Workflow Engine Developer Guide

Overview

Package: @bluefly/workflow-engine Alias: @bluefly/langflow-drupal-bridge Version: Latest License: MIT

Enterprise AI workflow orchestration platform bridging Langflow with Drupal CMS for visual AI workflow management.

Key Features

• Langflow Integration: Visual workflow builder, custom components, real-time sync
• Drupal CMS Integration: Native module, entity management, visual builder UI
• GraphQL Server: Complete workflow management API, real-time subscriptions
• MCP Integration: Langflow MCP server, tool registration, Claude Desktop integration
• OSSA 1.0 Compliance: Agent protocol, workflow orchestration, capability discovery
• Observability: OpenTelemetry (server + browser), Phoenix Arize, Prometheus

Installation

# 1. Install workflow engine
npm install -g @bluefly/workflow-engine

# 2. Install Drupal module
composer require drupal/ai_agentic_workflows
drush en ai_agentic_workflows visual_builder -y

# 3. Setup Langflow
pip install langflow && langflow run

# 4. Configure
workflow-engine setup --drupal-url https://your-site.com

Quick Start

1. Visit /admin/config/ai/visual-builder in Drupal
2. Click "Test Connection" to Langflow
3. Click "Register Drupal" to add CMS components
4. Start building visual AI workflows!

Configuration

Environment Variables

# Langflow
LANGFLOW_URL=http://localhost:7860
LANGFLOW_API_KEY=your-api-key

# Drupal
DRUPAL_URL=https://your-site.com
DRUPAL_API_KEY=your-drupal-key

# GraphQL Server
GRAPHQL_PORT=4000

# MCP Server
MCP_SERVER_PORT=3001
MCP_ENABLED=true

# OpenTelemetry
OTEL_EXPORTER_OTLP_TRACES_ENDPOINT=http://localhost:4318/v1/traces

Development

npm run graphql:serve     # GraphQL server
npm run mcp:server        # MCP server
npm run components:deploy # Deploy Langflow components
npm run dev               # Development mode
npm test                  # Tests

GraphQL API

npm run graphql:serve
# GraphiQL UI: http://localhost:4000/graphql

Example Queries

query ListWorkflows {
  workflows {
    id
    name
    status
    components {
      type
      config
    }
  }
}

mutation CreateWorkflow($input: WorkflowInput!) {
  createWorkflow(input: $input) {
    id
    name
    status
  }
}

MCP Integration

Add to Claude Desktop's MCP settings:

{
  "mcpServers": {
    "langflow": {
      "command": "node",
      "args": ["/path/to/workflow-engine/dist/mcp/langflow-mcp-server.js"],
      "env": {
        "LANGFLOW_URL": "http://localhost:7860"
      }
    }
  }
}

Available Tools

• execute_workflow - Execute Langflow workflow
• list_workflows - List available workflows
• get_workflow - Get workflow details
• deploy_component - Deploy custom component

Observability

OpenTelemetry

Server and browser-side tracing enabled.

curl -I http://localhost:4318/v1/traces
open http://localhost:16686  # Jaeger
open http://localhost:6006   # Phoenix

Deployment

Kubernetes

kubectl apply -f infrastructure/kubernetes/
kubectl get pods -n workflow-engine

Docker

npm run docker:build
npm run docker:run

Documentation

• GitLab: https://gitlab.bluefly.io/llm/npm/workflow-engine
• OpenAPI Specs: openapi/