Class OllamaModelProvider

java.lang.Object
com.oracle.coherence.rag.model.ollama.OllamaModelProvider
All Implemented Interfaces:
ModelProvider
@ApplicationScoped @Named("Ollama") public class OllamaModelProvider extends Object implements ModelProvider
ModelProvider implementation for Ollama models.

This provider enables integration with Ollama, a tool for running large language models locally. Ollama supports a wide variety of models including Llama, Mistral, Code Llama, and many others, providing both chat and embedding capabilities.

The provider supports all three model types: embedding models for creating vector representations of text, chat models for conversational AI, and streaming chat models for real-time response generation.

Configuration is managed through MicroProfile Config with the following property:

  • ollama.base.url - Base URL for Ollama service (defaults to http://localhost:11434)

Example usage: 


 @Inject
 @Named("Ollama")
 ModelProvider ollamaProvider;
 
 EmbeddingModel embeddingModel = ollamaProvider.getEmbeddingModel("nomic-embed-text");
 ChatModel chatModel = ollamaProvider.getChatModel("llama3.1:8b");
 StreamingChatModel streamingModel = ollamaProvider.getStreamingChatModel("llama3.1:8b");

Since:
25.09
Author:
Aleks Seovic 2025.07.04

  • Constructor Details

    • OllamaModelProvider

      @Inject public OllamaModelProvider(org.eclipse.microprofile.config.Config config, ConfigRepository jsonConfig)
      Default constructor for CDI initialization.

  • Method Details

    • getEmbeddingModel

      public dev.langchain4j.model.embedding.EmbeddingModel getEmbeddingModel(String sName)
      Returns an embedding model instance for the specified model name.

      Embedding models convert text into vector representations that can be used for semantic similarity search and document retrieval operations.

      Creates an Ollama embedding model for generating vector representations of text. Supports models like nomic-embed-text, mxbai-embed-large, and other embedding-capable models available in Ollama.

      Specified by:
      getEmbeddingModel in interface ModelProvider
      Parameters:
      sName - the name of the embedding model to create
      Returns:
      a configured OllamaEmbeddingModel instance
      Throws:
      IllegalArgumentException - if the model name is null or empty

    • getChatModel

      public dev.langchain4j.model.chat.ChatModel getChatModel(String sName)
      Returns a chat model instance for the specified model name.

      Chat models provide conversational AI capabilities for generating responses to user queries in a blocking manner. These models are suitable for synchronous chat interactions.

      Creates an Ollama chat model for conversational AI. Supports a wide variety of models including Llama, Mistral, Code Llama, and others.

      Specified by:
      getChatModel in interface ModelProvider
      Parameters:
      sName - the name of the chat model to create (e.g., "llama3.1:8b")
      Returns:
      a configured OllamaChatModel instance
      Throws:
      IllegalArgumentException - if the model name is null or empty

    • getStreamingChatModel

      public dev.langchain4j.model.chat.StreamingChatModel getStreamingChatModel(String sName)
      Returns a streaming chat model instance for the specified model name.

      Streaming chat models provide conversational AI capabilities with streaming response generation, allowing for real-time token-by-token response delivery. This is useful for creating responsive chat interfaces.

      Creates an Ollama streaming chat model for real-time response generation. Enables progressive response streaming suitable for interactive applications.

      Specified by:
      getStreamingChatModel in interface ModelProvider
      Parameters:
      sName - the name of the streaming chat model to create (e.g., "llama3.1:8b")
      Returns:
      a configured OllamaStreamingChatModel instance
      Throws:
      IllegalArgumentException - if the model name is null or empty

    • ollamaBaseUrl

      protected String ollamaBaseUrl()
      Returns the Ollama base URL from configuration.
      Returns:
      the base URL for Ollama service, defaults to http://localhost:11434