What is XunLong?
XunLong (寻龙, meaning "Dragon Seeker") is an AI-powered multi-modal content generation system that automates the creation of research reports, creative fiction, and professional presentations.
The Vision
In today's information age, content creation is both crucial and time-consuming. Whether you're a researcher compiling findings, a writer crafting stories, or a professional preparing presentations, the process often involves:
- 🔍 Hours of research - Searching, reading, and extracting relevant information
- ✍️ Manual writing - Organizing thoughts and crafting coherent narratives
- 🎨 Formatting - Designing layouts and ensuring visual appeal
- 🔄 Iterations - Revising and refining until satisfactory
XunLong automates this entire workflow, allowing you to generate professional-quality content in minutes instead of hours.
What Makes XunLong Different?
1. Multi-Modal Generation
Unlike single-purpose AI tools, XunLong supports three distinct content types:
| Mode | Purpose | Output Formats |
|---|---|---|
| Report | Research reports, analysis documents, white papers | MD, HTML, PDF, DOCX |
| Fiction | Novels, short stories, creative writing | MD, HTML, PDF, DOCX |
| PPT | Business presentations, pitch decks, slides | JSON, HTML, PPTX |
2. Intelligence Through Multi-Agent Architecture
XunLong doesn't rely on a single AI prompt. Instead, it orchestrates multiple specialized agents that collaborate:
- Coordinator - Breaks down complex tasks
- Search Agent - Gathers relevant information
- Generation Agent - Creates content
- Review Agent - Ensures quality
- Iteration Agent - Refines based on feedback
3. Real Web Search
XunLong doesn't hallucinate facts. It performs actual web searches using Playwright and Perplexity to gather current, factual information before generating content.
4. Iterative Refinement
Not satisfied with the initial output? XunLong supports intelligent iterations:
# Initial generation
python xunlong.py report "AI Industry 2025"
# Iterate and refine
python xunlong.py iterate <project_id> "Add more case studies in chapter 3"The system understands context and makes targeted modifications without regenerating everything.
5. Full Observability
Every LLM call, search query, and generation step is tracked via LangFuse integration, giving you complete transparency into how content is created.
Use Cases
📊 For Researchers & Analysts
Generate comprehensive research reports:
- Industry analysis
- Market research
- Technical documentation
- Literature reviews
Example:
python xunlong.py report "Quantum Computing Applications in Finance" \
--style academic \
--depth comprehensive✍️ For Writers & Content Creators
Create engaging fiction:
- Novels and novellas
- Short stories
- Creative narratives
- Character-driven plots
Example:
python xunlong.py fiction "A detective story set in cyberpunk Tokyo" \
--style mystery \
--chapters 15📈 For Business Professionals
Build professional presentations:
- Business pitches
- Product launches
- Quarterly reviews
- Training materials
Example:
python xunlong.py ppt "Q4 2025 Sales Review" \
--slides 20 \
--style business \
--speech-notes "Executive presentation"How It Works
Simple 3-Step Process
Input Your Request
bashpython xunlong.py report "Your Topic Here"Let XunLong Work
- Analyzes your request
- Searches for information
- Generates structured content
- Creates formatted output
Review & Iterate
- Check the generated content
- Request modifications if needed
- Export to desired format
Behind the Scenes
Technology Stack
XunLong is built on proven, industry-standard technologies:
- LangGraph - Workflow orchestration
- LangChain - LLM integration framework
- OpenAI / Anthropic / DeepSeek - Language models
- Playwright - Web automation
- WeasyPrint - PDF generation
- python-pptx - PowerPoint creation
- LangFuse - Observability and monitoring
Key Features
🔍 Intelligent Search
- Real-time web search
- Content extraction and summarization
- Source citation and verification
✨ Smart Generation
- Context-aware writing
- Style adaptation
- Consistent formatting
- Proper structure
🎨 Professional Export
- Multiple format support
- Template customization
- High-quality PDF rendering
- PPTX with layouts and styling
🔄 Iterative Improvement
- Targeted modifications
- Version management
- Context preservation
- Incremental refinement
📈 Full Transparency
- LLM call tracking
- Token usage monitoring
- Performance metrics
- Error logging
Comparison
| Feature | XunLong | ChatGPT | Traditional Tools |
|---|---|---|---|
| Multi-modal | ✅ Reports, Fiction, PPT | ❌ Text only | ❌ Specialized |
| Real search | ✅ Live web search | ⚠️ Limited | ❌ Manual |
| Export | ✅ PDF, DOCX, PPTX | ❌ Copy-paste | ✅ Native |
| Iteration | ✅ Context-aware | ⚠️ New chat | ❌ Manual edit |
| Observability | ✅ Full tracking | ❌ Black box | N/A |
| Offline | ❌ API required | ❌ Internet needed | ✅ Offline |
| Cost | 💰 API costs | 💰 Subscription | 💰 License |
Limitations
XunLong is designed for content generation, not as a:
- ❌ Real-time chat interface
- ❌ Code execution environment
- ❌ Data analysis platform
- ❌ Image generation tool
Current constraints:
- Requires stable internet connection
- LLM API costs apply
- Generation time: 5-20 minutes per project
- Quality depends on LLM model chosen
Philosophy
XunLong follows these core principles:
- Quality over Speed - Takes time to research and generate properly
- Transparency - Every step is visible and trackable
- Flexibility - Supports multiple models and can be customized
- Iterative - Content can be refined, not just regenerated
- Open Source - Community-driven development under MIT license
Next Steps
Ready to get started?
Community & Support
XunLong - Empowering content creators with AI