conductor-oss/conductor · error · RuntimeException

OpenAI Responses API call failed:

Error message

OpenAI Responses API call failed: 

What it means

OpenAIResponsesChatModel.call() wraps an IOException from OpenAIResponsesApi.createResponse() in a RuntimeException with message "OpenAI Responses API call failed: ". This is the primary chat model for the OpenAI provider (using the newer /v1/responses endpoint, not /chat/completions). The IOException originates from error 219 (non-2xx HTTP) or a network failure.

Source

Thrown at ai/src/main/java/org/conductoross/conductor/ai/providers/openai/OpenAIResponsesChatModel.java:70

 * converts the response back to Spring AI's {@link ChatResponse}.
 */
@Slf4j
public class OpenAIResponsesChatModel implements ChatModel {

    private final OpenAIResponsesApi responsesApi;

    public OpenAIResponsesChatModel(OpenAIResponsesApi responsesApi) {
        this.responsesApi = responsesApi;
    }

    @Override
    public ChatResponse call(Prompt prompt) {
        try {
            ResponseRequest request = buildRequest(prompt);
            ResponseResult result = responsesApi.createResponse(request);
            return toSpringChatResponse(result);
        } catch (IOException e) {
            throw new RuntimeException("OpenAI Responses API call failed: " + e.getMessage(), e);
        }
    }

    private ResponseRequest buildRequest(Prompt prompt) throws JsonProcessingException {
        List<Message> messages = prompt.getInstructions();
        ChatOptions options = prompt.getOptions();

        // Extract system messages → instructions field
        String instructions = null;
        List<Message> nonSystemMessages = new ArrayList<>();
        for (Message msg : messages) {
            if (msg.getMessageType() == MessageType.SYSTEM) {
                // Concatenate multiple system messages
                String sysText = ((SystemMessage) msg).getText();
                instructions = instructions == null ? sysText : instructions + "\n" + sysText;
            } else {
                nonSystemMessages.add(msg);
            }

View on GitHub (pinned to cf7c3e4a8a)

Solutions

1. Inspect getCause() for the IOException — its message includes the HTTP status and response body (see error 219).
2. Verify the model supports the Responses API (gpt-4o, gpt-4o-mini, o1, o3, o4-mini, gpt-5 series).
3. Check that the API key is valid and has access to the requested model.
4. For 400 errors mentioning temperature with o-series models, the code already auto-retries without temperature — if it still fails, check other unsupported parameters.

Example fix

// before: using a model only available via Chat Completions with Responses API
{"model": "gpt-3.5-turbo", "prompt": "..."}
// after
{"model": "gpt-4o", "prompt": "..."}

// Validate chat request and model before calling
if (input.getModel() == null || input.getModel().isBlank()) {
    throw new IllegalArgumentException("Model name is required");
}
// Verify model supports the Responses API
Set<String> responsesApiModels = Set.of(
    "gpt-4o", "gpt-4o-mini", "o1", "o1-mini", "o3", "o3-mini", "o4-mini",
    "gpt-5", "gpt-5-mini", "gpt-5-nano");
String modelPrefix = input.getModel().split("-")[0];
if (!responsesApiModels.contains(input.getModel())
    && !input.getModel().startsWith("o1")
    && !input.getModel().startsWith("o3")
    && !input.getModel().startsWith("o4")
    && !input.getModel().startsWith("gpt-4o")
    && !input.getModel().startsWith("gpt-5")) {
    log.warn("Model '{}' may not support the Responses API", input.getModel());
}

null

try {
    ChatResponse response = chatModel.call(prompt);
} catch (RuntimeException e) {
    Throwable cause = e.getCause();
    if (cause instanceof IOException io) {
        String msg = io.getMessage();
        if (msg.contains("429") || msg.contains("500") || msg.contains("503") || msg.contains("529")) {
            return retryWithBackoff(() -> chatModel.call(prompt));
        }
        log.error("Responses API error: {}", msg);
    }
    throw e;
}

When it happens

Trigger: The OpenAI provider's chat model call(Prompt) is invoked and POST /v1/responses fails: 401 invalid key, 429 rate limit, 400 invalid model or parameters, model not available for the Responses API, or network error.

Common situations: Expired API key; using a model that doesn't support the Responses API (some older models only work with Chat Completions); rate limit on high-throughput chat tasks; network timeout on long context windows; reasoning-model parameter mismatch (temperature rejected by o-series — note the code already retries without temperature).

Related errors

• Responses API failed with status %d: %s
• Embeddings API call failed:
• Speech API call failed:
• Chat Completions API call failed:
• Image generation API call failed:

AI-assisted analysis of conductor-oss/conductor@cf7c3e4a8a (2026-08-14). Data as JSON: /api/errors/e8d2e55f76d756ca. Report an issue: GitHub.