Klea Utils¶
klea_utils is the shared foundation that every other Klea package is
built on. It is published to PyPI, so its framework pieces can be reused
independently of the applications.
BaseLangGraph¶
At its core is BaseLangGraph, a template
method for building LangGraph orchestrators. setup() runs a fixed
sequence – load configuration, set up models, configure resources,
connect the MCP client, load stores, compile the graph – and subclasses
implement only the seams they need:
_setup_models– declare the model roles;_configure_resources– declare MCP servers and knowledge stores;_create_graph– define the nodes, edges and routing.
Execution (invoke, stream, structured event stream), session checkpointing and per-request model overrides are provided by the base, so an application gets them for free. Both Klea RAG and Klea Agent are subclasses.
Interfaces¶
The package also ships the klea-stores-create CLI for building vector
and BM25 stores, and the shared web UI components used by both
applications (see Web interface).
See also
API Reference – the Python API reference
Developer info – architecture and the C4 model
Contributing – development setup