invoke-ai/InvokeAI · error · ValueError

Model '{model_key}' is not a TextLLM model (got {model_confi

Error message

Model '{model_key}' is not a TextLLM model (got {model_config.type})

What it means

ValueError raised by _run_expand_prompt when the model identified by model_key exists in the model store but its configured type is not ModelType.TextLLM. It is surfaced by expand_prompt as HTTP 422 with the same message. This is a caller-supplied model_key mismatch against the type-strict inference path.

Source

Thrown at invokeai/app/api/routers/utilities.py:152

    return progress_callback


def _run_expand_prompt(
    prompt: str,
    model_key: str,
    max_tokens: int,
    system_prompt: str | None,
    seed: int | None,
    task_id: str | None,
    user_id: str,
) -> tuple[str, int]:
    """Run text LLM inference synchronously (called from thread)."""
    model_manager = ApiDependencies.invoker.services.model_manager
    events = ApiDependencies.invoker.services.events
    model_config = model_manager.store.get_model(model_key)

    if model_config.type != ModelType.TextLLM:
        raise ValueError(f"Model '{model_key}' is not a TextLLM model (got {model_config.type})")

    if task_id is not None:
        events.emit_llm_task_progress(task_id=task_id, user_id=user_id, phase="loading_model", message="Loading model")

    with _model_load_lock:
        loaded_model = model_manager.load.load_model(model_config, user_id=user_id)

    with torch.no_grad(), loaded_model.model_on_device() as (_, model):
        model_abs_path = _resolve_model_path(model_config.path)
        tokenizer = AutoTokenizer.from_pretrained(model_abs_path, local_files_only=True)

        pipeline = TextLLMPipeline(model, tokenizer)
        model_device = next(model.parameters()).device

        progress_callback = _make_progress_callback(events, task_id, user_id)

        effective_seed = seed if seed is not None else get_random_seed()
        output = pipeline.run(

View on GitHub (pinned to 0b6a024f2f)

Solutions

Pass a model_key whose model type is TextLLM (check via the model manager store or /api/v1/models listing)
Fix the default model_key in configuration used by the prompt-expansion feature
Install/convert a TextLLM model if none is present
Upgrade InvokeAI if your model's type string isn't recognized as TextLLM

Example fix

// before
expand_prompt(model_key="main:stablediffusion-xl")
// after
expand_prompt(model_key="text_llm:my-llm-model")

Defensive patterns

Strategy: type-guard

Validate before calling

const models = await api.listModels();
const cfg = models.find(m => m.key === modelKey);
if (!cfg || cfg.type !== 'text_llm') {
  throw new Error(`model_key must reference a TextLLM model (got ${cfg ? cfg.type : 'unknown'})`);
}

Type guard

function isTextLlmModel(config) {
  return config != null && config.type === 'text_llm' && typeof config.key === 'string';
}

Try / catch

try {
  await api.expandPrompt({ modelKey, prompt });
} catch (e) {
  if (e.status === 422 && /not a TextLLM model/.test(e.detail)) {
    console.error('Wrong model type; pick a TextLLM model');
  } else throw e;
}

Prevention

Filter model dropdowns to type text_llm for prompt expansion
Store the model key from an API listing, never hand-written
Validate model type at startup if the key comes from config/env

When it happens

Trigger: POST expand_prompt (or the test path test_expand_prompt_uses_fresh_seed) with a model_key whose model_config.type is e.g. Main, LlavaOnevision, or Embedding instead of TextLLM.

Common situations: Selecting a diffusion/LLaVA model in a UI dropdown that doesn't filter by TextLLM; config/env pointing prompt expansion at the wrong model key; older installers/models predating the TextLLM type.

Related errors

AI-assisted analysis of invoke-ai/InvokeAI@0b6a024f2f (2026-08-29). Data as JSON: /api/errors/93769b91ca3b8c86. Report an issue: GitHub.