manufacturing-graphrag
Платформа интеллектуальной прослеживаемости производства
Агентная ИИ-платформа на основе GraphRAG для прослеживаемости производства с использованием Neo4j, Amazon Bedrock и MCP.
Архитектура
┌─────────────────────────────────────────────────────────────────────┐
│ INGESTION & ENRICHMENT LAYER │
│ Connectors → Chunking → Embeddings → Entity Extraction → │
│ Entity Resolution → Graph Construction │
└──────────────────────────┬──────────────────────────────────────────┘
│
┌──────────────────────────▼──────────────────────────────────────────┐
│ KNOWLEDGE GRAPH (Neo4j) │
│ Product → Requirement → Component → TestCase → TestRun → │
│ Defect → ChangeRequest │
└──────────────────────────┬──────────────────────────────────────────┘
│
┌──────────────────────────▼──────────────────────────────────────────┐
│ RETRIEVAL LAYER │
│ VectorRetriever │ GraphRAGRetriever │ NL2CypherRetriever │ Hybrid │
└──────────────────────────┬──────────────────────────────────────────┘
│
┌──────────────────────────▼──────────────────────────────────────────┐
│ AGENT LAYER │
│ LangGraph Agent (plan→retrieve→reason→validate→respond) │
│ Strands Agent (tool-calling with 5 specialized graph tools) │
└──────────────────────────┬──────────────────────────────────────────┘
│
┌──────────────────────────▼──────────────────────────────────────────┐
│ INTERFACE LAYER │
│ FastAPI REST │ MCP Server (Claude Desktop / Cursor compatible) │
└─────────────────────────────────────────────────────────────────────┘Related MCP server: RAG MCP Server
Быстрый старт
1. Предварительные требования
Neo4j 5.x (с плагином APOC)
Python 3.11+
Учетные данные AWS с доступом к Bedrock (Claude 3.5 Sonnet + Titan Embed v2)
2. Настройка
cd manufacturing-graphrag
python -m venv .venv
.venv\Scripts\activate # Windows
pip install -r requirements.txt
python -m spacy download en_core_web_sm
copy .env.example .env # Edit with your credentials3. Запуск Neo4j (Docker)
docker-compose up neo4j -d4. Заполнение графа знаний
set PYTHONPATH=src
python scripts/seed_data.py5. Запуск платформы
# API server
python main.py api
# MCP server (for Claude Desktop)
python main.py mcp
# Both
python main.py all6. Загрузка документов
python scripts/ingest.py path/to/spec.pdf path/to/requirements.csvAPI-эндпоинты
Метод | Endpoint | Описание |
GET |
| Проверка работоспособности |
GET |
| Количество узлов по меткам |
POST |
| Загрузка и импорт документа |
POST |
| Импорт структурированной записи API |
POST |
| Гибридный GraphRAG-запрос |
POST |
| Агент рассуждений на LangGraph |
POST |
| Агент вызова инструментов Strands |
GET |
| Полная цепочка прослеживаемости дефектов |
GET |
| Сводка по прослеживаемости продукта |
Примеры запросов
# Hybrid GraphRAG query
curl -X POST http://localhost:8000/query \
-H "Content-Type: application/json" \
-d '{"question": "What components are affected by the thermal runaway defect?"}'
# LangGraph agent — multi-step reasoning
curl -X POST http://localhost:8000/agent/langgraph \
-d '{"question": "Trace the full impact chain of DEF-001 and identify all change requests needed"}'
# Strands agent — tool-calling
curl -X POST http://localhost:8000/agent/strands \
-d '{"question": "Which critical defects are blocking the EV BMS release?"}'
# Defect traceability
curl http://localhost:8000/traceability/defect/DEF-002Интеграция с MCP (Claude Desktop)
Скопируйте содержимое config/claude_desktop_mcp.json в ваш Claude Desktop claude_desktop_config.json.
Доступные MCP-инструменты:
ask_manufacturing_ai— гибридный GraphRAG Q&Asemantic_search— векторный поиск по сходствуgraph_trace— извлечение с обходом графаnatural_language_to_cypher— NL2Cypherget_defect_chain— полная прослеживаемость дефектовget_requirement_traceability— покрытие требованийproduct_health_dashboard— метрики продукта
Схема графа знаний
(Product)-[:HAS_REQUIREMENT]->(Requirement)
(Component)-[:IMPLEMENTS]->(Requirement)
(TestCase)-[:VALIDATES]->(Requirement)
(TestRun)-[:INSTANCE_OF]->(TestCase)
(TestRun)-[:FOUND_IN]->(Defect)
(Defect)-[:AFFECTS]->(Component)
(Defect)-[:TRIGGERS_CHANGE]->(ChangeRequest)
(ChangeRequest)-[:MODIFIES]->(Component)
(Document)-[:CONTAINS_CHUNK]->(Chunk)
(Chunk)-[:MENTIONS]->(any entity)Запуск тестов
set PYTHONPATH=src
pytest tests/ -vThis server cannot be deployed
Maintenance
Related MCP Connectors
Answer questions using your semantic layer and company knowledge graph.
Knowledge graph ingestion, entity search, ontology analysis, and CoSync scoring.
Shared semantic graph for AI reviews, classification and structured memory across AI assistants.
AI knowledge graph for architecture, portfolio, and digital strategy management.
Related MCP Servers
- FlicenseNot gradedqualityDmaintenanceEnables storage and retrieval of knowledge in a graph database format, allowing users to create, update, search, and delete entities and relationships in a Neo4j-powered knowledge graph through natural language.5-
- AlicenseCqualityDmaintenanceCombines a knowledge graph with RAG (Retrieval-Augmented Generation) capabilities for semantic code indexing and search. Enables creating entity relationships, managing observations, and performing semantic searches across indexed codebases.13MIT
- FlicenseNot gradedqualityCmaintenanceTransforms code repositories and development documentation into a queryable Neo4j knowledge graph, enabling AI assistants to perform intelligent code analysis, dependency mapping, impact assessment, and automated documentation generation across 15+ programming languages.7-
- FlicenseNot gradedqualityDmaintenanceMCP server for manufacturing defect investigation using knowledge graphs and graph analytics. Enables LLM agents to retrieve context, find similar cases, and generate investigation reports from manufacturing data.-