Support Ollama as other LLM provider

[arey]

LangChain4j enables integration with 15+ LLM providers, allowing us to incorporate those that support the specific LangChain4j features used in this application: Streaming, Tools, and JSON mode.
Potential candidates include Google Vertex AI Gemini, Ollama, and Mistral AI.

This task should also take into account LLM switching, possibly through a Maven profile.

[arey]

Test result with Ollama and the llama3.3 model.

Issue 1: toolExecutor is null
When asking the question How many veterinary cardiologists are there? the LLM asks to call the function getAllVets which does not exist.
This is a known bug: langchain4j/langchain4j#1052 and ollama/ollama#8294

Details

2024-12-29T16:28:16.219+01:00 ERROR 58618 --- [nio-8080-exec-2] o.a.c.c.C.[.[.[/].[dispatcherServlet] : Servlet.service() for servlet [dispatcherServlet] in context with path [] threw exception [Request processing failed: java.lang.NullPointerException: Cannot invoke "dev.langchain4j.service.tool.ToolExecutor.execute(dev.langchain4j.agent.tool.ToolExecutionRequest, Object)" because "toolExecutor" is null] with root cause

java.lang.NullPointerException: Cannot invoke "dev.langchain4j.service.tool.ToolExecutor.execute(dev.langchain4j.agent.tool.ToolExecutionRequest, Object)" because "toolExecutor" is null
at dev.langchain4j.service.AiServiceStreamingResponseHandler.onCompleteResponse(AiServiceStreamingResponseHandler.java:93) ~[langchain4j-1.0.0-alpha1.jar:na]
at dev.langchain4j.model.chat.StreamingChatLanguageModel$1.onComplete(StreamingChatLanguageModel.java:69) ~[langchain4j-core-1.0.0-alpha1.jar:na]
at dev.langchain4j.model.ollama.OllamaClient$2.onResponse(OllamaClient.java:180) ~[langchain4j-ollama-1.0.0-alpha1.jar:na]
at retrofit2.OkHttpCall$1.onResponse(OkHttpCall.java:161) ~[retrofit-2.9.0.jar:na]
at okhttp3.internal.connection.RealCall$AsyncCall.run(RealCall.kt:519) ~[okhttp-4.12.0.jar:na]
at java.base/java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1144) ~[na:na]
at java.base/java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:642) ~[na:na]
at java.base/java.lang.Thread.run(Thread.java:1583) ~[na:na]

To remove this hallucination, we try add the following command to the command prompt:

Don't provide a function name that doesn't exist.

When dealing with vets, use the user augmented prompt instead of hallucinating a function tool name that does not exist (ie. getVets, getVetsWithSpecialties, filterVetsBySpecialty).

We also try to decrease the temperature to 0.5 and 0.2

Issue 2: display technical ID

When asking the question Please list the owners that come to the clinic. the LLM may display the ID of each owners.

Solution: add the following command to the command prompt.

Don't display technical information such as ID.

[arey]

I've tried the deepseek-r1:14b model with the LangChain4j 1.0.0-alpha1 version and I've got the following exception:

2025-02-01T19:12:40.061+01:00 ERROR 93418 --- [lhost:11434/...] o.s.s.p.chat.AssistantController         : Unexpected chat error

java.lang.NullPointerException: Cannot invoke "okhttp3.ResponseBody.byteStream()" because the return value of "retrofit2.Response.body()" is null
	at dev.langchain4j.model.ollama.OllamaClient$2.onResponse(OllamaClient.java:164) ~[langchain4j-ollama-1.0.0-alpha1.jar:na]
	at retrofit2.OkHttpCall$1.onResponse(OkHttpCall.java:161) ~[retrofit-2.9.0.jar:na]
	at okhttp3.internal.connection.RealCall$AsyncCall.run(RealCall.kt:519) ~[okhttp-4.12.0.jar:na]
	at java.base/java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1144) ~[na:na]
	at java.base/java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:642) ~[na:na]
	at java.base/java.lang.Thread.run(Thread.java:1583) ~[na:na]