{answer}
-{va.perspective_gaps}
+{va.integrative_insights}
+{state.main_query}
" - else: - html += "No main query specified
" - - html += """ -Found {len(results)} results
-{info.synthesized_answer}
- """ - - if info.key_sources: - html += """ -Key Sources:
-{state.viewpoint_analysis.perspective_gaps}
- """ - - if state.viewpoint_analysis.integrative_insights: - html += f""" -{state.viewpoint_analysis.integrative_insights}
- """ - - html += """ -Improvements Made:
-Final HTML report is available but not displayed here. View the HTML artifact to see the complete report.
-No additional queries recommended
- """ - - html += """ -No critique summary available
- """ - - html += """ -No additional questions identified
- """ - - html += """ -Main Query: {}
-Current Stage: {}
-Sub-questions: {}
-Search Results: {} queries with results
-Synthesized Info: {} topics synthesized
-| Provider | +Cost | +Percentage | +
|---|---|---|
| {provider} | +${cost:.4f} | +{(cost / total_cost * 100 if total_cost > 0 else 0):.1f}% | +
{info.information_gaps}
+{enhanced_info.synthesized_answer}
+{info.synthesized_answer}
+ +{code_content}{html.escape(code_content)}\1", html.escape(line))
+ # Process bullet points
+ if line.strip().startswith("•") or line.strip().startswith("-"):
+ line = re.sub(r"^(\s*)[•-]\s*", r"\1", line)
+ formatted_lines.append(f"{line}
") + + # Handle case where code block wasn't closed + if in_code_block and code_lines: + code_content = "\n".join(code_lines) + formatted_lines.append( + f'{html.escape(code_content)} blocks)
- parts = []
- in_pre = False
- for line in formatted_text.split("\n"):
- if "" in line:
- in_pre = True
- parts.append(line)
- elif "
" in line:
- in_pre = False
- parts.append(line)
- elif in_pre:
- # Inside a code block, preserve newlines
- parts.append(line)
+ # Wrap list items in ul tags
+ result = []
+ in_list = False
+ for line in formatted_lines:
+ if line.startswith("This report examines the question: {html.escape(state.main_query)}
" - summary += f"The research explored {len(state.sub_questions)} key dimensions of this topic, " + summary = f"
This report examines the question: {html.escape(query_context.main_query)}
" + summary += f"The research explored {len(query_context.sub_questions)} key dimensions of this topic, " summary += "synthesizing findings from multiple sources to provide a comprehensive analysis.
" # Add confidence overview confidence_counts = {"high": 0, "medium": 0, "low": 0} - for info in state.enhanced_info.values(): + info_source = ( + synthesis_data.enhanced_info + if synthesis_data.enhanced_info + else synthesis_data.synthesized_info + ) + for info in info_source.values(): level = info.confidence_level.lower() if level in confidence_counts: confidence_counts[level] += 1 @@ -292,10 +336,10 @@ def generate_fallback_executive_summary(state: ResearchState) -> str: return summary -def generate_fallback_introduction(state: ResearchState) -> str: +def generate_fallback_introduction(query_context: QueryContext) -> str: """Generate a fallback introduction when LLM fails.""" - intro = f"This report addresses the research query: {html.escape(state.main_query)}
" - intro += f"The research was conducted by breaking down the main query into {len(state.sub_questions)} " + intro = f"
This report addresses the research query: {html.escape(query_context.main_query)}
" + intro += f"The research was conducted by breaking down the main query into {len(query_context.sub_questions)} "
intro += (
"sub-questions to explore different aspects of the topic in depth. "
)
@@ -304,7 +348,9 @@ def generate_fallback_introduction(state: ResearchState) -> str:
def generate_conclusion(
- state: ResearchState,
+ query_context: QueryContext,
+ synthesis_data: SynthesisData,
+ analysis_data: AnalysisData,
conclusion_generation_prompt: Prompt,
llm_model: str = "sambanova/DeepSeek-R1-Distill-Llama-70B",
langfuse_project_name: str = "deep-research",
@@ -312,9 +358,12 @@ def generate_conclusion(
"""Generate a comprehensive conclusion using LLM based on all research findings.
Args:
- state: The ResearchState containing all research findings
+ query_context: The query context with main query and sub-questions
+ synthesis_data: The synthesis data with all synthesized information
+ analysis_data: The analysis data with viewpoint analysis
conclusion_generation_prompt: Prompt for generating conclusion
llm_model: The model to use for conclusion generation
+ langfuse_project_name: Name of the Langfuse project for tracking
Returns:
str: HTML-formatted conclusion content
@@ -323,15 +372,21 @@ def generate_conclusion(
# Prepare input data for conclusion generation
conclusion_input = {
- "main_query": state.main_query,
- "sub_questions": state.sub_questions,
+ "main_query": query_context.main_query,
+ "sub_questions": query_context.sub_questions,
"enhanced_info": {},
}
# Include enhanced information for each sub-question
- for question in state.sub_questions:
- if question in state.enhanced_info:
- info = state.enhanced_info[question]
+ info_source = (
+ synthesis_data.enhanced_info
+ if synthesis_data.enhanced_info
+ else synthesis_data.synthesized_info
+ )
+
+ for question in query_context.sub_questions:
+ if question in info_source:
+ info = info_source[question]
conclusion_input["enhanced_info"][question] = {
"synthesized_answer": info.synthesized_answer,
"confidence_level": info.confidence_level,
@@ -339,93 +394,99 @@ def generate_conclusion(
"key_sources": info.key_sources,
"improvements": getattr(info, "improvements", []),
}
- elif question in state.synthesized_info:
- # Fallback to synthesized info if enhanced info not available
- info = state.synthesized_info[question]
- conclusion_input["enhanced_info"][question] = {
- "synthesized_answer": info.synthesized_answer,
- "confidence_level": info.confidence_level,
- "information_gaps": info.information_gaps,
- "key_sources": info.key_sources,
- "improvements": [],
- }
- # Include viewpoint analysis if available
- if state.viewpoint_analysis:
+ # Include viewpoint analysis
+ if analysis_data.viewpoint_analysis:
+ va = analysis_data.viewpoint_analysis
conclusion_input["viewpoint_analysis"] = {
- "main_points_of_agreement": state.viewpoint_analysis.main_points_of_agreement,
+ "main_points_of_agreement": va.main_points_of_agreement,
"areas_of_tension": [
- {"topic": tension.topic, "viewpoints": tension.viewpoints}
- for tension in state.viewpoint_analysis.areas_of_tension
+ {"topic": t.topic, "viewpoints": t.viewpoints}
+ for t in va.areas_of_tension
],
- "perspective_gaps": state.viewpoint_analysis.perspective_gaps,
- "integrative_insights": state.viewpoint_analysis.integrative_insights,
+ "integrative_insights": va.integrative_insights,
}
# Include reflection metadata if available
- if state.reflection_metadata:
- conclusion_input["reflection_metadata"] = {
- "critique_summary": state.reflection_metadata.critique_summary,
- "additional_questions_identified": state.reflection_metadata.additional_questions_identified,
- "improvements_made": state.reflection_metadata.improvements_made,
+ if analysis_data.reflection_metadata:
+ rm = analysis_data.reflection_metadata
+ conclusion_input["reflection_insights"] = {
+ "improvements_made": rm.improvements_made,
+ "additional_questions_identified": rm.additional_questions_identified,
}
try:
- # Use the conclusion generation prompt
- conclusion_prompt_str = str(conclusion_generation_prompt)
-
- # Generate conclusion using LLM
- conclusion_html = run_llm_completion(
- prompt=json.dumps(conclusion_input, indent=2),
- system_prompt=conclusion_prompt_str,
+ # Call LLM to generate conclusion
+ result = run_llm_completion(
+ prompt=json.dumps(conclusion_input),
+ system_prompt=str(conclusion_generation_prompt),
model=llm_model,
- clean_output=True,
- max_tokens=1500, # Sufficient for comprehensive conclusion
+ temperature=0.7,
+ max_tokens=800,
project=langfuse_project_name,
+ tags=["conclusion_generation"],
)
- # Clean up any formatting issues
- conclusion_html = conclusion_html.strip()
+ if result:
+ content = remove_reasoning_from_output(result)
+ # Clean up the HTML
+ content = extract_html_from_content(content)
+ logger.info("Successfully generated LLM-based conclusion")
+ return content
+ else:
+ logger.warning("Failed to generate conclusion via LLM")
+ return generate_fallback_conclusion(query_context, synthesis_data)
+
+ except Exception as e:
+ logger.error(f"Error generating conclusion: {e}")
+ return generate_fallback_conclusion(query_context, synthesis_data)
- # Remove any h2 tags with "Conclusion" text that LLM might have added
- # Since we already have a Conclusion header in the template
- conclusion_html = re.sub(
- r" This research has explored the question: {html.escape(query_context.main_query)} Through systematic investigation of {len(query_context.sub_questions)} sub-questions, "
+ conclusion += (
+ "we have gathered insights from multiple sources and perspectives. "):
- # Wrap in paragraph tags if not already formatted
- conclusion_html = f" {conclusion_html} The research yielded {high_confidence} high-confidence findings out of "
+ conclusion += f"{len(info_source)} total areas investigated. This report has explored {html.escape(state.main_query)} through a structured research approach, examining {len(state.sub_questions)} focused sub-questions and synthesizing information from diverse sources. The findings provide a comprehensive understanding of the topic, highlighting key aspects, perspectives, and current knowledge. While some information gaps remain, as noted in the respective sections, this research provides a solid foundation for understanding the topic and its implications. Further research may be beneficial to address remaining information gaps "
+ conclusion += "and explore emerging questions identified during this investigation. {html.escape(content)} Note: This is a fallback report generated due to an error in the report generation process. This report presents findings related to the main research query. It explores multiple aspects of the topic through structured sub-questions and synthesizes information from various sources. This report addresses the research query: "{state.main_query}" The analysis is structured around {len(state.sub_questions)} sub-questions that explore different dimensions of this topic. This report investigates the research query: {html.escape(query_context.main_query)} The investigation was structured around {len(query_context.sub_questions)} key sub-questions to provide comprehensive coverage of the topic. Confidence Level: {confidence.upper()} {answer} {html.escape(info.synthesized_answer)} {info.information_gaps} Information Gaps: {html.escape(info.information_gaps)} No information available for this question. This report has explored the research query through multiple sub-questions, providing synthesized information based on available sources. While limitations exist in some areas, the report provides a structured analysis of the topic. This research has provided insights into the various aspects of the main query through systematic investigation. The findings represent a synthesis of available information, with varying levels of confidence across different areas. No references available. Sources were gathered from various search providers and synthesized to create this report.
- {category_title}
-
- """
-
+ for tension in va.areas_of_tension:
+ viewpoints_list = "\n".join(
+ [
+ f"
+
"""
- # Format the viewpoint analysis section using the template
+ # Format agreements (just the list items)
+ agreements_html = ""
+ if va.main_points_of_agreement:
+ agreements_html = "\n".join(
+ [
+ f"{html.escape(tension.topic)}
-
- {viewpoints_html}
-
+
+ {viewpoints_list}
+
"
+ # Generate references section
+ references_html = '
'
if all_sources:
for source in sorted(all_sources):
if source.startswith(("http://", "https://")):
- references_html += f'
- Research Report: {state.main_query}
-
-
-
+ Research Report: {html.escape(query_context.main_query)}
-
-
- Table of Contents
-
-
+
+ Note: This is a simplified version of the report generated due to processing limitations.
-
-
-
-
- Executive Summary
-
+
+
+ Introduction
-
+
+
Research Findings
"""
- # Add each sub-question and its synthesized information
- for i, sub_question in enumerate(state.sub_questions):
- safe_id = f"question-{i + 1}"
- info = state.enhanced_info.get(sub_question, None)
+ # Add findings for each sub-question
+ info_source = (
+ synthesis_data.enhanced_info
+ if synthesis_data.enhanced_info
+ else synthesis_data.synthesized_info
+ )
- if not info:
- # Try to get from synthesized info if not in enhanced info
- info = state.synthesized_info.get(sub_question, None)
-
- if info:
- answer = info.synthesized_answer
- confidence = info.confidence_level
-
- # Add appropriate confidence class
- confidence_class = ""
- if confidence == "high":
- confidence_class = "confidence-high"
- elif confidence == "medium":
- confidence_class = "confidence-medium"
- elif confidence == "low":
- confidence_class = "confidence-low"
+ for i, question in enumerate(query_context.sub_questions, 1):
+ if question in info_source:
+ info = info_source[question]
+ confidence_class = info.confidence_level.lower()
html += f"""
-
-
- """
- else:
- html += f"""
- {i + 1}. {sub_question}
-
-
+
+
- """
-
- html += """
- {i}. {html.escape(question)}
+ Confidence: {info.confidence_level.upper()}
+
-
- """
-
- # Add improvements if available
- if hasattr(info, "improvements") and info.improvements:
- html += """
- Information Gaps
-
-
- """
-
- # Add key sources if available
- if hasattr(info, "key_sources") and info.key_sources:
- html += """
- Improvements Made
-
- """
-
- for improvement in info.improvements:
- html += f"
-
- "
- html += """
-
- Key Sources
-
- """
+ if info.information_gaps:
+ html += f"
-
- """
-
- # Add conclusion section
html += """
+ {i + 1}. {sub_question}
-
- """
-
- # Add sources if available
- sources_set = set()
- for info in state.enhanced_info.values():
- if info.key_sources:
- sources_set.update(info.key_sources)
-
- if sources_set:
- html += """
- Conclusion
-
-
- """
- else:
- html += """
- References
-
- """
-
- for source in sorted(sources_set):
- html += f"
-
+
+
-
- """
-
- # Close the HTML structure
- html += """
References
-