conductor-oss/conductor · error · UnsupportedOperationException

Not supported

Error message

Not supported

What it means

HuggingFace.java throws UnsupportedOperationException from generateEmbeddings() with message "Not supported". HuggingFace is wired to OpenAI Responses API for chat (see constructor at line 60-62), but embeddings are not implemented. AIModel.generateEmbeddings() (line 97) is abstract, so the provider must override it — HuggingFace declines by throwing. Called via LLMHelper.generateEmbeddings() which delegates directly to llm.generateEmbeddings().

Source

Thrown at ai/src/main/java/org/conductoross/conductor/ai/providers/huggingface/HuggingFace.java:67

    }

    public HuggingFace(HuggingFaceConfiguration config, OkHttpClient httpClient) {
        this.config = config;
        // Bearer auth (azureAuth=false via the 3-arg constructor). baseURL is the
        // router /v1 root; the client appends /responses.
        OpenAIResponsesApi responsesApi =
                new OpenAIResponsesApi(httpClient, config.getApiKey(), config.getBaseURL());
        this.chatModel = new OpenAIResponsesChatModel(responsesApi);
    }

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

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

    @Override
    public ChatOptions getChatOptions(ChatCompletion input) {
        List<Tool> tools = convertTools(input);
        return OpenAIResponsesChatOptions.builder()
                .model(input.getModel())
                .temperature(input.getTemperature())
                .topP(input.getTopP())
                .frequencyPenalty(input.getFrequencyPenalty())
                .presencePenalty(input.getPresencePenalty())
                .maxTokens(input.getMaxTokens())
                .stopSequences(input.getStopWords())
                .jsonOutput(input.isJsonOutput())
                .responsesApiTools(tools.isEmpty() ? null : tools)
                .build();
    }

View on GitHub (pinned to cf7c3e4a8a)

Solutions

  1. Route the embedding task to a provider that implements generateEmbeddings(): openai, gemini, azureopenai, bedrock, ollama, or cohere.
  2. If you need HuggingFace embeddings, subclass HuggingFace and implement generateEmbeddings() by calling HuggingFace's /embeddings or sentence-transformers endpoint.
  3. Split workflow: use HuggingFace for chat, openai/ollama for embedding.

Example fix

// before
{"llmProvider": "huggingface", "model": "bge-large-en", "text": "hello"}
// after
{"llmProvider": "openai", "model": "text-embedding-3-small", "text": "hello"}
Defensive patterns

Strategy: validation

Validate before calling

// Check provider supports embeddings before calling generateEmbeddings()
private static final Set<String> EMBEDDING_CAPABLE = Set.of(
    "openai", "gemini", "azureopenai", "bedrock", "ollama", "cohere", "mistral");

String provider = request.getLlmProvider();
if (!EMBEDDING_CAPABLE.contains(provider)) {
    throw new IllegalArgumentException(
        "Provider '" + provider + "' does not support embeddings. " +
        "Supported: " + EMBEDDING_CAPABLE);
}

Type guard

null

Try / catch

try {
    List<Float> embeddings = llm.generateEmbeddings(request);
} catch (UnsupportedOperationException e) {
    throw new IllegalArgumentException(
        "Provider '" + providerName + "' does not support embeddings: " + e.getMessage(), e);
}

Prevention

When it happens

Trigger: A VectorDB worker or LLMEmbeddingGen task is configured with llmProvider="huggingface". VectorDBWorkers.generateEmbeddings() (line 75) or LLMWorkers calls llm.generateEmbeddings(), which throws UnsupportedOperationException before any HTTP request.

Common situations: Using HuggingFace TGI (Text Generation Inference) for chat but routing an embedding/indexing pipeline to the same provider; assuming HuggingFace's inference endpoints cover embeddings (they do have an embedding API but this provider doesn't wire it).

Related errors

AI-assisted analysis of conductor-oss/conductor@cf7c3e4a8a (2026-08-14). Data as JSON: /api/errors/875105b3c78a71f4. Report an issue: GitHub.