microsoft/autogen · error · ValueError

default_chat_model not found in config.models

Error message

default_chat_model not found in config.models

What it means

GraphRAG GlobalSearchTool.from_settingsDir loads settings.yaml via load_config and looks up config.models['default_chat_model']. If that key is absent (models dict missing or renamed), it raises ValueError('default_chat_model not found in config.models'). The global search indexer/querier requires a chat model under that exact reserved ID to build its internal ModelManager model.

Source

Thrown at python/packages/autogen-ext/src/autogen_ext/tools/graphrag/_global_search.py:206

        Args:
            root_dir: Path to the GraphRAG root directory
            config_filepath: Path to the GraphRAG settings file (optional)

        Returns:
            An initialized GlobalSearchTool instance
        """
        # Load GraphRAG config
        if isinstance(root_dir, str):
            root_dir = Path(root_dir)
        if isinstance(config_filepath, str):
            config_filepath = Path(config_filepath)
        config = load_config(root_dir=root_dir, config_filepath=config_filepath)

        # Get the language model configuration from the models section
        chat_model_config = config.models.get(defs.DEFAULT_CHAT_MODEL_ID)

        if chat_model_config is None:
            raise ValueError("default_chat_model not found in config.models")

        # Initialize token encoder based on the model being used
        try:
            token_encoder = tiktoken.encoding_for_model(chat_model_config.model)
        except KeyError:
            # Fallback to cl100k_base if model is not recognized by tiktoken
            token_encoder = tiktoken.get_encoding("cl100k_base")

        # Create the LLM using ModelManager
        model = ModelManager().get_or_create_chat_model(
            name="global_search_model",
            model_type=chat_model_config.type,
            config=chat_model_config,
        )

        # Create data config from storage paths
        data_config = DataConfig(
            input_dir=str(config.output.base_dir),

View on GitHub (pinned to 027ecf0a37)

Solutions

1. Open settings.yaml under root_dir and ensure a models: section contains a default_chat_model entry with type, model, and credentials (api_key etc.).
2. If your settings predate the models schema, regenerate or migrate it: move the chat model block under models with key default_chat_model.
3. Confirm root_dir/config_filepath actually resolve to the settings file you edited (log the resolved path).
4. Verify load_config did not silently fall back to defaults because the yaml filename is non-standard.

Example fix

# settings.yaml — before
model:
  type: openai_chat
  model: gpt-4o

# settings.yaml — after
models:
  default_chat_model:
    type: openai_chat
    model: gpt-4o
    api_key: ${GRAPHRAG_API_KEY}

from autogen_ext.tools.graphrag._config_loader import load_config

cfg = load_config(root_dir=root_dir, config_filepath=config_filepath)
if "default_chat_model" not in cfg.models:
    raise ValueError("settings.yaml missing models.default_chat_model")

def has_default_chat_model(cfg) -> TypeGuard[type("GraphRAGConfig")]:
    return isinstance(cfg.models, dict) and "default_chat_model" in cfg.models

try:
    tool = await GlobalSearchTool.from_settings_dir(root_dir)
except ValueError as e:
    if "default_chat_model" in str(e):
        raise ConfigError("Add models.default_chat_model to settings.yaml") from e
    raise

When it happens

Trigger: Calling GraphRAG GlobalSearchTool.from_settings_dir(root_dir, config_filepath) where the loaded settings.yaml has no models section, or a models section without the key 'default_chat_model'. The lookup is config.models.get(defs.DEFAULT_CHAT_MODEL_ID) and raises when it returns None.

Common situations: Using an old GraphRAG settings.yaml from before the models-migration (model config lived at top level); hand-editing settings.yaml and renaming the model entry; running the newer AutoGen GraphRAG tools against a workspace generated by the graphrag CLI of a different version; pointing root_dir at the wrong directory so a stale/empty settings.yaml is found.

Related errors

• default_chat_model not found in config.models
• default_embedding_model not found in config.models
• Failed to fetch settings
• Failed to update settings
• modelName is a required property for LMStudioConfig and cann

AI-assisted analysis of microsoft/autogen@027ecf0a37 (2026-08-15). Data as JSON: /api/errors/185aaf772632aba7. Report an issue: GitHub.