In recent years, the advancements in AI, particularly in Large Language Models (LLMs) such as GPT-3, have revolutionized various sectors by providing nearly human-like text generation and interaction capabilities. However, when combined with Retrieval-Augmented Generation (RAG) and Knowledge Graphs, these models’ efficiency and accuracy reach unprecedented levels, offering more contextually relevant and precise information.<\/p>\n\n\n\n
In this article, we will introduce the basics of RAG and its limitation, and introduce a better RAG solution too.<\/p>\n\n\n\n
LLMs, such as GPT (Generative Pre-trained Transformer) series, are deep learning algorithms trained on vast amounts of text data. They can predict the next word in a sentence, making them powerful tools for generating human-like text based on the input they’re fed.<\/p>\n\n\n\n
Retrieval-Augmented Generation (RAG) combines the generative power of LLMs with an external knowledge source, allowing the model to “retrieve” relevant information before “generating” a response. This process enhances the model’s ability to provide informed and precise answers.<\/p>\n\n\n\n
Knowledge Graphs organize and store complex data in an interconnected, graph-like structure. They enable logical reasoning over the stored data, facilitating precise information retrieval and data interconnectivity essential for various AI applications.<\/p>\n\n\n\n
While LLMs have shown remarkable capabilities, their effectiveness is often limited by the data on which they were trained. They can generate outdated or irrelevant information and struggle with queries requiring up-to-date or domain-specific knowledge.<\/p>\n\n\n\n
While RAG models enhance LLMs by integrating external sources, they still face challenges such as integrating these sources smoothly and efficiently extracting the most relevant information from vast databases.<\/p>\n\n\n\n
Combining LLMs with Knowledge Graphs bridges the gap between generative models’ linguistic capabilities and the structured, detailed data stored in knowledge bases.<\/p>\n\n\n\n
Go through the code of demo: pingcap\/tidb-vector-python\/examples\/graphrag-demo<\/a>, we can find the code how we design schema SQL to store knowledge graph:<\/p>\n\n\n\n Here, we use TiDB Cloud \uc2a4\ud0c0\ud130<\/a> – MySQL compatible database with built-in vector storage – to store the entity and relationship data:<\/p>\n\n\n\n Sure, it’s a MySQL compatible table schema definition, and the data format would be like:<\/p>\n\n\n\n The full demo data is \uc5ec\uae30<\/a>.<\/p>\n\n\n\n After create table and inserting demo data, we can perform a search on top of them with just MySQL skills but with vector search!<\/p>\n\n\n\n It’s clear that we have implemented a RAG but:<\/p>\n\n\n\n That’s amazing!<\/p>\n\n\n\n (The full demo code is \uc5ec\uae30<\/a>.)<\/p>\n\n\n\n The integration allows LLMs to leverage structured data from Knowledge Graphs, enriching the context and accuracy of the generated responses. This symbiosis enables AI to not just predict text but to understand and generate content with nuanced, fact-based insights.<\/p>\n\n\n\n This combination supports complex query answering, factual consistency, and the exploration of relationships between entities, significantly improving task-oriented and informational dialogues in AI applications.<\/p>\n\n\n\n GraphRAG revolutionizes the traditional RAG approach by directly integrating Knowledge Graph data into the indexing and retrieval process, leading to more relevant data extraction and smarter, contextual response generation.<\/p>\n\n\n\n Try TiDB Serverless Vector Storage<\/p><\/div>\n Build AI applications confidently with the most advanced MySQL vector solution.<\/p><\/div>\n Prepare Tables<\/h4>\n\n\n\n
CREATE TABLE `entities` (\n `id` int(11) NOT NULL AUTO_INCREMENT,\n `name` varchar(512) DEFAULT NULL,\n `description` text DEFAULT NULL,\n `description_vec` vector<float> DEFAULT NULL,\n PRIMARY KEY (`id`) \/*T![clustered_index] CLUSTERED *\/\n) ENGINE=InnoDB DEFAULT CHARSET=utf8mb4 COLLATE=utf8mb4_bin;\n\nCREATE TABLE `relationships` (\n `id` int(11) NOT NULL AUTO_INCREMENT,\n `source_entity_id` int(11) DEFAULT NULL,\n `target_entity_id` int(11) DEFAULT NULL,\n `relationship_desc` text DEFAULT NULL,\n PRIMARY KEY (`id`) \/*T![clustered_index] CLUSTERED *\/,\n KEY `fk_1` (`source_entity_id`),\n KEY `fk_2` (`target_entity_id`),\n CONSTRAINT `fk_1` FOREIGN KEY (`source_entity_id`) REFERENCES `entities` (`id`),\n CONSTRAINT `fk_2` FOREIGN KEY (`target_entity_id`) REFERENCES `entities` (`id`)\n) ENGINE=InnoDB DEFAULT CHARSET=utf8mb4 COLLATE=utf8mb4_bin;<\/code><\/pre>\n\n\n\nInsert Entity & Relationship<\/h4>\n\n\n\n
INSERT INTO entities (id, name, description, description_vec) VALUES (1, 'Elon Reeve Musk', 'Elon Reeve Musk is a businessman and investor, born on June 28, 1971, in Pretoria, South Africa. He is the founder, chairman, CEO, and CTO of SpaceX; angel investor, CEO, product architect, and former chairman of Tesla, Inc.; owner, executive chairman, and CTO of X Corp.; founder of the Boring Company and xAI; co-founder of Neuralink and OpenAI; and president of the Musk Foundation.', [1536 length of vector]);<\/code><\/pre>\n\n\n\nSearch knowledge with SQL<\/h4>\n\n\n\n
from sqlalchemy import create_engine, text\nimport openai\nimport getpass\n\n# TiDB Connection String Pattern:\n# mysql+pymysql:\/\/{TIDB_USER}:{TIDB_PASSWORD}@{TIDB_HOST}:{TIDB_PORT}\/{TIDB_DB_NAME}?ssl_verify_cert=True&ssl_verify_identity=True\n\ndb_engine = create_engine(getpass.getpass(\"Input your TIDB connection string:\"))\noai_cli = openai.OpenAI(api_key=getpass.getpass(\"Input your OpenAI API Key:\"))\nquestion = input(\"Enter your question:\")\nembedding = str(oai_cli.embeddings.create(input=[question], model=\"text-embedding-3-small\").data[0].embedding)\n\nwith db_engine.connect() as conn:\n result = conn.execute(text(\"\"\"\n WITH initial_entity AS (\n SELECT id FROM `entities`\n ORDER BY VEC_Cosine_Distance(description_vec, :embedding) LIMIT 1\n ), entities_ids AS (\n SELECT source_entity_id i FROM relationships r INNER JOIN initial_entity i ON r.target_entity_id = i.id\n UNION SELECT target_entity_id i FROM relationships r INNER JOIN initial_entity i ON r.source_entity_id = i.id\n UNION SELECT initial_entity.id i FROM initial_entity\n ) SELECT description FROM `entities` WHERE id IN (SELECT i FROM entities_ids);\"\"\"), {\"embedding\": embedding}).fetchall()\n\n print(oai_cli.chat.completions.create(model=\"gpt-4o\", messages=[\n {\"role\": \"system\", \"content\": f\"Please carefully answer the question by {str(result)}\"},\n {\"role\": \"user\", \"content\": question}]).choices[0].message.content)<\/code><\/pre>\n\n\n\n\n
How LLMs and Knowledge Graphs complement each other<\/h3>\n\n\n\n
Benefits of combining LLMs\u2019 natural language understanding with Knowledge Graphs\u2019 structured data<\/h3>\n\n\n\n
<\/span>Workflow Design of GraphRAG<\/strong><\/span><\/h2>\n\n\n\n
Explain the workflow to use graphrag<\/h3>\n\n\n\n
\n
\n