LightRAG Skill

This skill allows you to interact with one or more LightRAG API servers. You can perform queries in various modes (local, global, hybrid, mix, naive) and use the retrieved context for further processing.

Configuration

The skill uses a configuration file at ~/.lightrag_config.json to store server details.

Format:

{
  "servers": {
    "alias1": {
      "url": "http://server1:9621",
      "api_key": "optional_key"
    },
    "alias2": {
      "url": "http://server2:9621",
      "api_key": "optional_key"
    }
  },
  "default_server": "alias1"
}

Workflows

1. Direct Search

To find information, use scripts/query_lightrag.py.

Modes: local, global, hybrid, mix, naive.

2. Using Context for Writing

To use a knowledge base as context (e.g., for a test or article):

1. Run query_lightrag.py with the --only-context flag.
2. Pass the resulting context to your writing task/model.

Reference

See API_DOCS.md for endpoint details.