Agent Engineering Studio

Il laboratorio dove le idee diventano AI agents.

Starter kit, template, demo end-to-end e best practice — dal concept alla prima release.

Le aree dello studio

Starter Kit

Template pronti per partire: agenti, tool, orchestrazioni. Fork e inizia.

Progetti End-to-End

Demo complete dal concept al deploy. Stack reali, codice funzionante.

Best Practice

Pattern, architetture e convenzioni testate. Impara dai nostri errori.

Community Lab

Contribuisci, testa, migliora. Ogni progetto รจ aperto e documentato.

Supporta il Progetto

Se questo progetto open-source ti รจ utile, considera di offrire un caffรจ per aiutare a mantenerlo attivo e in crescita.

๐Ÿค– Help the project

Progetti

Progetti open-source per sistemi multi-agente e AI engineering

In Development

Knowledge Graph

Sistema completo con pipeline di ingestion, vector store, graph database, hybrid RAG, interfaccia web e orchestrazione multi-agent basata su LangGraph. Progetto companion del libro "Knowledge Graph: dalla Teoria alla Pratica".

Architettura

Client / LLM Host
MCP Server :8080
Agent API :8002
FastAPI :8000
Neo4j :7474
Redis :6379
Ollama :11434
Next.js UI :3000

Moduli

knowledge-graph-api FastAPI backend โ€” ingestion, RAG, graph Python ยท FastAPI ยท Pydantic v2
knowledge-graph-ui Next.js 15 web frontend Next.js 15 ยท React 19 ยท Tailwind CSS 4
knowledge-graph-mcp MCP server โ€” espone le API come tool LLM Python ยท FastMCP ยท SSE transport
knowledge-graph-agents Multi-agent orchestration (LangGraph) Python ยท LangGraph ยท StateGraph

Multi-Agent System (Supervisor + Specialists)

Orchestrator Classifica intent, crea il piano โ€” mai tool diretti
Ingestion Health check, dedup, ingest documenti
Analyst Vector search / graph traversal / hybrid (3 strategie)
Validator 4 Cypher query, KGQualityReport con overall_health
KGC Graph Completion โ€” relazioni mancanti
Synthesis RAG context + LLM, report Markdown
Monitor Health check + quality check, alert summary

Pipeline

Ingestion Document โ†’ Chunking โ†’ Embedding โ†’ Dedup โ†’ Entity Extraction โ†’ Redis + Neo4j
RAG Query Question โ†’ Intent Classification โ†’ Vector Search โ†’ Graph Traversal โ†’ LLM Generation
Multi-Agent Request โ†’ Orchestrator (LangGraph) โ†’ Specialist Agent โ†’ HTTP Tools โ†’ Response

Stack

Neo4j 5.18Redis StackOllamaFastAPILangGraphNext.js 15FastMCPDocker Compose

Modelli LLM

Llama 3Qwen 2.5:14bnomic-embed-text

Servizi Docker Compose

UI localhost:3000 Next.js frontend
API localhost:8000 FastAPI REST + Swagger
Agent API localhost:8002 Multi-Agent Orchestration
MCP Server localhost:8080 MCP tool layer (SSE)
Neo4j localhost:7474 Graph DB Browser
RedisInsight localhost:8001 Redis web UI