Agentic Flows Developer Guide

Overview

Package: @bluefly/agentic-flows Version: Latest License: MIT

Multi-Agent Workflow Orchestration with N8N custom nodes for CrewAI, Langflow integration, MLflow experiment tracking, and vector operations.

Key Features

Agent Orchestration: CrewAI integration, multi-agent workflows, role-based agents
Langflow Integration: N8N nodes for Langflow, visual flow builder, component deployment
MLflow Experiment Tracking: Experiment logging, model registry, metrics tracking
Vector Operations: Qdrant, Pinecone, Weaviate, Milvus support
LLM Integration: Multi-provider support (OpenAI, Anthropic, Ollama, Bedrock)
Observability: Phoenix Arize, MLflow tracking, Prometheus metrics

Installation

For N8N Users

cd ~/.n8n/custom
npm install @bluefly/agentic-flows
n8n start

For Developers

npm install @bluefly/agentic-flows

Quick Start

N8N Node Examples

CrewAI Agent Node

{
  "node": "CrewAI Agent",
  "parameters": {
    "operation": "executeAgent",
    "agentRole": "researcher",
    "goal": "Research market trends",
    "backstory": "Expert market analyst",
    "task": "Analyze Q4 2024 AI market trends",
    "tools": ["search", "calculator"]
  }
}

Vector Search Node

{
  "node": "Vector Search",
  "parameters": {
    "operation": "search",
    "database": "qdrant",
    "collection": "knowledge-base",
    "query": "How to implement RAG?",
    "topK": 5,
    "scoreThreshold": 0.7
  }
}

Programmatic API

import { AgenticFlows } from '@bluefly/agentic-flows';

const flows = new AgenticFlows({
  n8nUrl: 'http://localhost:5678',
  crewAiUrl: 'http://localhost:8000'
});

const workflow = await flows.createWorkflow({
  name: 'Research and Analysis',
  agents: [
    {
      role: 'researcher',
      goal: 'Gather information',
      tools: ['web_search', 'arxiv']
    },
    {
      role: 'analyst',
      goal: 'Analyze findings',
      tools: ['data_analysis']
    }
  ],
  flow: 'sequential'
});

const result = await workflow.execute({
  input: 'Analyze AI impact on healthcare'
});

Available N8N Nodes

Node	Purpose	Operations
CrewAI Agent	Agent orchestration	Create, execute, coordinate
Vector Search	Semantic search	Index, search, update
LLM Gateway	LLM operations	Generate, embed, moderate
RAG Pipeline	Retrieval augmentation	Retrieve, generate, validate
Agent Memory	State management	Store, retrieve, clear
Tool Executor	Tool invocation	Execute, validate, retry

Configuration

Environment Variables

# LLM Configuration
OPENAI_API_KEY=sk-...
ANTHROPIC_API_KEY=sk-ant-...

# Vector Databases
QDRANT_URL=http://localhost:6333
PINECONE_API_KEY=...

# CrewAI
CREWAI_API_URL=http://localhost:8000

Langflow + MLflow Integration

Execute Langflow from N8N

const langflowNode = {
  node: "Execute Langflow",
  parameters: {
    workflowId: "my-ai-workflow",
    inputs: { query: "Analyze market data" },
    trackWithMLflow: true
  }
}

MLflow Experiment Tracking

import { MLflowTracker } from '@bluefly/agentic-flows';

const tracker = new MLflowTracker({
  trackingUri: 'http://localhost:5000'
});

await tracker.startRun({
  experimentName: 'CrewAI Research Team',
  runName: 'market-analysis-2024'
});

await tracker.logMetrics({
  accuracy: 0.95,
  cost: 0.42,
  duration_seconds: 12.5
});

await tracker.endRun();

Testing

npm test
npm run test:coverage
npm run lint

Deployment

Kubernetes

kubectl create namespace agentic-flows
kubectl apply -f infrastructure/kubernetes/ -n agentic-flows

Documentation

GitLab: https://gitlab.bluefly.io/llm/npm/agentic-flows
OpenAPI Specs: openapi/