Skip to content

Hypergraph is more powerful. Transform unstructured text into structured knowledge with LLMs. Graphs, hypergraphs, and spatio-temporal extractions — with one command.

5.6 KB
Raw Blame

Hyper-Extract CLI

"Stop reading. Start understanding."

A powerful command-line tool for extracting structured knowledge from unstructured text.


⚡ Quick Start

# Install
uv tool install hyperextract

# First-time setup
he config init

# Extract knowledge from a document
he parse document.md -o my_ka -l zh

# Visualize the knowledge graph
he show my_ka

# Search within your Knowledge Abstract
he search my_ka "key insights"

⚙️ Configuration

One-Command Setup (Simplest)

# Just provide your API key - defaults work for most users
he config init --api-key YOUR_API_KEY

# With custom base URL
he config init -k YOUR_KEY -u https://your-endpoint.com/v1

This automatically configures both LLM and Embedder with:

  • LLM: gpt-4o-mini
  • Embedder: text-embedding-3-small

Interactive Setup

he config init

Manual Configuration

For advanced users who need separate configurations:

# Configure LLM
he config llm --api-key YOUR_KEY --model gpt-4o

# Configure Embedder
he config embedder --api-key YOUR_KEY --model text-embedding-3-small

Environment Variables

You can also use environment variables as an alternative:

export OPENAI_API_KEY=your_api_key
export OPENAI_BASE_URL=https://api.openai.com/v1  # Optional

Environment variables take precedence over config file settings.

View Current Configuration

he config show

📄 Parse Command

Extract knowledge from documents into structured Knowledge Abstracts.

Basic Usage

# Parse with interactive template selection
he parse document.md -o my_ka -l zh

# Parse with specific template
he parse document.md -o my_ka -t general/knowledge_graph -l zh

# Parse with specific method
he parse document.md -o my_ka -m light_rag

Options

  • <input> - Input file path, directory, or - for stdin
  • -o, --output - Output Knowledge Abstract directory (required)
  • -t, --template - Template ID (omit for interactive selection)
  • -m, --method - Method template (e.g., light_rag, hyper_rag)
  • -l, --lang - Language (zh or en, required for knowledge templates)
  • -f, --force - Force overwrite existing output
  • --no-index - Skip building search index

List Available Templates

he list template
he list template -l zh  # Filter by language
he list template -a graph  # Filter by type
he list template -q finance  # Search by keyword

List Available Methods

he list method
he list method -q rag  # Search by keyword

🔍 Other Commands

Build Search Index

Required for semantic search and chat functionality.

he build-index my_ka
he build-index my_ka --force  # Rebuild existing index

Search Knowledge Abstract

Perform semantic search within your Knowledge Abstract.

he search my_ka "What are the key findings?"
he search my_ka "key insights" -n 5  # Return top 5 results

Chat with Knowledge Abstract

Ask questions about your Knowledge Abstract.

# Single query
he talk my_ka -q "What was discussed in the meeting?"

# Interactive mode
he talk my_ka -i

Visualize Knowledge Graph

View your Knowledge Abstract as an interactive graph.

he show my_ka

View Knowledge Abstract Info

Display statistics and metadata about your Knowledge Abstract.

he info my_ka

Add Knowledge to Existing KA

Append new knowledge to an existing Knowledge Abstract.

he feed my_ka new_document.md

📝 Examples

Example 1: Extract Financial Data

# Configure API keys
he config init

# List finance templates
he list template -l zh | grep finance

# Extract earnings report
he parse earnings_report.md -o finance_ka -t finance/earnings_summary -l zh

# Build index for search
he build-index finance_ka

# Search for insights
he search finance_ka "What was the revenue growth?"

Example 2: Extract Legal Contracts

# List legal templates
he list template -l zh | grep legal

# Extract contract information
he parse contract.md -o legal_ka -t legal/contract_summary -l zh

# View as knowledge graph
he show legal_ka

Example 3: Use Method Templates

# Use Light RAG method
he parse document.md -o ka -m light_rag

# Use Hyper RAG method
he parse document.md -o ka -m hyper_rag

❓ FAQ

Q: How do I choose between template and method?

A: Use templates when you need domain-specific extraction (finance, legal, medicine, etc.). Use methods when you want algorithm-driven extraction (RAG-based approaches).

Q: Why do I need to build an index?

A: The index enables semantic search and chat functionality. Without it, you can still extract and visualize knowledge, but search and talk commands won't work.

Q: How do I switch between languages?

A: Use the -l or --lang option:

  • -l zh for Chinese
  • -l en for English

Q: Can I use custom API endpoints?

A: Yes! Configure with --base-url to use API-compatible endpoints:

he config llm --api-key YOUR_KEY --base-url https://your-custom-api.com/v1

Q: Where is the configuration stored?

A: Configuration is stored in ~/.he/config.toml

Q: How do I update my API key?

A: Simply run the configuration command again:

he config llm --api-key NEW_API_KEY

🆘 Need Help?

# View all available commands
he --help

# View help for specific command
he parse --help
he config --help
he search --help

📚 Learn More