Lawyer Knowledge Graph

What Lawyer Knowledge Graph does

Lawyer Knowledge Graph has two modules. The first is a knowledge system: a lawyer pastes a judgment or case summary, an LLM extracts knowledge cards (dispute focus, adjudication rule, statute application, practice experience), and the cards are vectorized and stored in PostgreSQL 16 with pgvector. Entering a new case description returns semantically similar cards with similarity scores. The backend is FastAPI, the interface is Streamlit, vectors come from text2vec-large-chinese, and the extraction model can be DeepSeek, Kimi or GLM.

The second module is a skill collection in skills/legal for Hermes Agent, covering the case workflow, including case analysis, evidence organization, legal research, document drafting, trial preparation and trial response. The skills run in conversation and need a Hermes Agent environment. The knowledge system starts with a deploy script or docker-compose, and two standalone scripts, extract_cards.py and vector_search.py, run without the full service. A statute-tracking feature is marked as planned.

Key features

• Knowledge cards extracted from judgments: dispute focus, adjudication rule, statute application and practice experience
• Semantic similar-case search over pgvector with similarity scores
• Hermes Agent skills for case analysis, evidence organization, legal research, document drafting and trial preparation
• Docker Compose stack with a FastAPI backend, Streamlit frontend and text2vec vector service
• Standalone scripts extract_cards.py and vector_search.py
• deploy.sh script that guides setup of the LLM model and API key

When to use it

• Finding earlier cases with a similar dispute before starting new research
• Drafting legal documents and preparing for trial from a Hermes chat
• Building a personal archive of case experience that grows with each matter

Who it is for: Litigation lawyers working with Chinese civil and commercial law who want an AI-assisted case archive and Hermes skills.

How it fits with Hermes Agent

The skills module is written for Hermes Agent and requires a Hermes Agent runtime, while the knowledge system runs as a separate Docker stack.

How to install Lawyer Knowledge Graph

These commands are copied from the project's README. Check the repository for the latest steps before you run them.

git clone https://github.com/wxlawyers/lawyer-knowledge-graph.git
cd lawyer-knowledge-graph
bash deploy.sh

Requirements: Docker Engine 24+ with Docker Compose v2, 4 CPU cores, 8 GB RAM (16 GB recommended), 20 GB free disk, an LLM API key and network access to the LLM API; Hermes Agent for the skills

Note: The README is written in Chinese, and its statute-tracking feature is marked as planned rather than built.

FAQ