Use 'model' config type for LLM models

2 files changed, 3 insertions, 16 deletions

diff --git a/container-search/src/main/java/ai/vespa/llm/clients/LocalLLM.java b/container-search/src/main/java/ai/vespa/llm/clients/LocalLLM.java
index 3b99e5f0a09..1e204d29a19 100644
--- a/container-search/src/main/java/ai/vespa/llm/clients/LocalLLM.java
+++ b/container-search/src/main/java/ai/vespa/llm/clients/LocalLLM.java
@@ -44,7 +44,7 @@ public class LocalLLM extends AbstractComponent implements LanguageModel {
         // Only used if GPU is not used
         var defaultThreadCount = Runtime.getRuntime().availableProcessors() - 2;
 
-        var modelFile = selectModelFile(config);
+        var modelFile = config.model().toFile().getAbsolutePath();
         var modelParams = new ModelParameters()
                 .setModelFilePath(modelFile)
                 .setContinuousBatching(true)
@@ -69,16 +69,6 @@ public class LocalLLM extends AbstractComponent implements LanguageModel {
                 new ThreadPoolExecutor.AbortPolicy());
     }
 
-    private String selectModelFile(LlmLocalClientConfig config) {
-        if ( ! config.localLlmFile().isEmpty()) {  // primarily for testing
-            return config.localLlmFile();
-        } else if (config.modelUrl().exists()) {
-            return config.modelUrl().getAbsolutePath();
-        }
-        throw new IllegalArgumentException("Local LLM model not set. " +
-                "Either set 'localLlmFile' or 'modelUrl' in 'llm-local-client' config.");
-    }
-
     @Override
     public void deconstruct() {
         logger.info("Closing LLM model...");
diff --git a/container-search/src/main/resources/configdefinitions/llm-local-client.def b/container-search/src/main/resources/configdefinitions/llm-local-client.def
index 08eab19f0f8..c06c24b33e5 100755
--- a/container-search/src/main/resources/configdefinitions/llm-local-client.def
+++ b/container-search/src/main/resources/configdefinitions/llm-local-client.def
@@ -1,11 +1,8 @@
 # Copyright Vespa.ai. Licensed under the terms of the Apache 2.0 license. See LICENSE in the project root.
 package=ai.vespa.llm.clients
 
-# Url to the model to use
-modelUrl url default=""
-
-# Local file path to the model to use - will have precedence over model_url if set - mostly for testing
-localLlmFile string default=""
+# The LLM model to use
+model model
 
 # Maximum number of requests to handle in parallel pr container node
 parallelRequests int default=10