x

libs/core/langchain_core/__init__.py

@@ -14,6 +14,15 @@ from langchain_core._api import (
     surface_langchain_beta_warnings,
     surface_langchain_deprecation_warnings,
 )
+from langchain_core._api.shorthand import (
+    chat_model,
+    document_loader,
+    embeddings,
+    retriever,
+    tool,
+    toolkit,
+    vectorstore,
+)
 
 try:
     __version__ = metadata.version(__package__)
@@ -23,3 +32,14 @@ except metadata.PackageNotFoundError:
 
 surface_langchain_deprecation_warnings()
 surface_langchain_beta_warnings()
+
+__all__ = [
+    "__version__",
+    "chat_model",
+    "retriever",
+    "tool",
+    "toolkit",
+    "document_loader",
+    "vectorstore",
+    "embeddings",
+]

libs/core/langchain_core/_api/shorthand.py

@@ -0,0 +1,93 @@
+"""
+Functions that support shorthand access of LangChain classes like
+
+::code
+    import langchain_core as lc
+
+    model = lc.chat_model("claude-3-5-sonnet-20240620", provider="anthropic")
+"""
+
+import importlib
+from typing import TYPE_CHECKING, Literal, cast
+
+if TYPE_CHECKING:
+    from langchain_core.document_loaders import BaseLoader
+    from langchain_core.embeddings import Embeddings
+    from langchain_core.language_models import BaseChatModel
+    from langchain_core.retrievers import BaseRetriever
+    from langchain_core.tools import BaseTool, BaseToolkit
+    from langchain_core.vectorstores import VectorStore
+
+ProviderManifestType = dict[
+    Literal[
+        "chat_model",
+        "retriever",
+        "embeddings",
+        "vectorstore",
+        "tool",
+        "toolkit",
+        "document_loader",
+    ],
+    list[str],
+]
+_registered_providers: dict[
+    str,
+    ProviderManifestType,
+] = {}
+
+
+def register_manifest(
+    provider: str,
+    provider_manifest: ProviderManifestType,
+    *,
+    dangerously_allow_provider_overwrite: bool = False,
+) -> None:
+    if provider in _registered_providers and not dangerously_allow_provider_overwrite:
+        msg = (
+            f"Provider `{provider}` was already registered. To allow overwriting, "
+            "pass `dangerously_allow_provider_overwrite=True`"
+        )
+        raise ValueError(msg)
+    _registered_providers[provider] = provider_manifest
+
+
+def register(package_root: str) -> None:
+    manifest_module = importlib.import_module("lc_manifest", package_root)
+    if not hasattr(manifest_module, "manifest"):
+        msg = (
+            f"Package registration requires a {package_root}.lc_manifest.manifest "
+            "dictionary to be declared"
+        )
+        raise ValueError(msg)
+
+    manifest = cast(dict[str, ProviderManifestType], manifest_module.manifest)
+    for provider, provider_manifest in manifest.items():
+        register_manifest(provider, provider_manifest)
+
+
+def chat_model(model: str, *, provider: str | None = None, **kwargs) -> BaseChatModel:
+    pass
+
+
+def retriever(name: str, **kwargs) -> BaseRetriever:
+    pass
+
+
+def tool(name: str, **kwargs) -> BaseTool:
+    pass
+
+
+def toolkit(name: str, **kwargs) -> BaseToolkit:
+    pass
+
+
+def document_loader(name: str, **kwargs) -> BaseLoader:
+    pass
+
+
+def vectorstore(name: str, **kwargs) -> VectorStore:
+    pass
+
+
+def embeddings(model: str, *, provider: str | None = None, **kwargs) -> Embeddings:
+    pass

libs/partners/openai/langchain_openai/lc_manifest.py

@@ -0,0 +1,36 @@
+from typing import Literal
+
+# we want to register integration classes, as well as provide a hook
+
+manifest: dict[
+    str,
+    dict[
+        Literal[
+            "chat_model",
+            "retriever",
+            "embeddings",
+            "vectorstore",
+            "tool",
+            "toolkit",
+            "document_loader",
+        ],
+        list[str],
+    ],
+] = {
+    "openai": {
+        "chat_model": [
+            "langchain_openai.chat_models.base.ChatOpenAI",
+        ],
+        "embeddings": [
+            "langchain_openai.embeddings.base.OpenAIEmbeddings",
+        ],
+    },
+    "azure": {
+        "chat_model": [
+            "langchain_openai.chat_models.azure.AzureChatOpenAI",
+        ],
+        "embeddings": [
+            "langchain_openai.embeddings.azure.AzureOpenAIEmbeddings",
+        ],
+    },
+}