Rate This Document
Findability
Accuracy
Completeness
Readability

Introduction

This document describes how to develop langchain-core on Kunpeng servers, including source code build, installation, and development verification.

LangChain uses modular abstraction and LangChain Expression Language (LCEL) to encapsulate model calling, data retrieval, tool execution, and memory management into standardized, composable components. This provides developers with a full-process standardized solution from prototype verification to production deployment. When LangChain is deployed on a Kunpeng server, the BiSheng compiler, KML math library, and Hyper MPI high-performance communication library in HPCKit can be used to perform full-link collaborative optimization on computational graphs and operators.

The verification objective is to ensure that LangChain can complete basic installation, import, and minimum running links on the Arm platform. Complex scenarios such as vector databases, RAG, Agent, external model services, and third-party tool invoking are not covered.

Verification Scope

Table 1 Component scope

Component

Description

Mandatory for Verification

LangChain

Main package of LangChain. Verify that the main entry can be installed and imported.

Mandatory

langchain-core

Core abstraction and runtime. Verify basic capabilities such as Runnable and Prompt.

Mandatory

langchain-text-splitters

Text segmentation component. Verify basic text processing capabilities.

Recommended

Table 2 Components not included in the verification scope

Component/Scenario

Reason

Vector database

Native dependencies and server deployment will be introduced, which is beyond the scope of basic compatibility verification.

RAG

It depends on external components such as the vector library, Embedding, and document parsing.

Agent

It depends on model services and tool calling, and involves many variables.

langchain-community

It involves a large number of third-party integrations and is suitable for extended verification, not the minimum verification.

External LLM API

It depends on the stability of the network, keys, and model services, and is not suitable as a prerequisite for basic verification.