how does logging work?

[roncking]

Hi, I have included langchain4j in my maven based project. It runs fine, but I can't see any logging output at all! Here's my code:

    ChatLanguageModel chatLanguageModel = OpenAiChatModel.builder()
            .apiKey(ApiKeys.OPENAI_API_KEY)
            .modelName("gpt-4-0613") //gpt-4-0163
            .maxTokens(7000)
            .topP(1.0)
            .temperature(1.0)
            .timeout(ofSeconds(120))
            .logRequests(true)
            .logResponses(true)
            .build();

What do I need to do to activate logging, and where will the output be?

[langchain4j]

@roncking sorry for the late response.

You need to include one of SLF4J implementations (e.g. log4j, logback, tinylog etc) in your classpath and configure it to print out logs to stdout/file/etc.

For example:

            <dependency>
                <groupId>org.tinylog</groupId>
                <artifactId>slf4j-tinylog</artifactId>
                <version>2.7.0</version>
            </dependency>

            <dependency>
                <groupId>org.tinylog</groupId>
                <artifactId>tinylog-impl</artifactId>
                <version>2.7.0</version>
            </dependency>

And also enable logging when creating the *Model instance:

.logRequests(true)
.logResponses(true)

[benzourry]

Hi @langchain4j
Do we need to do the same for spring boot project. Supposed that I'm using spring boot 3 without adding any logback or log4j dependency, and doesn't configure any logging configuration in application.properties.

[benzourry]

Or maybe what I want to ask is actually can we have access to the actual text that is being sent to language model? Say if I want to store it in the db for later improvement.

Regardless, thank you very much for your amazing work. 😊

[langchain4j]

@benzourry logging of actual requests and responses to/from LLM should be enabled explicitly.
If you create the model class manually, use:

.logRequests(true)
.logResponses(true)

If you use spring boot starter, use:

langchain4j.open-ai.chat-model.log-requests=true
langchain4j.open-ai.chat-model.log-responses=true
logging.level.dev.langchain4j=DEBUG
logging.level.dev.ai4j.openai4j=DEBUG // when using langchain4j-open-ai or langchain4j-local-ai module

There is no way to get access to request/response programmatically yet, but we are going to add this feature quite soon
(hopefully in the next few weeks).
In the meantime, you could probably intercept calls to the slf4j implementation with your custom callback.

[benzourry]

I'm using normal java version of langchain4j with spring boot (without using langchain4j spring boot starter) with default spring boot logging.

As mention by @langchain4j, i set

.logRequests(true)
.logResponses(true)

and in my application.properties, I just add

logging.level.dev.langchain4j=DEBUG

[langchain4j]

@benzourry yes, when using Spring Boot, you don't need to include any extra loggers, as spring-boot-starter-logging already has everything needed.