ApsaraDB RDS:Build an AI application based on ApsaraDB RDS for PostgreSQL and the Dify platform

Create an ECS instance

Create an ECS instance. For more information, see Custom purchase of ECS instances.

Install Docker

Install Docker on the ECS instance. For more information, see Install Docker.

Get the Dify source code

Clone the Dify repository:

git clone https://github.com/langgenius/dify.git

To install a specific version, use the --branch parameter. For example, to install version v1.0.0:

git clone https://github.com/langgenius/dify.git --branch 1.0.0

For available versions, see Dify tags.

Configure ApsaraDB RDS for PostgreSQL as the database and vector store

Set ApsaraDB RDS for PostgreSQL as both the default database and the default vector store for Dify. Choose one of the following configuration methods.

export DB_USERNAME=testdbuser
export DB_PASSWORD=dbPassword
export DB_HOST=pgm-****.pg.rds.aliyuncs.com
export DB_PORT=5432
export DB_DATABASE=testdb01

export VECTOR_STORE=pgvector
export PGVECTOR_HOST=pgm-****.pg.rds.aliyuncs.com
export PGVECTOR_PORT=5432
export PGVECTOR_USER=testdbuser
export PGVECTOR_PASSWORD=dbPassword
export PGVECTOR_DATABASE=testdb01

(Optional) Disable the default database and Weaviate containers

To avoid running the default PostgreSQL database and Weaviate containers on ECS, which saves traffic and storage space, edit the .env and docker-compose.yaml files.

Start Dify

Start Dify:

cd /root/dify/docker
docker compose -f docker-compose.yaml up -d

Add Tongyi Qianwen through an API key

1. Log on to the Dify platform.

2. In the upper-right corner, click User Name > Settings.

3. On the Settings page, go to Model Provider > Tongyi Qianwen > (settings).

4. On the Tongyi Qianwen Settings page, click the link to retrieve the API key from Alibaba Cloud Bailian.

5. Enter the API key and click Save.

Upload documents

1. Go to Knowledge Base > Create Knowledge Base > Import Existing Text > Select File > Next to upload your prepared file.

1. Click Next and follow the on-screen guide to perform Text Segmentation And Traffic Scrubbing. You can use the default settings. Dify automatically scrubs, segments, and indexes the uploaded documents, improving the Q&A application's retrieval accuracy.

Verify the knowledge base in ApsaraDB RDS for PostgreSQL

After setting up the knowledge base, verify its content in the ApsaraDB RDS for PostgreSQL database and confirm the index for each knowledge base table.

1. Connect to the ApsaraDB RDS for PostgreSQL database used by Dify. For connection instructions, see Connect to an ApsaraDB RDS for PostgreSQL instance.

2. Query the knowledge base IDs:

      SELECT * FROM datasets;

3. Construct the table name from the knowledge base ID: replace - with _ in the ID, add the prefix embedding_vector_index_, and append the suffix _nod. For example:

      SELECT * FROM embedding_vector_index_6b169753_****_****_be66_9bddc44a4848_nod;

4. Confirm the knowledge base index. By default, Dify creates an HNSW index for each knowledge base table to speed up vector similarity queries through the pgvector plug-in. The system uses the following SQL statement for vector similarity queries: To check whether the index exists and its parameters meet your recall rate requirements, run: For details on how the HNSW index parameters m and ef_construction affect recall rate, see pgvector performance testing (based on HNSW index).

      SELECT
          meta,
          text,
          embedding <=> $1 AS distance
      FROM
          embedding_vector_index_6b169753_****_****_be66_9bddc44a4848_nod
      ORDER BY
          distance
      LIMIT
          $2;

      SELECT *
      FROM pg_indexes
      WHERE tablename = 'embedding_vector_index_6b169753_****_****_be66_9bddc44a4848_nod';

5. (Optional) If the index was not created automatically or the default parameters do not meet your recall rate requirements, create the index manually.

   1. Delete the existing index (if needed):

      DROP INDEX IF EXISTS embedding_vector_index_6b169753_****_****_be66_9bddc44a4848_nod;

   2. Create a new HNSW index:

      CREATE INDEX ON embedding_vector_index_6b169753_****_****_be66_9bddc44a4848_nod
      USING hnsw (embedding vector_cosine_ops)
      WITH (m = 16, ef_construction = 100);

   Note

   Replace embedding_vector_index_6b169753_****_****_be66_9bddc44a4848_nod with your actual table name.