data-extract: Upgrade to granite 4

Author: aldettinger

data-extract-langchain4j/README.adoc

@@ -15,7 +15,7 @@ In this example, we'll convert those text conversations into Java Objects that c
 image::schema.png[]
 
 In order to achieve this extraction, we'll need a https://en.wikipedia.org/wiki/Large_language_model[Large Language Model (LLM)] and related serving framework that natively supports https://ollama.com/blog/structured-outputs[JSON structured output].
-Here, we choose https://ollama.com/library/granite3.3:2b[granite3.3:2b] served through https://ollama.com/[ollama] as it seems compute friendly and under Apache V2 license.
+Here, we choose https://ollama.com/library/granite4:3b-h[granite4:3b-h] served through https://ollama.com/[ollama] as it seems compute friendly and under Apache V2 license.
 In order to request inference to the served model, we'll use the high-level LangChain4j APIs like https://docs.langchain4j.dev/tutorials/ai-services[AiServices].
 
 === Start the Large Language Model
@@ -24,21 +24,21 @@ Let's start a container to serve the LLM with Ollama, in a first shell type:
 
 [source,shell]
 ----
-docker run --rm -it -v cqex-data-extract-ollama:/root/.ollama -p 11434:11434 --name cqex-data-extract-ollama ollama/ollama:0.9.3
+docker run --rm -it -v cqex-data-extract-ollama:/root/.ollama -p 11434:11434 --name cqex-data-extract-ollama ollama/ollama:0.14.1
 ----
 
 After a moment, a log like below should be output:
 
 [source,shell]
 ----
-time=2026-01-06T14:21:55.578Z level=INFO source=types.go:130 msg="inference compute" id=0 library=cpu variant="" compute="" driver=0.0 name="" total="62.2 GiB" available="49.9 GiB"
+time=2026-01-07T14:28:19.092Z level=INFO source=types.go:60 msg="inference compute" id=cpu library=cpu compute="" name=cpu description=cpu libdirs=ollama driver="" pci_id="" type="" total="62.2 GiB" available="62.2 GiB"
 ----
 
 Then, download the codellama model, in a second shell type:
 
 [source,shell]
 ----
-docker exec -it cqex-data-extract-ollama ollama pull granite3.3:2b
+docker exec -it cqex-data-extract-ollama ollama pull granite4:3b-h
 ----
 
 After a moment, log like below should be output:

data-extract-langchain4j/src/main/java/org/acme/extraction/CustomPojoExtractionService.java

@@ -53,7 +53,8 @@ public String toString() {
     String CUSTOM_POJO_EXTRACT_PROMPT = "Extract information about a customer from the transcript delimited by triple backticks: ```{{text}}```."
             + "The customerBirthday field should be formatted as {{dateFormat}}."
             + "The summary field should concisely relate the customer main ask."
-            + "Source any extracted field values from what is explicitly mentioned in the transcript.";
+            + "Source any extracted field values from what is explicitly mentioned in the transcript."
+            + "Extracted field values should be as accurate as possible.";
 
     /**
      * The text and dateFormat parameters of this method are automatically injected as {{text}} & {{dateFormat}} in the

data-extract-langchain4j/src/main/resources/application.properties

@@ -20,4 +20,4 @@ quarkus.default-locale=en_US
 # Adjust as per your where your Ollama instance is running
 langchain4j.ollama.base-url=http://localhost:11434
 # The chat model to use
-langchain4j.ollama.chat-model.model-id=granite3.3:2b
+langchain4j.ollama.chat-model.model-id=granite4:3b-h

data-extract-langchain4j/src/test/java/org/acme/extraction/OllamaTestResource.java

@@ -36,7 +36,7 @@ public class OllamaTestResource implements QuarkusTestResourceLifecycleManager {
 
     private static final Logger LOG = LoggerFactory.getLogger(OllamaTestResource.class);
 
-    private static final String OLLAMA_IMAGE = "ollama/ollama:0.9.3";
+    private static final String OLLAMA_IMAGE = "ollama/ollama:0.14.1";
     private static final int OLLAMA_SERVER_PORT = 11434;
 
     private static final String MODE_MOCK = "mock";

data-extract-langchain4j/src/test/resources/mappings/api_chat-3867080e-87d5-40d6-b9a8-b90817d980f9.json

@@ -0,0 +1,24 @@
+{
+  "id" : "4b63856f-cac5-4bc1-b66a-d01255ac40e0",
+  "name" : "api_chat",
+  "request" : {
+    "url" : "/api/chat",
+    "method" : "POST",
+    "bodyPatterns" : [ {
+      "equalToJson" : "{\n  \"model\" : \"granite4:3b-h\",\n  \"messages\" : [ {\n    \"role\" : \"user\",\n    \"content\" : \"Extract information about a customer from the transcript delimited by triple backticks: ```Operator: Hello, how may I help you ?\\nCustomer: Hello, I'm calling because I need to declare an accident on my main vehicle.\\nOperator: Ok, can you please give me your name ?\\nCustomer: My name is Sarah London.\\nOperator: Could you please give me your birth date ?\\nCustomer: 1986, July the 10th.\\nOperator: Ok, I've got your contract and I'm happy to share with you that we'll be able to reimburse all expenses linked to this accident.\\nCustomer: Oh great, many thanks.```.The customerBirthday field should be formatted as YYYY-MM-DD.The summary field should concisely relate the customer main ask.Source any extracted field values from what is explicitly mentioned in the transcript.Extracted field values should be as accurate as possible.\"\n  } ],\n  \"options\" : {\n    \"temperature\" : 0.0,\n    \"top_k\" : 1,\n    \"top_p\" : 0.1,\n    \"stop\" : [ ]\n  },\n  \"format\" : {\n  \"type\" : \"object\",\n  \"properties\" : {\n    \"customerSatisfied\" : {\n      \"type\" : \"boolean\"\n    },\n    \"customerName\" : {\n      \"type\" : \"string\"\n    },\n    \"customerBirthday\" : {\n      \"type\" : \"object\",\n      \"properties\" : {\n        \"year\" : {\n          \"type\" : \"integer\"\n        },\n        \"month\" : {\n          \"type\" : \"integer\"\n        },\n        \"day\" : {\n          \"type\" : \"integer\"\n        }\n      },\n      \"required\" : [ ]\n    },\n    \"summary\" : {\n      \"type\" : \"string\"\n    }\n  },\n  \"required\" : [ \"customerSatisfied\", \"customerName\", \"customerBirthday\", \"summary\" ]\n},\n  \"stream\" : false,\n  \"tools\" : [ ]\n}",
+      "ignoreArrayOrder" : true,
+      "ignoreExtraElements" : true
+    } ]
+  },
+  "response" : {
+    "status" : 200,
+    "body" : "{\"model\":\"granite4:3b-h\",\"created_at\":\"2026-01-15T15:03:26.59496985Z\",\"message\":{\"role\":\"assistant\",\"content\":\"{\\n  \\\"customerSatisfied\\\": true,\\n  \\\"customerName\\\": \\\"Sarah London\\\",\\n  \\\"customerBirthday\\\": {\\\"year\\\": 1986, \\\"month\\\": 7, \\\"day\\\": 10},\\n  \\\"summary\\\": \\\"The customer, Sarah London, called to declare an accident on her main vehicle. The operator confirmed that all expenses linked to the accident would be reimbursed.\\\"\\n}\"},\"done\":true,\"done_reason\":\"stop\",\"total_duration\":10290825797,\"load_duration\":1189765918,\"prompt_eval_count\":205,\"prompt_eval_duration\":3886962278,\"eval_count\":81,\"eval_duration\":4312366938}",
+    "headers" : {
+      "Date" : "Thu, 15 Jan 2026 15:03:26 GMT",
+      "Content-Type" : "application/json; charset=utf-8"
+    }
+  },
+  "uuid" : "4b63856f-cac5-4bc1-b66a-d01255ac40e0",
+  "persistent" : true,
+  "insertionIndex" : 6
+}