LangChain Executor

LangChain-based execution engine for Ark agents with optional Retrieval-Augmented Generation (RAG) support. Communicates using the A2A (Agent-to-Agent) protocol.

Overview

The LangChain Executor provides:

LangChain Framework Integration - Execute agents using LangChain with OpenAI and Azure OpenAI models
Conversation History - Server-side conversation memory keyed by conversationId, using LangChain’s ChatMessageHistory
RAG Support - Retrieval-Augmented Generation with FAISS vector search for code-aware responses
A2A Protocol - Compliant with the Agent-to-Agent protocol for seamless integration with Ark
ExecutionEngine CRD - Registers as a Kubernetes-native execution engine

Install


ark install marketplace/executors/executor-langchain

Or with DevSpace:


cd executors/langchain
devspace deploy

Or with Helm:


helm install executor-langchain ./chart -n default --create-namespace

Access via:

Port forward: kubectl port-forward svc/executor-langchain 8000:8000
Health check: curl http://localhost:8000/health

ExecutionEngine Resource

The LangChain executor registers as an ExecutionEngine with A2A protocol:


apiVersion: ark.mckinsey.com/v1prealpha1
kind: ExecutionEngine
metadata:
  name: executor-langchain
spec:
  address:
    valueFrom:
      serviceRef:
        name: executor-langchain
  description: "LangChain Executor - A2A server with RAG support"

Creating Agents

Reference the execution engine in your agent spec:


apiVersion: ark.mckinsey.com/v1alpha1
kind: Agent
metadata:
  name: my-langchain-agent
spec:
  executionEngine:
    name: executor-langchain
  modelRef:
    name: my-model
  prompt: |
    You are a helpful assistant.

RAG Support

Enable RAG for an agent by adding the langchain: rag label:


apiVersion: ark.mckinsey.com/v1alpha1
kind: Agent
metadata:
  name: my-rag-agent
  labels:
    langchain: rag
spec:
  executionEngine:
    name: executor-langchain
  modelRef:
    name: my-model
  prompt: |
    You are an expert developer assistant with deep knowledge of the codebase.
    When RAG context is provided, use it to give accurate, specific answers.

When RAG is enabled, the engine indexes local Python files and provides relevant code context via FAISS vector search.

Custom Embeddings Model

Specify a custom embeddings model via the langchain-embeddings-model label:


metadata:
  labels:
    langchain: rag
    langchain-embeddings-model: text-embedding-3-small

Configuration

Parameter	Description	Default
`app.image.repository`	Container image	`ghcr.io/mckinsey/agents-at-scale-marketplace/executor-langchain`
`service.port`	Service port	`8000`
`executionEngine.enabled`	Register ExecutionEngine CRD	`true`
`httpRoute.enabled`	Enable Gateway API HTTPRoute	`false`

Environment Variables

Variable	Description	Default
`HOST`	Server bind address	`0.0.0.0`
`PORT`	Server port	`8000`

Supported Model Types

Type	Description
`openai`	OpenAI API with configurable properties
`azure`	Azure OpenAI with deployment URLs

Troubleshooting

Service Not Starting

Check pod status and logs:


kubectl get pods -l app=executor-langchain
kubectl logs -l app=executor-langchain

ExecutionEngine Not Registered

Verify the ExecutionEngine CRD exists:


kubectl get executionengines

Health Check Failing

Test the health endpoint directly:


kubectl port-forward svc/executor-langchain 8000:8000
curl http://localhost:8000/health

Uninstall

Using DevSpace:


cd executors/langchain
devspace purge

Using Helm:


helm uninstall executor-langchain -n default