langgenius/dify · error · ValueError

NEW_USER_DEFAULT_MODELS contains duplicate model type: {mode

Error message

NEW_USER_DEFAULT_MODELS contains duplicate model type: {model_type}

What it means

Raised by NEW_USER_DEFAULT_MODEL_LIST when two entries share the same model_type (first segment). The default-models logic reserves one default per model type, so duplicates would non-deterministically override each other at app-provisioning time.

Source

Thrown at api/configs/feature/__init__.py:335

        default="",
    )

    @property
    def NEW_USER_DEFAULT_MODEL_LIST(self) -> list[tuple[str, str, str]]:
        default_models: list[tuple[str, str, str]] = []
        configured_model_types: set[str] = set()

        for item in self.NEW_USER_DEFAULT_MODELS.split(","):
            if not item.strip():
                continue

            parts = tuple(part.strip() for part in item.split(":", 2))
            if len(parts) != 3 or not all(parts):
                raise ValueError("NEW_USER_DEFAULT_MODELS entries must use 'model_type:provider:model' format")

            model_type, provider, model = parts
            if model_type in configured_model_types:
                raise ValueError(f"NEW_USER_DEFAULT_MODELS contains duplicate model type: {model_type}")

            configured_model_types.add(model_type)
            default_models.append((model_type, provider, model))

        return default_models


class MarketplaceConfig(BaseSettings):
    """
    Configuration for marketplace
    """

    MARKETPLACE_ENABLED: bool = Field(
        description="Enable or disable marketplace",
        default=True,
    )

    MARKETPLACE_API_URL: HttpUrl = Field(

View on GitHub (pinned to ef8544b173)

Solutions

1. Keep at most one entry per model_type (e.g. only one 'llm' entry).
2. If you need multiple LLMs, set them per-workspace after provisioning rather than in NEW_USER_DEFAULT_MODELS.
3. Switch the second entry to a different model_type if a second default category is intended.

Example fix

// before
NEW_USER_DEFAULT_MODELS=llm:openai:gpt-4o,llm:anthropic:claude-3
// after
NEW_USER_DEFAULT_MODELS=llm:openai:gpt-4o

def no_duplicate_model_type(entries: str) -> bool:
    types = [e.split(':', 1)[0].strip() for e in entries.split(',') if e.strip()]
    return len(types) == len(set(types))

When it happens

Trigger: Setting NEW_USER_DEFAULT_MODELS='llm:openai:gpt-4o,llm:anthropic:claude-3' — both entries have model_type 'llm'.

Common situations: Wanting to default multiple providers for the same type; misunderstanding that the keying is by model_type, not by provider.

Related errors

• NEW_USER_DEFAULT_MODELS entries must use 'model_type:provide
• KNOWLEDGE_FS_BASE_URL must be an absolute HTTP(S) URL
• KNOWLEDGE_FS_BASE_URL must include a valid port
• KNOWLEDGE_FS_BASE_URL must not include credentials, query, o
• KNOWLEDGE_FS_BASE_URL and KNOWLEDGE_FS_JWT_SECRET must be co

AI-assisted analysis of langgenius/dify@ef8544b173 (2026-08-12). Data as JSON: /api/errors/4cbf5ef046dfd3dc. Report an issue: GitHub.