What causes "Not supported" in conductor-oss/conductor?

Use an embeddings-capable provider (OpenAI) for embedding tasks instead of perplexity.

conductor-oss/conductor · error · UnsupportedOperationException

Not supported

Error message

Not supported

What it means

Thrown by PerplexityAI.generateEmbeddings as UnsupportedOperationException. Perplexity (Sonar models) is a search-augmented chat provider and does not offer an embeddings endpoint, so the AIModel interface method is intentionally unimplemented.

Source

Thrown at ai/src/main/java/org/conductoross/conductor/ai/providers/perplexity/PerplexityAI.java:55

        this(config, AIHttpClients.defaultClient());
    }

    public PerplexityAI(PerplexityAIConfiguration config, OkHttpClient httpClient) {
        this.config = config;
        OpenAIChatCompletionsApi api =
                new OpenAIChatCompletionsApi(
                        httpClient, config.getApiKey(), config.getBaseURL(), "/chat/completions");
        this.chatModel = new OpenAICompatChatModel(api);
    }

    @Override
    public String getModelProvider() {
        return NAME;
    }

    @Override
    public List<Float> generateEmbeddings(EmbeddingGenRequest embeddingGenRequest) {
        throw new UnsupportedOperationException("Not supported");
    }

    @Override
    public ChatOptions getChatOptions(ChatCompletion input) {
        return ToolCallingChatOptions.builder()
                .model(input.getModel())
                .maxTokens(input.getMaxTokens())
                .topP(input.getTopP())
                .temperature(input.getTemperature())
                .toolCallbacks(getToolCallback(input))
                .internalToolExecutionEnabled(false)
                .frequencyPenalty(input.getFrequencyPenalty())
                .topK(input.getTopK())
                .presencePenalty(input.getPresencePenalty())
                .build();
    }

    @Override

View on GitHub (pinned to cf7c3e4a8a)

Solutions

Use an embeddings-capable provider (OpenAI) for embedding tasks instead of perplexity.
Check the AIModel capabilities or getModelProvider() before calling generateEmbeddings.
Set the correct llmProvider on the embedding task input.

Example fix

// before
AIModel model = provider.getModel(req); // resolves to perplexity
List<Float> emb = model.generateEmbeddings(embReq); // UnsupportedOperationException

// after
if (!"perplexity".equals(model.getModelProvider())) {
    List<Float> emb = model.generateEmbeddings(embReq);
}

Defensive patterns

Strategy: validation

Validate before calling

if ("perplexity".equals(model.getModelProvider())) {
    throw new IllegalArgumentException(
        "Perplexity does not support embeddings; use an OpenAI provider");
}
List<Float> emb = model.generateEmbeddings(req);

Type guard

boolean supportsEmbeddings(AIModel m) {
    return !"perplexity".equals(m.getModelProvider())
        && !"stabilityai".equals(m.getModelProvider());
}

Try / catch

try {
    List<Float> emb = model.generateEmbeddings(req);
} catch (UnsupportedOperationException e) {
    // provider lacks embeddings; switch provider or skip
}

Prevention

Check getModelProvider() before calling generateEmbeddings.
Keep a known-capabilities map per provider name.
Route embedding tasks to OpenAI explicitly.

When it happens

Trigger: A workflow or caller invokes generateEmbeddings on an AIModel resolved to the 'perplexity' provider, e.g. an embedding task whose llmProvider is misconfigured to perplexity.

Common situations: Copy-pasting a chat-oriented provider config into an embeddings task, or a generic AIModel consumer that does not check provider capabilities before calling generateEmbeddings.

Related errors

AI-assisted analysis of conductor-oss/conductor@cf7c3e4a8a (2026-08-14). Data as JSON: /api/errors/bd181bb97f7c6e50. Report an issue: GitHub.