Report Generation

XunLong excels at generating comprehensive, well-researched reports on any topic with minimal input.

Overview

The Report Generation feature automatically:

🔍 Researches your topic across the web
📊 Structures findings logically
✍️ Writes professional content
📚 Cites all sources
📄 Exports to multiple formats

What's New (2025)

Recent pipeline upgrades deliver higher quality output with lower latency:

Section-Level Agents – every outline node spins up its own writer and reviewer, so chapters are drafted, evaluated, and enhanced in parallel.
Built-in Visualization Pass – each section is analysed for tabular or chart-worthy data; when required the data-visualization agent emits HTML-ready tables or chart specs.
User Document Ingestion – upload .txt, .pdf, or .docx files via --input-file; the content is parsed, summarised, and treated as a high-priority knowledge source throughout the pipeline (image files are not supported yet).
Time-Aware Research – the deep-search layer now understands queries such as “daily report” or explicit dates, applies DuckDuckGo date filters, and prioritises chronologically relevant sources.
Clean HTML Rendering – headings are de-duplicated, Markdown is normalised before rendering, and the academic template consumes pre-rendered section fragments to avoid formatting drift.

Quick Start

bash

python xunlong.py report "AI Industry Trends 2025"

# Bootstrap from an existing company brief
python xunlong.py report "AI Startup Business Plan" \
  --type business \
  --input-file ./docs/company_overview.pdf

That's it! XunLong handles the rest.

Report Styles

XunLong supports three professional report styles:

Business Style 💼

Best for: Market analysis, industry reports, business intelligence

Characteristics:

Executive summary with key takeaways
Data-driven insights
Professional tone
Charts and tables
ROI and metrics focus

Example:

bash

python xunlong.py report "Electric Vehicle Market Analysis" \
  --style business \
  --depth comprehensive

Sample Output Structure:

├── Executive Summary
├── Market Overview
│   ├── Market Size & Growth
│   └── Key Players
├── Trend Analysis
├── Competitive Landscape
├── Opportunities & Challenges
└── Recommendations

Academic Style 🎓

Best for: Research summaries, literature reviews, academic papers

Characteristics:

Abstract and introduction
Rigorous citations
Methodological approach
Literature review sections
Formal academic tone

Example:

bash

python xunlong.py report "Machine Learning in Healthcare" \
  --style academic \
  --depth comprehensive

Sample Output Structure:

├── Abstract
├── Introduction
├── Literature Review
├── Methodology
├── Findings
├── Discussion
├── Conclusion
└── References

Technical Style 🔧

Best for: Technology deep-dives, API documentation, technical specifications

Characteristics:

Technical accuracy
Code examples
Architecture diagrams
Implementation details
Best practices

Example:

bash

python xunlong.py report "GraphQL vs REST APIs" \
  --style technical \
  --depth comprehensive

Sample Output Structure:

├── Overview
├── Technical Architecture
├── Core Concepts
├── Implementation Guide
├── Code Examples
├── Performance Analysis
└── Best Practices

Depth Levels

Control the level of detail with the --depth parameter:

Depth	Time	Words	Best For
overview	~5 min	1,500-2,000	Quick summaries, initial research
standard	~10 min	3,000-4,000	Most use cases, balanced detail
comprehensive	~20 min	6,000-8,000	In-depth analysis, presentations

Examples:

bash

# Quick overview
python xunlong.py report "Quantum Computing" --depth overview

# Balanced report (default)
python xunlong.py report "Quantum Computing" --depth standard

# Deep dive
python xunlong.py report "Quantum Computing" --depth comprehensive

Advanced Features

Section-Oriented Pipeline

Outline Planning – the coordinator produces a structured outline (id, title, requirements, word budget).
Parallel Section Agents – each outline item is assigned to a dedicated writer that drafts content with awareness of neighbouring sections.
Automatic Visualisation – if a section contains structured data, a visualiser agent decides whether to embed tables or charts and generates ready-to-render HTML fragments.
Quality Gate & Rewrite Loop – evaluators score each section for coverage and coherence; low scoring sections are rewritten until they pass or reach the retry limit.
HTML Assembly – the final assembler streams the pre-rendered section HTML into the document template, ensuring headings, tables, and charts remain stable.

Because every stage runs asynchronously the full report finishes noticeably faster, especially for comprehensive or daily reports with strict timelines.

Custom Sections

Specify exactly what sections you want:

bash

python xunlong.py report "AI Ethics" \
  --sections "Introduction,Current Challenges,Case Studies,Future Outlook"

Time-Bounded Research

The search layer now applies automatic date filters:

Queries containing phrases like “daily report”, specific dates, or “today/yesterday” are scoped to the appropriate day or week automatically.
DuckDuckGo searches include df= parameters so only fresh sources are retrieved.

You can still force a manual window when needed:

bash

python xunlong.py report "COVID-19 Vaccines" \
  --time-range "last-6-months"

Language Support

Generate reports in multiple languages:

bash

python xunlong.py report "气候变化影响" --language zh-CN
python xunlong.py report "Climate Change Impact" --language en-US

Source Filtering

Control what sources are used:

bash

# Only academic sources
python xunlong.py report "Dark Matter" \
  --sources academic

# News sources only
python xunlong.py report "Tech Industry Layoffs" \
  --sources news

# All sources (default)
python xunlong.py report "AI Trends" \
  --sources all

Output Formats

Markdown (Default)

bash

python xunlong.py report "Topic" --format md

Features:

Clean, readable text
Easy to edit
Version control friendly
Portable

HTML

bash

python xunlong.py report "Topic" --format html

Features:

Academic-styled template with pre-rendered section HTML
Stable table of contents and anchors
Responsive design with inline tables/charts
Print-ready output

PDF

bash

python xunlong.py report "Topic" --format pdf

Features:

Professional layout
Page numbers
Headers/footers
Ready to share

DOCX

bash

python xunlong.py report "Topic" --format docx

Features:

Microsoft Word compatible
Editable formatting
Comments support
Track changes ready

Multiple Formats

bash

python xunlong.py report "Topic" --format md,html,pdf,docx

All formats are generated simultaneously.

Report Quality

Citations

Every fact is cited with:

Source URL
Publication date
Author (when available)
Access date

Example Citation:

markdown

According to recent research, AI adoption has increased by 67% [1].

## References
[1] Smith, J. (2025). "AI in Enterprise." TechReview.
    https://example.com/ai-enterprise
    Accessed: 2025-10-05

Quality Metrics

XunLong tracks:

Citation Coverage: % of claims cited (target: >80%)
Source Diversity: Number of unique sources (target: >10)
Readability Score: Flesch reading ease (target: 60-70)
Coherence Score: Logical flow rating (target: >0.85)

View metrics:

bash

python xunlong.py stats <project-id>

Fact-Checking

Reports undergo automatic fact-checking:

✅ Date verification
✅ Statistical consistency
✅ Source credibility check
✅ Claim cross-referencing

Example Workflow

1. Generate Initial Report

bash

python xunlong.py report "Renewable Energy Trends 2025" \
  --style business \
  --depth standard \
  --format md,pdf

Output:

✅ Report generated successfully!

📊 Statistics:
   - Duration: 8m 34s
   - Words: 3,847
   - Citations: 23 sources
   - Quality Score: 0.89

📁 Files:
   - storage/20251005_143022_renewable_energy/reports/FINAL_REPORT.md
   - storage/20251005_143022_renewable_energy/exports/report.pdf

🔗 Project ID: 20251005_143022

2. Review Content

bash

cat storage/20251005_143022_renewable_energy/reports/FINAL_REPORT.md

3. Request Modifications

bash

python xunlong.py iterate 20251005_143022 \
  "Add a section on solar energy costs and expand the wind energy section"

4. Export Additional Formats

bash

python xunlong.py export 20251005_143022 --format docx,html

Best Practices

📝 Writing Effective Queries

Good:

"AI ethics challenges in autonomous vehicles"
"Impact of remote work on productivity 2025"
"Comparison of Python vs JavaScript for web development"

Less Effective:

"AI" (too broad)
"stuff about work" (too vague)
"tell me everything about programming" (unfocused)

🎯 Choosing the Right Style

Your Goal	Recommended Style
Investor presentation	Business
Research paper	Academic
Internal tech doc	Technical
Blog post	Business (lighter tone)
White paper	Academic or Business

⚡ Optimizing Generation Time

Fast (~5 min):

bash

python xunlong.py report "Topic" \
  --depth overview \
  --model gpt-3.5-turbo

Balanced (~10 min):

bash

python xunlong.py report "Topic" \
  --depth standard \
  --model gpt-4o-mini

High Quality (~20 min):

bash

python xunlong.py report "Topic" \
  --depth comprehensive \
  --model gpt-4o

Troubleshooting

Issue: "No relevant sources found"

Solutions:

Make your query more specific
Check your internet connection
Try different search terms
Verify topic is searchable

Issue: Report is too short

Solutions:

Increase depth: --depth comprehensive
Add more sections: --sections "Section1,Section2,..."
Use slower, more powerful model: --model gpt-4o

Issue: Citations missing

Solutions:

Enable strict citations: --strict-citations
Increase source count: --min-sources 15
Check search results quality

Issue: Generation failed mid-way

Solutions:

bash

# Resume from checkpoint
python xunlong.py resume <project-id>

# Check error logs
cat storage/<project-id>/logs/generation.log

API Reference

Command Structure

bash

python xunlong.py report <query> [options]

Required Arguments

<query>: Topic or research question

Optional Arguments

Argument	Type	Default	Description
`--style`	str	`business`	Report style (business/academic/technical)
`--depth`	str	`standard`	Detail level (overview/standard/comprehensive)
`--format`	str	`md`	Output formats (md,html,pdf,docx)
`--language`	str	`en-US`	Report language
`--sections`	str	Auto	Custom section list
`--time-range`	str	`all`	Research time window
`--sources`	str	`all`	Source type filter
`--model`	str	`gpt-4o-mini`	LLM model to use
`--min-sources`	int	`10`	Minimum sources required
`--strict-citations`	flag	`false`	Enforce citation on every claim

Examples

Market Research Report

bash

python xunlong.py report "Global Smartphone Market 2025" \
  --style business \
  --depth comprehensive \
  --sections "Market Size,Top Players,Trends,Forecast,Recommendations" \
  --format pdf,docx

Academic Literature Review

bash

python xunlong.py report "Deep Learning in Medical Imaging" \
  --style academic \
  --depth comprehensive \
  --sources academic \
  --strict-citations \
  --format md,pdf

Technical Documentation

bash

python xunlong.py report "Kubernetes Best Practices 2025" \
  --style technical \
  --depth standard \
  --format md,html

Quick News Summary

bash

python xunlong.py report "Latest AI Developments" \
  --depth overview \
  --time-range last-week \
  --sources news \
  --format md

Next Steps

Learn about Fiction Writing
Explore PPT Creation
Understand Content Iteration
Check Export Formats

Report Generation ​

Overview ​

What's New (2025) ​

Quick Start ​

Report Styles ​

Business Style 💼 ​

Academic Style 🎓 ​

Technical Style 🔧 ​

Depth Levels ​

Advanced Features ​

Section-Oriented Pipeline ​

Custom Sections ​

Time-Bounded Research ​

Language Support ​

Source Filtering ​

Output Formats ​

Markdown (Default) ​

HTML ​

PDF ​

DOCX ​

Multiple Formats ​

Report Quality ​

Citations ​

Quality Metrics ​

Fact-Checking ​

Example Workflow ​

1. Generate Initial Report ​

2. Review Content ​

3. Request Modifications ​

4. Export Additional Formats ​

Best Practices ​

📝 Writing Effective Queries ​

🎯 Choosing the Right Style ​

⚡ Optimizing Generation Time ​

Troubleshooting ​

Issue: "No relevant sources found" ​

Issue: Report is too short ​

Issue: Citations missing ​

Issue: Generation failed mid-way ​

API Reference ​

Command Structure ​

Required Arguments ​

Optional Arguments ​

Examples ​

Market Research Report ​

Academic Literature Review ​

Technical Documentation ​

Quick News Summary ​

Next Steps ​

Report Generation

Overview

What's New (2025)

Quick Start

Report Styles

Business Style 💼

Academic Style 🎓

Technical Style 🔧

Depth Levels

Advanced Features

Section-Oriented Pipeline

Custom Sections

Time-Bounded Research

Language Support

Source Filtering

Output Formats

Markdown (Default)

HTML

PDF

DOCX

Multiple Formats

Report Quality

Citations

Quality Metrics

Fact-Checking

Example Workflow

1. Generate Initial Report

2. Review Content

3. Request Modifications

4. Export Additional Formats

Best Practices

📝 Writing Effective Queries

🎯 Choosing the Right Style

⚡ Optimizing Generation Time

Troubleshooting

Issue: "No relevant sources found"

Issue: Report is too short

Issue: Citations missing

Issue: Generation failed mid-way

API Reference

Command Structure

Required Arguments

Optional Arguments

Examples

Market Research Report

Academic Literature Review

Technical Documentation

Quick News Summary

Next Steps