Experimental: RAG: SQL database content retriever

[langchain4j]

Issue

#232

Change

An experimental SqlDatabaseContentRetriever has been added.

Simplest usage example:

ContentRetriever contentRetriever = SqlDatabaseContentRetriever.builder()
    .dataSource(dataSource)
    .chatLanguageModel(openAiChatModel)
    .build();

In this case SQL dialect and table structure will be determined from the DataSource.

But it can be customized:

ContentRetriever contentRetriever = SqlDatabaseContentRetriever.builder()
    .dataSource(dataSource)
    .sqlDialect("PostgreSQL")
    .databaseStructure(...)
    .promptTemplate(...)
    .chatLanguageModel(openAiChatModel)
    .maxRetries(2)
    .build();

See SqlDatabaseContentRetrieverIT for a full example.

General checklist

• There are no breaking changes
• I have added unit and integration tests for my change
• I have manually run all the unit and integration tests in the module I have added/changed, and they are all green
• I have manually run all the unit and integration tests in the core and main modules, and they are all green

• I have added/updated the documentation
• I have added an example in the examples repo (only for "big" features)

Checklist for adding new model integration

• I have added my new module in the BOM

[langchain4j]

cc @dandreadis @geoand

[geoand]

Also cc @jmartisk

[jmartisk]

Looks cool, we have something like it in the Quarkus samples, but this looks more general
Did you try with various models? I remember trying to use codellama and mistral for the sample that we have, but it just didn't work properly (only GPT produced reasonable queries).

[langchain4j]

@jmartisk yes, I played with gpt-3.5-turbo, open-mistral-7b (via Mistral API) and tinydolphin (via Ollama).
For those integration tests that I have, gpt-3.5-turbo works most of the time, open-mistral-7b in like 70% cases and tinydolphin only with simple cases. I can imagine it will work much better with https://ollama.com/library/sqlcoder or other text-to-SQL models.