quarkiverse/quarkus-langchain4j

vertex-ai-gemini bad configuration error message

Closed this issue · 8 comments

exaV commented

Hi, I was eagerly awaiting the release of quarkus-langchain4j-vertex-ai-gemini, so props for the release 👍 .

Unfortunately I get a configuration error. I added a log excerpt below. The key "quarkus.langchain4j.vertexai.devopi.location" is logged as missing during startup and later also mentioned in the error message as required. I later realized that the config key had to be quarkus.langchain4j.vertexai.gemini.devopi.location but the error message did not help with that.

2024-05-09 10:21:59,013 WARN  [io.qua.config] (main) Unrecognized configuration key "quarkus.langchain4j.vertexai.devopi.location" was provided; it will be ignored; verify that the dependency extension for this configuration is set or that you did not make a typo
2024-05-09 10:21:59,016 WARN  [io.qua.config] (main) Unrecognized configuration key "quarkus.langchain4j.vertexai.devopi.projectId" was provided; it will be ignored; verify that the dependency extension for this configuration is set or that 
...
Caused by: io.smallrye.config.ConfigValidationException: Configuration validation failed:
	SRCFG00014: The config property quarkus.langchain4j.vertexai.devopi.location is required but it could not be found in any config source
	at io.quarkiverse.langchain4j.vertexai.runtime.gemini.VertexAiGeminiRecorder.chatModel(VertexAiGeminiRecorder.java:26)
	at io.quarkus.deployment.steps.VertexAiGeminiProcessor$generateBeans1607297205.deploy_0(Unknown Source)
	at io.quarkus.deployment.steps.VertexAiGeminiProcessor$generateBeans1607297205.deploy(Unknown Source)
	... 8 more

my inital config that lead to the error:

  langchain4j:
    devopi:
      chat-model:
        provider: vertexai-gemini

    vertexai:
      devopi:
        projectId: "<my-project>"
        location: "<region>"
      gemini:
        chat-model:
          model-id: "gemini-1.0-pro"

config that worked after I later realized that the config path had to be quarkus.langchain4j.vertexai.gemini.devopi.location:

  langchain4j:
    devopi:
      chat-model:
        provider: vertexai-gemini

    vertexai:
      gemini:
        devopi:
          projectId: "<my-project>"
          location: "<region>"
        chat-model:
          chat-model: # why is chat-model repeated?
            model-id: "gemini-1.0-pro"
geoand commented

Hi,

Actually there are two different extensions, quarkus-langchain4j-vertex-ai-gemini which is meant to be used with Google's latest Gemini models and quarkus-langchain4j-vertex-ai which is meant to be used with the older Palm models.

The configuration for the former is under quarkus.langchain4j.vertexai.gemini while for the latter it is quarkus.langchain4j.vertexai.

Are you trying to use both of them at the same time?

exaV commented

Hi

I'm trying to use only gemini and I only have a dependency to quarkus-langchain4j-vertex-ai-gemini in my pom. I also have a dependency on langchain4j-vertex-ai-gemini (without quarkus) because that is was I was using before, but I don't think that is causing this issue.

geoand commented

also have a dependency on langchain4j-vertex-ai-gemini (without quarkus) because that is was I was using before

You can safely remove this.

Also, if you only have this provider, then you don't need to set quarkus.langchain4j.chat-model.provider or quarkus.langchain4j.whatever.chat-model.provider.

Question, in your configuration, why are you using devopi?
Quarkus is complaining about missing mymodel configuration because that is the named model you seem to be using in your code.

exaV commented

I replaced devopi with mymodel when I created the issue, but apparently I didn't do a good job there, sorry about the confusion. I edited my first message and it should now say "devopi" everywhere.

I also use quarkus-langchain4j-openai that's why I configured a provider.

geoand commented

Any chance you can attach a sample project so I can have a look?

geoand commented

That indeed looks wrong!

Would you like to contribute a fix as you have already pinpointed the problem?

exaV commented

Sure. #558