
-
|
| 421 |
|
| 422 |
# Format the response for compatibility
|
| 423 |
# Return: (anchor_text, formatted_content_with_position)
|
| 424 |
-
position_text = f"[Insert after paragraph {insert_after + 1}]
|
| 425 |
|
| 426 |
return chosen_keyword, f"{position_text}\n\n{new_content}"
|
| 427 |
|
|
@@ -645,7 +535,7 @@ def gpt_rewrite(sentence_html, anchor_text, target_url, style="neutral", languag
|
|
| 645 |
"(2) Do NOT use an em dash or any dash. "
|
| 646 |
'(3) Avoid phrases like "for details", "click here", "learn more", "visit", "read more". '
|
| 647 |
"Prefer integrating the anchor as part of the sentence. "
|
| 648 |
-
f"(4) Write in {language} and preserve ALL special characters (Δ,
|
| 649 |
"Return a compact JSON object with key sentence_html only."
|
| 650 |
)
|
| 651 |
|
|
@@ -770,7 +660,21 @@ def gpt_get_search_keywords(target_content, target_url):
|
|
| 770 |
|
| 771 |
return obj.get("keywords", ["related content"])
|
| 772 |
|
| 773 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 774 |
"The addition should flow seamlessly with the existing content. "
|
| 775 |
"\n\nYOUR TASK:\n"
|
| 776 |
"1. Choose the ONE keyword that fits most naturally with the article's context\n"
|
|
@@ -891,8 +795,8 @@ def run_tool(source_url, target_url, anchor_text, smart_rewrite, plain_text, sug
|
|
| 891 |
|
| 892 |
result = warn + f"β οΈ **Anchor text '{anchor_text}' not found in article**\n\n"
|
| 893 |
result += f"π Result 1 - Suggested placement:\n\n"
|
| 894 |
-
result += f"
|
| 895 |
-
result += f"
|
| 896 |
|
| 897 |
# Show alternative if requested and available
|
| 898 |
if suggest_alternative_anchor and res.get("alternative_anchor"):
|
|
@@ -901,47 +805,24 @@ def run_tool(source_url, target_url, anchor_text, smart_rewrite, plain_text, sug
|
|
| 901 |
|
| 902 |
if alt_content:
|
| 903 |
# Parse if there's position information
|
| 904 |
-
paragraph_num = None
|
| 905 |
-
actual_content = alt_content
|
| 906 |
-
|
| 907 |
if "[Insert after paragraph" in alt_content:
|
| 908 |
parts = alt_content.split("\n\n", 1)
|
| 909 |
position_info = parts[0] if len(parts) > 0 else ""
|
| 910 |
actual_content = parts[1] if len(parts) > 1 else alt_content
|
| 911 |
-
|
| 912 |
-
|
| 913 |
-
|
| 914 |
-
match = re.search(r'\[Insert after paragraph (\d+)\]', position_info)
|
| 915 |
-
if match:
|
| 916 |
-
paragraph_num = int(match.group(1)) - 1 # Convert to 0-based index
|
| 917 |
|
| 918 |
# The content already has the link included from GPT
|
| 919 |
alt_output = to_plain_text(actual_content) if plain_text else actual_content
|
| 920 |
|
| 921 |
# Add alternative as Result 2
|
| 922 |
result += f"\n\n{'='*50}\n\n"
|
| 923 |
-
result += f"π Result 2 - Suggested new
|
| 924 |
result += f"π‘ Using keyword: '{alt_anchor}'\n"
|
| 925 |
-
|
| 926 |
-
|
| 927 |
-
|
| 928 |
-
try:
|
| 929 |
-
# Get the last sentence of the target paragraph
|
| 930 |
-
target_paragraph = source_blocks[paragraph_num]
|
| 931 |
-
# Split into sentences and get the last one
|
| 932 |
-
sentences = re.split(r'(?<=[.!?])\s+', target_paragraph.strip())
|
| 933 |
-
last_sentence = sentences[-1] if sentences else target_paragraph[:100] + "..."
|
| 934 |
-
|
| 935 |
-
result += f"\n[Insert after paragraph {paragraph_num + 1}]\n"
|
| 936 |
-
result += f"After this sentence: {last_sentence}\n\n"
|
| 937 |
-
result += f"Add this: {alt_output}"
|
| 938 |
-
except:
|
| 939 |
-
# Fallback if we can't get the paragraph
|
| 940 |
-
result += f"\n[Insert after paragraph {paragraph_num + 1}]\n\n"
|
| 941 |
-
result += f"{alt_output}"
|
| 942 |
-
else:
|
| 943 |
-
# No paragraph info available
|
| 944 |
-
result += f"\n{alt_output}"
|
| 945 |
|
| 946 |
return result
|
| 947 |
|
|
@@ -998,7 +879,7 @@ with gr.Blocks(title=f"Link Insertion Helper β’ GPT: {gpt_status}") as demo:
|
|
| 998 |
|
| 999 |
gr.Markdown("""
|
| 1000 |
### Features:
|
| 1001 |
-
- π **Auto Language Detection**: Preserves special characters (Δ,
|
| 1002 |
- πΎ **Smart Caching**: Caches embeddings and API responses for faster repeated queries
|
| 1003 |
- π― **Main Content Focus**: Prioritizes first 5-7 paragraphs, ignores author bios
|
| 1004 |
- π **Alternative Anchor Suggestion**: When your anchor isn't in the article, suggests better anchors from existing text
|
|
|
|
| 1 |
+
import os, re, json, requests, urllib.parse
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 2 |
import torch, torch.nn.functional as F
|
| 3 |
from bs4 import BeautifulSoup
|
| 4 |
from transformers import AutoTokenizer, AutoModel
|
|
|
|
| 307 |
chosen_keyword = result.get("chosen_keyword", keywords[0] if keywords else original_anchor)
|
| 308 |
new_content = result.get("new_content", "")
|
| 309 |
insert_after = result.get("insert_after_paragraph", 0)
|
| 310 |
+
reasoning = result.get("reasoning", "")
|
| 311 |
|
| 312 |
# Format the response for compatibility
|
| 313 |
# Return: (anchor_text, formatted_content_with_position)
|
| 314 |
+
position_text = f"[Insert after paragraph {insert_after + 1}]: {reasoning}"
|
| 315 |
|
| 316 |
return chosen_keyword, f"{position_text}\n\n{new_content}"
|
| 317 |
|
|
|
|
| 535 |
"(2) Do NOT use an em dash or any dash. "
|
| 536 |
'(3) Avoid phrases like "for details", "click here", "learn more", "visit", "read more". '
|
| 537 |
"Prefer integrating the anchor as part of the sentence. "
|
| 538 |
+
f"(4) Write in {language} and preserve ALL special characters (Δ, Δ, Ε‘, ΕΎ, Δ, etc.). "
|
| 539 |
"Return a compact JSON object with key sentence_html only."
|
| 540 |
)
|
| 541 |
|
|
|
|
| 660 |
|
| 661 |
return obj.get("keywords", ["related content"])
|
| 662 |
|
| 663 |
+
def gpt_generate_content_with_keyword(source_blocks, keywords, target_url, language="English"):
|
| 664 |
+
"""
|
| 665 |
+
Generate new content with the best keyword and specify where to insert it.
|
| 666 |
+
"""
|
| 667 |
+
if not OPENAI_API_KEY or not keywords:
|
| 668 |
+
return None
|
| 669 |
+
|
| 670 |
+
# Create cache key
|
| 671 |
+
source_preview = " ".join(source_blocks[:3])[:500]
|
| 672 |
+
cache_key = hashlib.md5(f"generate_{source_preview}_{str(keywords)}_{target_url}_{language}".encode()).hexdigest()
|
| 673 |
+
|
| 674 |
+
system = (
|
| 675 |
+
f"You are a skilled content writer writing in {language}. "
|
| 676 |
+
"Given an article and a list of keywords related to a target page, "
|
| 677 |
+
"create a NATURAL addition to the article that incorporates the most suitable keyword. "
|
| 678 |
"The addition should flow seamlessly with the existing content. "
|
| 679 |
"\n\nYOUR TASK:\n"
|
| 680 |
"1. Choose the ONE keyword that fits most naturally with the article's context\n"
|
|
|
|
| 795 |
|
| 796 |
result = warn + f"β οΈ **Anchor text '{anchor_text}' not found in article**\n\n"
|
| 797 |
result += f"π Result 1 - Suggested placement:\n\n"
|
| 798 |
+
result += f"Original: {original_sentence}\n\n"
|
| 799 |
+
result += f"Suggested: {final_output}"
|
| 800 |
|
| 801 |
# Show alternative if requested and available
|
| 802 |
if suggest_alternative_anchor and res.get("alternative_anchor"):
|
|
|
|
| 805 |
|
| 806 |
if alt_content:
|
| 807 |
# Parse if there's position information
|
|
|
|
|
|
|
|
|
|
| 808 |
if "[Insert after paragraph" in alt_content:
|
| 809 |
parts = alt_content.split("\n\n", 1)
|
| 810 |
position_info = parts[0] if len(parts) > 0 else ""
|
| 811 |
actual_content = parts[1] if len(parts) > 1 else alt_content
|
| 812 |
+
else:
|
| 813 |
+
position_info = ""
|
| 814 |
+
actual_content = alt_content
|
|
|
|
|
|
|
|
|
|
| 815 |
|
| 816 |
# The content already has the link included from GPT
|
| 817 |
alt_output = to_plain_text(actual_content) if plain_text else actual_content
|
| 818 |
|
| 819 |
# Add alternative as Result 2
|
| 820 |
result += f"\n\n{'='*50}\n\n"
|
| 821 |
+
result += f"π Result 2 - Suggested new content to add:\n"
|
| 822 |
result += f"π‘ Using keyword: '{alt_anchor}'\n"
|
| 823 |
+
if position_info:
|
| 824 |
+
result += f"π {position_info}\n"
|
| 825 |
+
result += f"\n{alt_output}"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 826 |
|
| 827 |
return result
|
| 828 |
|
|
|
|
| 879 |
|
| 880 |
gr.Markdown("""
|
| 881 |
### Features:
|
| 882 |
+
- π **Auto Language Detection**: Preserves special characters (Δ, Δ, Ε‘, ΕΎ, Δ, etc.)
|
| 883 |
- πΎ **Smart Caching**: Caches embeddings and API responses for faster repeated queries
|
| 884 |
- π― **Main Content Focus**: Prioritizes first 5-7 paragraphs, ignores author bios
|
| 885 |
- π **Alternative Anchor Suggestion**: When your anchor isn't in the article, suggests better anchors from existing text
|