Hyper-Extract is a Python CLI tool that transforms unstructured text into structured knowledge, supporting graphs, hypergraphs, and spatio-temporal extractions via one command.
Collecting history — the radar snapshots this repo daily. The trend line appears after 3 days of data (1 so far).
What it is
Hyper-Extract is an intelligent, LLM-powered knowledge extraction CLI. It transforms documents into structured knowledge with one command and supports a range of structures including Graphs, Hypergraphs, and Spatio-Temporal Graphs.
How it works
The project provides a CLI and templates to extract information into predefined data structures. It supports multiple providers (OpenAI, Anthropic Claude, DeepSeek, Bailian, Local vLLM) and embedding options. It offers a three-layer architecture: Auto-Types (8 data structures), Methods (KG-Gen, GraphRAG, LightRAG, Hyper-RAG, Cog-RAG, etc.), and Templates (80+ presets across domains).
Getting started
Install (30-Second Quick Start):
# Install uv first (if you haven't)
curl -LsSf https://astral.sh/uv/install.sh | sh
# Install Hyper-Extract CLI
uv tool install hyperextract
# or: pipx install hyperextract
Configure a provider (example):
# OpenAI
he config init -p openai -k YOUR_OPENAI_API_KEY
Extract, query & visualize:
# Extract knowledge from a document
he parse examples/en/tesla.md -t general/biography_graph -o ./output/ -l en
# Query it
he search ./output/ "What are Tesla's major achievements?"
# Visualize
he show ./output/
# Export to an Obsidian vault (Markdown notes + [[wikilinks]])
he export obsidian ./output/ -o ./vault/
Python API example:
from hyperextract import Template
ka = Template.create("general/biography_graph")
with open("examples/en/tesla.md") as f:
result = ka.parse(f.read())
result.show()
Getting started (providers and models)
Supported platforms include OpenAI, Anthropic Claude, DeepSeek, Bailian, and Local vLLM. Embedding models include text-embedding-3-small, text-embedding-v4, and bge-m3 for local setups. DeepSeek V4 models default to thinking mode and must be paired with an OpenAI-compatible embedder. Anthropic models can be used as LLMs (embedder via OpenAI-compatible model).
Recent releases
Latest releases include:
- v0.4.0 (2026-08-01): DeepSeek Native Provider via create_client(llm="deepseek"). DeepSeek V4 models default to "think" mode. Also introduces DeepSeek provider support.
- v0.3.1 (2026-08-01): Patch release with merge_batch_data empty-first-chunk fix and other reliability fixes landed on main since v0.3.0.
- v0.3.0 (2026-06-19): Anthropic Claude Provider native support for claude-opus-4-8, claude-sonnet-4-6, claude-haiku-4-5 via create_client("anthropic", ...).
- v0.2.0 (2026-05-18): Unified Provider System introducing create_client() interface with multiple providers.
- v0.1.2 (2026-04-11): CLI bug fix release.
Traction
Stars: 3237. Forks: 383. Open issues: 1.
Behind the repo
The project provides a MCP Server for exposing knowledge abstracts to MCP-capable assistants with commands like list_templates, info, search, ask, and export_obsidian. It can be installed via pip as hyperextract[mcp]. The repository is written in Python and has ongoing development activity with last_push dated 2026-08-03.
Caveats
License: Apache-2.0 is stated for the project. Supports OpenAI, Anthropic, DeepSeek, Bailian, and Local vLLM integrations; licensing terms for providers outside Apache-2.0 are not included in the README excerpt. Last push: 2026-08-03. Releases indicate ongoing updates including bug fixes and provider additions.