add model feat table (#10921)

.github/workflows/doc_lint.yml

@@ -19,4 +19,4 @@ jobs:
       run: |
         # We should not encourage imports directly from main init file
         # Expect for hub
-        git grep 'from langchain import' docs | grep -vE 'from langchain import (hub)' && exit 1 || exit 0
+        git grep 'from langchain import' docs/{extras,docs_skeleton,snippets} | grep -vE 'from langchain import (hub)' && exit 1 || exit 0

docs/_scripts/model_feat_table.py

@@ -0,0 +1,145 @@
+import os
+from pathlib import Path
+
+from langchain import chat_models, llms
+from langchain.chat_models.base import BaseChatModel, SimpleChatModel
+from langchain.llms.base import BaseLLM, LLM
+
+INTEGRATIONS_DIR = (
+    Path(os.path.abspath(__file__)).parents[1] / "extras" / "integrations"
+)
+LLM_IGNORE = ("FakeListLLM", "OpenAIChat", "PromptLayerOpenAIChat")
+LLM_FEAT_TABLE_CORRECTION = {
+    "TextGen": {"_astream": False, "_agenerate": False},
+    "Ollama": {
+        "_stream": False,
+    },
+    "PromptLayerOpenAI": {"batch_generate": False, "batch_agenerate": False},
+}
+CHAT_MODEL_IGNORE = ("FakeListChatModel", "HumanInputChatModel")
+CHAT_MODEL_FEAT_TABLE_CORRECTION = {
+    "ChatMLflowAIGateway": {"_agenerate": False},
+    "PromptLayerChatOpenAI": {"_stream": False, "_astream": False},
+    "ChatKonko": {"_astream": False, "_agenerate": False},
+}
+
+LLM_TEMPLATE = """\
+---
+sidebar_position: 0
+sidebar_class_name: hidden
+---
+
+# LLMs
+
+import DocCardList from "@theme/DocCardList";
+
+## Features (natively supported)
+All `LLM`s implement the LCEL `Runnable` interface, meaning they all expose functioning `invoke`, `ainvoke`, `batch`, `abatch`, `stream`, and `astream` methods.
+*That is, they all have functioning sync, async, streaming, and batch generation methods.*
+
+This table highlights specifically those integrations that **natively support** batching, streaming, and asynchronous generation (meaning these features are built into the 3rd-party integration).
+
+{table}
+
+<DocCardList />
+"""
+
+CHAT_MODEL_TEMPLATE = """\
+---
+sidebar_position: 1
+sidebar_class_name: hidden
+---
+
+# Chat models
+
+import DocCardList from "@theme/DocCardList";
+
+## Features (natively supported)
+All `ChatModel`s implement the LCEL `Runnable` interface, meaning they all expose functioning `invoke`, `ainvoke`, `stream`, and `astream` (and `batch`, `abatch`) methods.
+*That is, they all have functioning sync, async and streaming generation methods.*
+
+This table highlights specifically those integrations that **natively support** streaming and asynchronous generation (meaning these features are built into the 3rd-party integration).
+
+{table}
+
+<DocCardList />
+"""
+
+
+def get_llm_table():
+    llm_feat_table = {}
+    for cm in llms.__all__:
+        llm_feat_table[cm] = {}
+        cls = getattr(llms, cm)
+        if issubclass(cls, LLM):
+            for feat in ("_stream", "_astream", ("_acall", "_agenerate")):
+                if isinstance(feat, tuple):
+                    feat, name = feat
+                else:
+                    feat, name = feat, feat
+                llm_feat_table[cm][name] = getattr(cls, feat) != getattr(LLM, feat)
+        else:
+            for feat in [
+                "_stream",
+                "_astream",
+                ("_generate", "batch_generate"),
+                "_agenerate",
+                ("_agenerate", "batch_agenerate"),
+            ]:
+                if isinstance(feat, tuple):
+                    feat, name = feat
+                else:
+                    feat, name = feat, feat
+                llm_feat_table[cm][name] = getattr(cls, feat) != getattr(BaseLLM, feat)
+    final_feats = {
+        k: v
+        for k, v in {**llm_feat_table, **LLM_FEAT_TABLE_CORRECTION}.items()
+        if k not in LLM_IGNORE
+    }
+
+    header = [
+        "model",
+        "_agenerate",
+        "_stream",
+        "_astream",
+        "batch_generate",
+        "batch_agenerate",
+    ]
+    title = ["Model", "Generate", "Async generate", "Stream", "Async stream", "Batch", "Async batch"]
+    rows = [title, [":-"] + [":-:"] * (len(title) - 1)]
+    for llm, feats in sorted(final_feats.items()):
+        rows += [[llm, "✅"] + ["✅" if feats.get(h) else "❌" for h in header[1:]]]
+    return "\n".join(["|".join(row) for row in rows])
+
+
+def get_chat_model_table():
+    feat_table = {}
+    for cm in chat_models.__all__:
+        feat_table[cm] = {}
+        cls = getattr(chat_models, cm)
+        if issubclass(cls, SimpleChatModel):
+            comparison_cls = SimpleChatModel
+        else:
+            comparison_cls = BaseChatModel
+        for feat in ("_stream", "_astream", "_agenerate"):
+            feat_table[cm][feat] = getattr(cls, feat) != getattr(comparison_cls, feat)
+    final_feats = {
+        k: v
+        for k, v in {**feat_table, **CHAT_MODEL_FEAT_TABLE_CORRECTION}.items()
+        if k not in CHAT_MODEL_IGNORE
+    }
+    header = ["model", "_agenerate", "_stream", "_astream"]
+    title = ["Model", "Generate", "Async generate", "Stream", "Async stream"]
+    rows = [title, [":-"] + [":-:"] * (len(title) - 1)]
+    for llm, feats in sorted(final_feats.items()):
+        rows += [[llm, "✅"] + ["✅" if feats.get(h) else "❌" for h in header[1:]]]
+    return "\n".join(["|".join(row) for row in rows])
+
+
+if __name__ == "__main__":
+    llm_page = LLM_TEMPLATE.format(table=get_llm_table())
+    with open(INTEGRATIONS_DIR / "llms" / "index.mdx", "w") as f:
+        f.write(llm_page)
+    chat_model_page = CHAT_MODEL_TEMPLATE.format(table=get_chat_model_table())
+    with open(INTEGRATIONS_DIR / "chat" / "index.mdx", "w") as f:
+        f.write(chat_model_page)

docs/api_reference/create_api_rst.py

@@ -1,7 +1,6 @@
 """Script for auto-generating api_reference.rst."""
 import importlib
 import inspect
-import os
 import typing
 from pathlib import Path
 from typing import TypedDict, Sequence, List, Dict, Literal, Union, Optional

docs/docs_skeleton/sidebars.js

@@ -99,8 +99,8 @@ module.exports = {
       label: "Components",
       collapsible: false,
       items: [
-        { type: "category", label: "LLMs", collapsed: true, items: [{type:"autogenerated", dirName: "integrations/llms" }], link: {type: "generated-index", slug: "integrations/llms" }},
+        { type: "category", label: "LLMs", collapsed: true, items: [{type:"autogenerated", dirName: "integrations/llms" }], link: { type: 'doc', id: "integrations/llms/index"}},
-        { type: "category", label: "Chat models", collapsed: true, items: [{type:"autogenerated", dirName: "integrations/chat" }], link: {type: "generated-index", slug: "integrations/chat" }},
+        { type: "category", label: "Chat models", collapsed: true, items: [{type:"autogenerated", dirName: "integrations/chat" }], link: { type: 'doc', id: "integrations/chat/index"}},
         { type: "category", label: "Document loaders", collapsed: true, items: [{type:"autogenerated", dirName: "integrations/document_loaders" }], link: {type: "generated-index", slug: "integrations/document_loaders" }},
         { type: "category", label: "Document transformers", collapsed: true, items: [{type: "autogenerated", dirName: "integrations/document_transformers" }], link: {type: "generated-index", slug: "integrations/document_transformers" }},
         { type: "category", label: "Text embedding models", collapsed: true, items: [{type: "autogenerated", dirName: "integrations/text_embedding" }], link: {type: "generated-index", slug: "integrations/text_embedding" }},

docs/extras/integrations/chat/index.mdx

@@ -0,0 +1,36 @@
+---
+sidebar_position: 1
+sidebar_class_name: hidden
+---
+
+# Chat models
+
+import DocCardList from "@theme/DocCardList";
+
+## Features (natively supported)
+All `ChatModel`s implement the LCEL `Runnable` interface, meaning they all expose functioning `invoke`, `ainvoke`, `stream`, and `astream` (and `batch`, `abatch`) methods.
+*That is, they all have functioning sync, async and streaming generation methods.*
+
+This table highlights specifically those integrations that **natively support** streaming and asynchronous generation (meaning these features are built into the 3rd-party integration).
+
+Model|Generate|Async generate|Stream|Async stream
+:-|:-:|:-:|:-:|:-:
+AzureChatOpenAI|✅|✅|✅|✅
+BedrockChat|✅|❌|✅|❌
+ChatAnthropic|✅|✅|✅|✅
+ChatAnyscale|✅|✅|✅|✅
+ChatGooglePalm|✅|✅|❌|❌
+ChatJavelinAIGateway|✅|✅|❌|❌
+ChatKonko|✅|❌|❌|❌
+ChatLiteLLM|✅|✅|✅|✅
+ChatMLflowAIGateway|✅|❌|❌|❌
+ChatOllama|✅|❌|✅|❌
+ChatOpenAI|✅|✅|✅|✅
+ChatVertexAI|✅|❌|✅|❌
+ErnieBotChat|✅|❌|❌|❌
+JinaChat|✅|✅|✅|✅
+MiniMaxChat|✅|✅|❌|❌
+PromptLayerChatOpenAI|✅|❌|❌|❌
+QianfanChatEndpoint|✅|✅|✅|✅
+
+<DocCardList />

docs/extras/integrations/llms/index.mdx

@@ -0,0 +1,91 @@
+---
+sidebar_position: 0
+sidebar_class_name: hidden
+---
+
+# LLMs
+
+import DocCardList from "@theme/DocCardList";
+
+## Features (natively supported)
+All `LLM`s implement the LCEL `Runnable` interface, meaning they all expose functioning `invoke`, `ainvoke`, `batch`, `abatch`, `stream`, and `astream` methods.
+*That is, they all have functioning sync, async, streaming, and batch generation methods.*
+
+This table highlights specifically those integrations that **natively support** batching, streaming, and asynchronous generation (meaning these features are built into the 3rd-party integration).
+
+Model|Generate|Async generate|Stream|Async stream|Batch|Async batch
+:-|:-:|:-:|:-:|:-:|:-:|:-:
+AI21|✅|❌|❌|❌|❌|❌
+AlephAlpha|✅|❌|❌|❌|❌|❌
+AmazonAPIGateway|✅|❌|❌|❌|❌|❌
+Anthropic|✅|✅|✅|✅|❌|❌
+Anyscale|✅|❌|❌|❌|❌|❌
+Aviary|✅|❌|❌|❌|❌|❌
+AzureMLOnlineEndpoint|✅|❌|❌|❌|❌|❌
+AzureOpenAI|✅|✅|✅|✅|✅|✅
+Banana|✅|❌|❌|❌|❌|❌
+Baseten|✅|❌|❌|❌|❌|❌
+Beam|✅|❌|❌|❌|❌|❌
+Bedrock|✅|❌|✅|❌|❌|❌
+CTransformers|✅|✅|❌|❌|❌|❌
+CTranslate2|✅|❌|❌|❌|✅|❌
+CerebriumAI|✅|❌|❌|❌|❌|❌
+ChatGLM|✅|❌|❌|❌|❌|❌
+Clarifai|✅|❌|❌|❌|❌|❌
+Cohere|✅|✅|❌|❌|❌|❌
+Databricks|✅|❌|❌|❌|❌|❌
+DeepInfra|✅|❌|❌|❌|❌|❌
+DeepSparse|✅|❌|❌|❌|❌|❌
+EdenAI|✅|✅|❌|❌|❌|❌
+Fireworks|✅|✅|❌|❌|✅|✅
+FireworksChat|✅|✅|❌|❌|✅|✅
+ForefrontAI|✅|❌|❌|❌|❌|❌
+GPT4All|✅|❌|❌|❌|❌|❌
+GooglePalm|✅|❌|❌|❌|✅|❌
+GooseAI|✅|❌|❌|❌|❌|❌
+GradientLLM|✅|✅|❌|❌|❌|❌
+HuggingFaceEndpoint|✅|❌|❌|❌|❌|❌
+HuggingFaceHub|✅|❌|❌|❌|❌|❌
+HuggingFacePipeline|✅|❌|❌|❌|❌|❌
+HuggingFaceTextGenInference|✅|✅|✅|✅|❌|❌
+HumanInputLLM|✅|❌|❌|❌|❌|❌
+JavelinAIGateway|✅|✅|❌|❌|❌|❌
+KoboldApiLLM|✅|❌|❌|❌|❌|❌
+LlamaCpp|✅|❌|✅|❌|❌|❌
+ManifestWrapper|✅|❌|❌|❌|❌|❌
+Minimax|✅|❌|❌|❌|❌|❌
+MlflowAIGateway|✅|❌|❌|❌|❌|❌
+Modal|✅|❌|❌|❌|❌|❌
+MosaicML|✅|❌|❌|❌|❌|❌
+NIBittensorLLM|✅|❌|❌|❌|❌|❌
+NLPCloud|✅|❌|❌|❌|❌|❌
+Nebula|✅|❌|❌|❌|❌|❌
+OctoAIEndpoint|✅|❌|❌|❌|❌|❌
+Ollama|✅|❌|❌|❌|❌|❌
+OpaquePrompts|✅|❌|❌|❌|❌|❌
+OpenAI|✅|✅|✅|✅|✅|✅
+OpenLLM|✅|✅|❌|❌|❌|❌
+OpenLM|✅|✅|✅|✅|✅|✅
+Petals|✅|❌|❌|❌|❌|❌
+PipelineAI|✅|❌|❌|❌|❌|❌
+Predibase|✅|❌|❌|❌|❌|❌
+PredictionGuard|✅|❌|❌|❌|❌|❌
+PromptLayerOpenAI|✅|❌|❌|❌|❌|❌
+QianfanLLMEndpoint|✅|✅|✅|✅|❌|❌
+RWKV|✅|❌|❌|❌|❌|❌
+Replicate|✅|❌|✅|❌|❌|❌
+SagemakerEndpoint|✅|❌|❌|❌|❌|❌
+SelfHostedHuggingFaceLLM|✅|❌|❌|❌|❌|❌
+SelfHostedPipeline|✅|❌|❌|❌|❌|❌
+StochasticAI|✅|❌|❌|❌|❌|❌
+TextGen|✅|❌|❌|❌|❌|❌
+TitanTakeoff|✅|❌|✅|❌|❌|❌
+Tongyi|✅|❌|❌|❌|❌|❌
+VLLM|✅|❌|❌|❌|✅|❌
+VLLMOpenAI|✅|✅|✅|✅|✅|✅
+VertexAI|✅|✅|❌|❌|❌|❌
+VertexAIModelGarden|✅|✅|❌|❌|❌|❌
+Writer|✅|❌|❌|❌|❌|❌
+Xinference|✅|❌|❌|❌|❌|❌
+
+<DocCardList />

libs/langchain/langchain/chat_models/bedrock.py

@@ -1,7 +1,6 @@
-from typing import Any, AsyncIterator, Dict, Iterator, List, Optional
+from typing import Any, Dict, Iterator, List, Optional
 
 from langchain.callbacks.manager import (
-    AsyncCallbackManagerForLLMRun,
     CallbackManagerForLLMRun,
 )
 from langchain.chat_models.anthropic import convert_messages_to_prompt_anthropic
@@ -59,17 +58,6 @@ class BedrockChat(BaseChatModel, BedrockBase):
             delta = chunk.text
             yield ChatGenerationChunk(message=AIMessageChunk(content=delta))
 
-    def _astream(
-        self,
-        messages: List[BaseMessage],
-        stop: Optional[List[str]] = None,
-        run_manager: Optional[AsyncCallbackManagerForLLMRun] = None,
-        **kwargs: Any,
-    ) -> AsyncIterator[ChatGenerationChunk]:
-        raise NotImplementedError(
-            """Bedrock doesn't support async requests at the moment."""
-        )
-
     def _generate(
         self,
         messages: List[BaseMessage],
@@ -98,14 +86,3 @@ class BedrockChat(BaseChatModel, BedrockBase):
 
         message = AIMessage(content=completion)
         return ChatResult(generations=[ChatGeneration(message=message)])
-
-    async def _agenerate(
-        self,
-        messages: List[BaseMessage],
-        stop: Optional[List[str]] = None,
-        run_manager: Optional[AsyncCallbackManagerForLLMRun] = None,
-        **kwargs: Any,
-    ) -> ChatResult:
-        raise NotImplementedError(
-            """Bedrock doesn't support async stream requests at the moment."""
-        )