Spring AI vectorstore pgvector support specify datasource host

pgvector support specify url\host such as vectorstore: chroma: initialize-schema: true client: host: ${CHROMA_HOST:http://192.168.18.64} port: ${CHROMA_PORT:8000}

In my scenario, MySQL is used for business-related and pgvector is used for vectors, but PgVectorStoreAutoConfiguration directly use JdbcTemplate, hoping to be configured separately like other Chroma instead of' spring.datasource'

Comment From: viosay

The same issue, I tried exclude PgVectorStoreAutoConfiguration and configuring support for multiple data sources, but it doesn't seem to be working.

Comment From: viosay

I have already implemented this requirement in the business using multi-datasource switching, and you can try it as well.

Comment From: markpollack

@viosay can you share your implementation?

Comment From: viosay

@markpollack I excluded the automatic injection of PgVectorStoreAutoConfiguration and manually constructed the jdbcTemplate using PgVectorStore.builder. The jdbcTemplate is configured with a data connection pool specified by annotations from certain multi-datasource plugins, and it works good. Of course, if using automatic injection, it might be a bit tricky at the source code level.

Comment From: markpollack

Thanks. Is this something that should go into our documentation?