AnalyticDB:Using the GraphRAG service

How it works

GraphRAG operates in three stages:

1. Indexing: A knowledge extraction model parses your documents, generates a knowledge graph, and saves it to the AnalyticDB for PostgreSQL graph analysis engine.

2. Retrieval: The model extracts keywords from a query and traverses the knowledge graph to find related subgraphs.

3. Generation: The query and the retrieved subgraph context are submitted to a Large Language Model (LLM) to generate a response.

Prerequisites

Before you begin, make sure you have:

• An AnalyticDB for PostgreSQL 7.0 instance with minor engine version 7.2.1.3 or later. Versions 7.3.0.0 and 7.3.1.0 do not support the adbpg_graphrag extension.

• The required extensions installed. See Install extensions.

Note: To check your minor version, go to the Basic Information page of your instance in the AnalyticDB for PostgreSQL console. If the version is outdated, update the minor engine version.

Quick start

The following end-to-end example initializes GraphRAG, uploads a document, and runs a query.

Note: If you are using DMS, select both statements in the SQL window and click Execute to run them in the same session. The initialize and query operations must run in the same session.

-- Step 1: Initialize the GraphRAG service (see Initialize the GraphRAG service for configuration options)
SELECT adbpg_graphrag.initialize(config json);

-- Step 2: Upload a document
SELECT adbpg_graphrag.upload(
    'ProductInfo.txt',
    'The Xiao Mi customer service system provides features such as question answering, leave requests, and knowledge base searches.'
);

-- Step 3: Query the knowledge graph
SELECT adbpg_graphrag.query('What features does the Xiao Mi system have?', 'hybrid');

Install extensions

Three extensions are required, in this order: plpython3u, age, and adbpg_graphrag.

On instances with minor engine version 7.2.1.4 or later, all three are installed automatically. Skip to step 2 to configure ag_catalog in the search_path.

1. Install plpython3u (installation guide). This extension is installed by default. To verify, run the following query in the public schema of your database:

   SELECT * FROM pg_extension WHERE extname = 'plpython3u';

   The following output confirms the extension is installed. No output means it is not installed in the public schema.

    oid  |  extname   | extowner | extnamespace | extrelocatable | extversion | extconfig | extcondition
   -------+------------+----------+--------------+----------------+------------+-----------+--------------
    14674 | plpython3u |       10 |           11 | f              | 1.0        |           |
   (1 row)

2. Install age (installation guide). After installing age, add ag_catalog to the search_path to simplify queries. Choose one of the following methods:

   • Session-level (applies to the current session only):

     SET search_path TO "$user", public, ag_catalog;

   • Database-level (permanent):

     ALTER DATABASE <database_name> SET search_path TO "$user", public, ag_catalog;

   (Optional) To grant other users access to the graph extension, use the initial account or a privileged user with the RDS_SUPERUSER role:

   GRANT USAGE ON SCHEMA ag_catalog TO <username>;

3. Install adbpg_graphrag. This extension depends on both plpython3u and age. Contact technical support for installation assistance.

Initialize the GraphRAG service

SELECT adbpg_graphrag.initialize(config json);

Call adbpg_graphrag.initialize before uploading documents or running queries. It accepts a JSON object with the following parameters.

SELECT adbpg_graphrag.initialize(
	$$
	{
		"llm_model" : "qwen-max-2025-01-25",
		"llm_api_key" : "sk-****************",
		"llm_url" : "https://dashscope.aliyuncs.com/compatible-mode/v1",
		"embedding_model" : "text-embedding-v3",
		"language" : "Simplified Chinese",
		"entity_types" : ["Business Scenario", "Product", "Feature", "Usage Logic", "Data Metric", "Data Caliber"],
		"relationship_types": ["Contains", "Belongs To", "Uses", "Depends On", "Associates With", "Solves"]
	}
	$$
);

Upload documents

SELECT adbpg_graphrag.upload(filename text, context text);

After a file is uploaded, GraphRAG performs a series of tasks, such as text chunking, vector generation, and knowledge graph extraction.

Example:

SELECT adbpg_graphrag.upload(
    'ProductInfo.txt',
    'The Xiao Mi customer service system provides features such as question answering, leave requests, and knowledge base searches.'
);

You are a professional image reading assistant. Your task is to generate a detailed interpretation and a comprehensive summary for the provided image.
---Requirements---
1. If the image contains distinct modules, extract and provide as much detail as possible for each module.
2. Based on the extracted information, generate a summary for each module.
3. If the image contains charts, summarize the available details from the charts.
4. If the image contains meaningful icons, such as flag icons, pictograms, metaphorical icons, tool icons, and hybrid icons, also extract details from these icons.
5. Based on the extracted information, create a summary and provide useful global information.
---Note---
Write in the third person and include module names to ensure complete context.
*Identify the primary language of the image* (the language that makes up more than 80% of the text). Use the *primary language of the image* as the output language.

Query for information

SELECT adbpg_graphrag.query(query_str text);
SELECT adbpg_graphrag.query(query_str text, query_mode text);

Example:

SELECT adbpg_graphrag.query('What features does the Xiao Mi system have?', 'hybrid');

What's next

• Tutorial: Build a GraphRAG with DTS RAGFlow and AnalyticDB for PostgreSQL

• View the minor engine version

• Multimodal large language models