Skip to main content

Model

What it is​

  • Lightweight data containers for describing AI models (primarily chat models) along with provider metadata.
  • Uses typing.Annotated + pydantic.Field for field descriptions, but these classes are not Pydantic models.

Public API​

ModelType (Enum)​

  • CHAT: identifies chat models.

class Model​

Container for a model instance and associated metadata.

Constructor

  • Model(model_id: str, provider: str, model: BaseChatModel, ..., default_parameters: Optional[dict] = None)

Attributes (set by __init__)

  • model_id: str — unique identifier.
  • provider: str — provider name (e.g., "openai", "anthropic").
  • model: BaseChatModel — LangChain chat model instance.
  • Optional metadata:
    • name, owner, description, image
    • created_at: datetime
    • canonical_slug, hugging_face_id
    • pricing: dict, architecture: dict, top_provider: dict
    • per_request_limits: dict
    • supported_parameters: list
    • default_parameters: dict

class ChatModel(Model)​

Specialization for chat models.

Constructor

  • ChatModel(model_id: str, provider: str, model: BaseChatModel, context_window: Optional[int] = None, ..., default_parameters: Optional[dict] = None)

Additional attributes

  • context_window: Optional[int] — maximum context length in tokens.
  • model_type: ModelType — set to ModelType.CHAT.

Configuration/Dependencies​

  • langchain_core.language_models.chat_models.BaseChatModel (required type for model).
  • pydantic.Field is used only to attach descriptions in type annotations; no runtime validation is performed by these classes.

Usage​

from langchain_core.language_models.chat_models import BaseChatModel
from naas_abi_core.models.Model import ChatModel

# You must supply an actual LangChain BaseChatModel implementation.
chat_backend: BaseChatModel = ...

m = ChatModel(
    model_id="gpt-4.1",
    provider="openai",
    model=chat_backend,
    context_window=128000,
    name="GPT-4.1",
)

print(m.model_type)        # ModelType.CHAT
print(m.context_window)    # 128000
print(m.provider)          # "openai"

Caveats​

  • These are plain Python classes (not pydantic.BaseModel / dataclasses):
    • No automatic validation, parsing, or serialization is provided.
    • Field(...) metadata is not enforced at runtime.