Spaces:
Runtime error
Runtime error
Update app.py
Browse files
app.py
CHANGED
|
@@ -33,6 +33,103 @@ TOTAL_SUMMARIZED_WORDS = 0
|
|
| 33 |
# Helper functions for external APIs and PDF Processing
|
| 34 |
# =============================================================================
|
| 35 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 36 |
def generate_graph_snippet(placeholder_text: str, context: str, initial_query: str, crumbs: str) -> str:
|
| 37 |
|
| 38 |
graph_examples = """
|
|
@@ -2327,7 +2424,8 @@ def main():
|
|
| 2327 |
report_status = gr.Textbox(label="Report Status", interactive=False, lines=2, value="Click 'Generate Report' to create your PDF report.")
|
| 2328 |
report_file = gr.File(label="Download Report", visible=False, interactive=False, file_types=[".pdf"])
|
| 2329 |
generate_button = gr.Button("Generate Report")
|
| 2330 |
-
|
|
|
|
| 2331 |
with gr.Accordion("6] Extra Context (Crumbs, Existing Report & Log, Processed Queries)", open=False):
|
| 2332 |
existing_report = gr.Textbox(label="Existing Report (if any)", placeholder="Paste previously generated report here...", lines=4)
|
| 2333 |
existing_log = gr.Textbox(label="Existing Process Log (if any)", placeholder="Paste previously generated log here...", lines=4)
|
|
@@ -2387,6 +2485,13 @@ def main():
|
|
| 2387 |
outputs=[report_status, report_file]
|
| 2388 |
)
|
| 2389 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 2390 |
demo.launch()
|
| 2391 |
|
| 2392 |
if __name__ == "__main__":
|
|
|
|
| 33 |
# Helper functions for external APIs and PDF Processing
|
| 34 |
# =============================================================================
|
| 35 |
|
| 36 |
+
def fine_tune_report(report_html: str, initial_request: str, qa: str, target_style: str,
|
| 37 |
+
knowledge_crumbs: str, complementary_guidance: str) -> (str, str):
|
| 38 |
+
"""
|
| 39 |
+
Fine-tunes the report by identifying improvable chunks, processing each one,
|
| 40 |
+
and updating the report HTML with improved content. Also returns an updated
|
| 41 |
+
plain text summary of changes that can be appended to the Q&A textbox.
|
| 42 |
+
|
| 43 |
+
Inputs:
|
| 44 |
+
- report_html: The full HTML code of the current report (non-empty)
|
| 45 |
+
- initial_request: The original research query or instructions
|
| 46 |
+
- qa: The existing Q&A text (plain text)
|
| 47 |
+
- target_style: The desired style (e.g., "Academic style") to apply uniformly
|
| 48 |
+
- knowledge_crumbs: The aggregated knowledge nuggets extracted from the searches
|
| 49 |
+
- complementary_guidance: Any additional guidance from the search parameters
|
| 50 |
+
|
| 51 |
+
Returns:
|
| 52 |
+
- A tuple of two elements:
|
| 53 |
+
(final_report_html, updated_qa)
|
| 54 |
+
Where:
|
| 55 |
+
* final_report_html is the updated HTML report (with reinjected improved chunks)
|
| 56 |
+
* updated_qa is the original Q&A text with a plain text summary of improvements appended.
|
| 57 |
+
"""
|
| 58 |
+
from bs4 import BeautifulSoup
|
| 59 |
+
import json
|
| 60 |
+
import logging
|
| 61 |
+
|
| 62 |
+
# Parse the existing report HTML
|
| 63 |
+
soup = BeautifulSoup(report_html, "html.parser")
|
| 64 |
+
|
| 65 |
+
# Try to find pre-marked improvable chunks.
|
| 66 |
+
# (For example, your report generator may wrap sections in <div class="improvable-chunk">...</div>)
|
| 67 |
+
chunks = soup.find_all("div", class_="improvable-chunk")
|
| 68 |
+
|
| 69 |
+
# If no marked chunks exist, as a fallback we group paragraphs into chunks.
|
| 70 |
+
if not chunks:
|
| 71 |
+
all_paragraphs = soup.find_all("p")
|
| 72 |
+
chunks = []
|
| 73 |
+
# Group every n paragraphs into a chunk (ensure at least 5 chunks)
|
| 74 |
+
group_size = max(1, len(all_paragraphs) // 10)
|
| 75 |
+
# Create new chunks and append them at the end of the body
|
| 76 |
+
for i in range(0, len(all_paragraphs), group_size):
|
| 77 |
+
new_div = soup.new_tag("div", **{"class": "improvable-chunk"})
|
| 78 |
+
for p in all_paragraphs[i:i+group_size]:
|
| 79 |
+
new_div.append(p.extract())
|
| 80 |
+
# Append the newly created chunk back to <body>
|
| 81 |
+
soup.body.append(new_div)
|
| 82 |
+
chunks.append(new_div)
|
| 83 |
+
|
| 84 |
+
improvements_summary = [] # to store plain text summary for each chunk
|
| 85 |
+
|
| 86 |
+
# Process each identified chunk sequentially.
|
| 87 |
+
for idx, chunk in enumerate(chunks, start=1):
|
| 88 |
+
original_chunk = str(chunk)
|
| 89 |
+
# Build a detailed prompt including all relevant inputs.
|
| 90 |
+
prompt = (
|
| 91 |
+
f"Improve the following report chunk to enhance clarity, incorporate additional knowledge, "
|
| 92 |
+
f"and ensure any citations are consistent. Replace any placeholders or visuals so that the text "
|
| 93 |
+
f"becomes more cohesive and well written, matching the target style.\n\n"
|
| 94 |
+
f"--- Chunk #{idx} Original Content ---\n{original_chunk}\n\n"
|
| 95 |
+
f"Initial Request: {initial_request}\n\n"
|
| 96 |
+
f"Clarification Q&A: {qa}\n\n"
|
| 97 |
+
f"Target Style: {target_style}\n\n"
|
| 98 |
+
f"Knowledge Crumbs: {knowledge_crumbs}\n\n"
|
| 99 |
+
f"Complementary Guidance: {complementary_guidance}\n\n"
|
| 100 |
+
f"Please output a JSON object with exactly two fields (no extra commentary):\n"
|
| 101 |
+
f'{{"improved": "<the improved chunk in valid HTML>", "summary": "<a brief summary of changes>"}}\n'
|
| 102 |
+
f"Do not include any markdown formatting or backticks."
|
| 103 |
+
)
|
| 104 |
+
|
| 105 |
+
# Call the model (using openai_call with a high token limit—for example, 5000 tokens)
|
| 106 |
+
result = openai_call(prompt, model="o3-mini", max_tokens_param=5000, temperature=0.5)
|
| 107 |
+
result = result.strip().strip("```")
|
| 108 |
+
try:
|
| 109 |
+
res_json = json.loads(result)
|
| 110 |
+
improved_chunk = res_json.get("improved")
|
| 111 |
+
chunk_summary = res_json.get("summary")
|
| 112 |
+
if improved_chunk and chunk_summary:
|
| 113 |
+
improvements_summary.append(f"Chunk {idx}: {chunk_summary}")
|
| 114 |
+
# Replace the old chunk with the improved HTML.
|
| 115 |
+
new_chunk = BeautifulSoup(improved_chunk, "html.parser")
|
| 116 |
+
chunk.replace_with(new_chunk)
|
| 117 |
+
else:
|
| 118 |
+
logging.error(f"Chunk {idx}: Incomplete JSON result: {result}")
|
| 119 |
+
except Exception as e:
|
| 120 |
+
logging.error(f"Error processing chunk {idx}: {e}. Raw result: {result}")
|
| 121 |
+
|
| 122 |
+
# Get the updated report HTML as a string.
|
| 123 |
+
final_report_html = str(soup)
|
| 124 |
+
|
| 125 |
+
# Create a plain text summary of improvements.
|
| 126 |
+
summary_text = "Summary of Fine-Tuning Improvements:\n" + "\n".join(improvements_summary)
|
| 127 |
+
|
| 128 |
+
# Append the summary (preceded by a separator) to the original QA text.
|
| 129 |
+
updated_qa = qa.strip() + "\n----------\n" + summary_text
|
| 130 |
+
|
| 131 |
+
return final_report_html, updated_qa
|
| 132 |
+
|
| 133 |
def generate_graph_snippet(placeholder_text: str, context: str, initial_query: str, crumbs: str) -> str:
|
| 134 |
|
| 135 |
graph_examples = """
|
|
|
|
| 2424 |
report_status = gr.Textbox(label="Report Status", interactive=False, lines=2, value="Click 'Generate Report' to create your PDF report.")
|
| 2425 |
report_file = gr.File(label="Download Report", visible=False, interactive=False, file_types=[".pdf"])
|
| 2426 |
generate_button = gr.Button("Generate Report")
|
| 2427 |
+
fine_tune_button = gr.Button("AI Improve the Report")
|
| 2428 |
+
|
| 2429 |
with gr.Accordion("6] Extra Context (Crumbs, Existing Report & Log, Processed Queries)", open=False):
|
| 2430 |
existing_report = gr.Textbox(label="Existing Report (if any)", placeholder="Paste previously generated report here...", lines=4)
|
| 2431 |
existing_log = gr.Textbox(label="Existing Process Log (if any)", placeholder="Paste previously generated log here...", lines=4)
|
|
|
|
| 2485 |
outputs=[report_status, report_file]
|
| 2486 |
)
|
| 2487 |
|
| 2488 |
+
fine_tune_button.click(
|
| 2489 |
+
fn=fine_tune_report,
|
| 2490 |
+
inputs=[final_report, research_query, clarification_text, reportstyle, crumbs_box, additional_clarifications],
|
| 2491 |
+
outputs=[fine_tuned_report, clarification_text]
|
| 2492 |
+
)
|
| 2493 |
+
|
| 2494 |
+
|
| 2495 |
demo.launch()
|
| 2496 |
|
| 2497 |
if __name__ == "__main__":
|