0.3.1

AgensGraph AI

Integrations, tools and starter material for building AI applications on AgensGraph, the PostgreSQL-based graph database. Every integration here runs on the agensgraph-python 2.0 driver.

Each directory has a README of its own with setup and usage, and an examples/demos/ suite that runs on real datasets.

Installation

The agensgraph-ai package installs any combination of the integrations under one name. Pick the ones you need:

pip install "agensgraph-ai[langchain]"
pip install "agensgraph-ai[langchain,lightrag]"
pip install "agensgraph-ai[all]"

Name at least one extra. pip install agensgraph-ai on its own installs no integrations, and an extra that is misspelled installs none either — pip warns about that, uv does not.

Each integration is also released on its own, so it can be installed by name instead. The two forms produce the same environment — agensgraph-ai ships no code, and is a convenience rather than a layer:

pip install langchain-agensgraph
pip install llama-index-agensgraph
pip install lightrag-agensgraph
pip install cognee-agensgraph

The MCP servers are commands rather than libraries, and an MCP client normally launches them itself with uvx, which needs no install at all:

"mcpServers": {
  "agensgraph-cypher": {
    "command": "uvx",
    "args": ["mcp-agensgraph-cypher", "--transport", "stdio"]
  }
}

The mcp extra is for the other case: hosting a server yourself over HTTP or SSE.

Requirements

• Python 3.11 or later.
• A running AgensGraph 2.17 or later; the driver refuses an older server at connect. SHOW agversion tells you which you have. From 2.18 a release reports four numbers, such as 2.18.4.0, with -rc1 on a release candidate; the first two are the line, so 2.18.6.0-rc1 is a 2.18 server.
• The vector-backed features need the pgvector extension, and schema introspection is faster with the meta extension. AgensGraph bundles neither; see how to build them.

License

Apache License 2.0 — see LICENSE.

Contact

Open an issue or a pull request.