Hypergraph is more powerful. Transform unstructured text into structured knowledge with LLMs. Graphs, hypergraphs, and spatio-temporal extractions — with one command.
{ width="500" }
Transform documents into structured knowledge with one command.
"告别文档焦虑,让信息一目了然"
Hyper-Extract is an intelligent, LLM-powered knowledge extraction framework. It transforms unstructured text into persistent, predictable, and strongly-typed knowledge structures—from simple lists to complex knowledge graphs, hypergraphs, and spatio-temporal graphs.
⚡ 5-Minute Quick Start
=== "CLI (Terminal)"
```bash
# 1. Install CLI tool
uv tool install hyperextract
# 2. Configure API Key (Bailian example)
he config init -p bailian -k YOUR_BAILIAN_API_KEY
# 3. Extract knowledge from a document
he parse tesla.md -t general/biography_graph -o ./output/ -l en
# 4. Visualize the knowledge graph
he show ./output/
```
=== "Python"
```python
from hyperextract import Template
# 1. Create a template
ka = Template.create("general/biography_graph", "en")
# 2. Extract knowledge
with open("tesla.md") as f:
result = ka.parse(f.read())
# 3. Visualize
result.show()
```
→ Ready to dive deeper? Check out the Getting Started Guide or jump to CLI / Python SDK documentation.
✨ What Makes Hyper-Extract Different?
:material-file-pdf-box:{ .lg .middle } Rich Document Ingestion
Feed PDF, Word, PowerPoint, Excel, HTML, EPUB and more — not just
.txt/.md. One optional extra:pip install "hyperextract[ingest]".:material-shape:{ .lg .middle } 9 Auto-Types
From simple
AutoList/AutoModelto advancedAutoGraph,AutoHypergraph, andAutoSpatioTemporalGraph. Pick the right structure for your data.:material-brain:{ .lg .middle } 11+ Extraction Engines
Built-in support for the
chunk_ragbaseline, GraphRAG, LightRAG, Hyper-RAG, KG-Gen, iText2KG, and more. Choose the best method for your use case.:material-file-document:{ .lg .middle } 80+ Domain Templates
Ready-to-use templates for Finance, Legal, Medical, TCM, and Industry. Zero configuration needed.
:material-sync:{ .lg .middle } Incremental Evolution & Provenance
Feed new documents to continuously expand your knowledge abstract — every source is attributed, and removal or rollback is incremental too. No full reprocessing, ever.
🎯 Choose Your Path
:material-console:{ .lg .middle } CLI User
Process documents directly from your terminal. Perfect for:
- Quick knowledge extraction
- Batch document processing
- Building knowledge abstracts without coding
:material-language-python:{ .lg .middle } Python Developer
Integrate into your Python applications. Perfect for:
- Custom extraction pipelines
- Integration with existing workflows
- Building AI-powered applications
:material-school:{ .lg .middle } Want to Learn More?
Understand the concepts and architecture:
- How Auto-Types work
- Choosing extraction methods
- Creating custom templates
🧩 The 8 Auto-Types at a Glance
| Type | Use Case | Example Output |
|---|---|---|
| AutoModel | Structured summaries | A pydantic model with specific fields |
| AutoList | Collections of items | A list of entities or facts |
| AutoSet | Deduplicated collections | A set of unique items |
| AutoGraph | Entity-relationship networks | Knowledge graph with nodes and edges |
| AutoHypergraph | Multi-entity relationships | Hyperedges connecting multiple nodes |
| AutoTemporalGraph | Time-based relationships | Graph with temporal information |
| AutoSpatialGraph | Location-based relationships | Graph with geographic data |
| AutoSpatioTemporalGraph | Time + Space combined | Full context with when and where |
→ Learn which Auto-Type to choose
🏗️ Architecture Overview
Hyper-Extract follows a three-layer architecture:
graph TD
A[Your Document] --> B[CLI / Python API]
B --> C[Template]
B --> D[Method]
C --> E[Auto-Type]
D --> E
E --> F[Structured Knowledge]
subgraph "Layer 3: Templates & Methods"
C
D
end
subgraph "Layer 2: Core Engine"
E
end
subgraph "Layer 1: Output"
F
end
- Auto-Types — Define the output data structure (8 types)
- Methods — Provide extraction algorithms (RAG-based and Typical)
- Templates — Offer domain-specific, ready-to-use configurations
You can use Hyper-Extract at any level: pick a template for quick results, choose a method for more control, or work directly with Auto-Types for full customization.
📊 Comparison with Other Tools
| Feature | GraphRAG | LightRAG | KG-Gen | Hyper-Extract |
|---|---|---|---|---|
| Knowledge Graph | ✅ | ✅ | ✅ | ✅ |
| Temporal Graph | ✅ | ❌ | ❌ | ✅ |
| Spatial Graph | ❌ | ❌ | ❌ | ✅ |
| Hypergraph | ❌ | ❌ | ❌ | ✅ |
| Domain Templates | ❌ | ❌ | ❌ | ✅ |
| CLI Tool | ✅ | ❌ | ❌ | ✅ |
| Multi-language | ✅ | ❌ | ❌ | ✅ |
📚 Documentation Structure
- Getting Started — Installation and your first extraction
- CLI Guide — Complete terminal workflow documentation
- Python SDK — API reference and developer guides
- Concepts — Understanding the architecture
- Templates — Domain-specific extraction templates
- Resources — FAQ, troubleshooting, and contributing
🤝 Contributing
Contributions are welcome! Whether it's bug reports, feature requests, or documentation improvements, please feel free to submit an issue or pull request.
:fontawesome-brands-github: GitHub Repository{ .md-button .md-button--primary }
📄 License
Hyper-Extract is licensed under the Apache-2.0 License.