Spring AI Stalling Indefinitely With LMStudio? Here's the simple fix.

Spring AI stalls indefinitely if you try to use a local model running in LMStudio. The issue is that LMStudio isn't properly handling HTTP/2 connections from Spring, which is the default being used.

It's a very simple fix though. When defining the OpenAiApi instance, set the webClientBuilder() and restClientBuilder() options to use HTTP/1.1 instead.

// define a custom HTTP connector for HTTP 1.1
JdkClientHttpConnector httpConnector = new JdkClientHttpConnector(HttpClient.newBuilder()
        .version(HttpClient.Version.HTTP_1_1)
        .connectTimeout(Duration.ofSeconds(30))
        .build());

// define a custom HTTP connector for HTTP 1.1
JdkClientHttpRequestFactory httpRequestFactory = new JdkClientHttpRequestFactory(HttpClient.newBuilder()
        .version(HttpClient.Version.HTTP_1_1)
        .connectTimeout(Duration.ofSeconds(30))
        .build());

// build the OpenAI API client
OpenAiApi api = OpenAiApi.builder()
        .baseUrl(url)
        .apiKey(new NoopApiKey())
        .webClientBuilder(
                // Force HTTP/1.1 for streaming
                WebClient.builder().clientConnector(httpConnector))
        .restClientBuilder(
                // Force HTTP/1.1 for non-streaming
                RestClient.builder().requestFactory(httpRequestFactory))
        .build();

Source: https://github.com/spring-projects/spring-ai/issues/2441