community(you): Integrate You.com conversational APIs (#23046)

You.com is releasing two new conversational APIs — Smart and Research. 

This PR:
- integrates those APIs with Langchain, as an LLM
- streaming is supported

If no one reviews your PR within a few days, please @-mention one of
baskaryan, efriis, eyurtsev, ccurme, vbarda, hwchase17.

libs/community/extended_testing_deps.txt

@@ -72,6 +72,7 @@ rspace_client>=2.5.0,<3
 scikit-learn>=1.2.2,<2
 simsimd>=4.3.1,<5
 sqlite-vss>=0.1.2,<0.2
+sseclient-py>=1.8.0,<2
 streamlit>=1.18.0,<2
 sympy>=1.12,<2
 telethon>=1.28.5,<2

libs/community/langchain_community/llms/__init__.py

@@ -640,6 +640,12 @@ def _import_yuan2() -> Type[BaseLLM]:
     return Yuan2
 
 
+def _import_you() -> Type[BaseLLM]:
+    from langchain_community.llms.you import You
+
+    return You
+
+
 def _import_volcengine_maas() -> Type[BaseLLM]:
     from langchain_community.llms.volcengine_maas import VolcEngineMaasLLM
 
@@ -847,6 +853,8 @@ def __getattr__(name: str) -> Any:
         return _import_yandex_gpt()
     elif name == "Yuan2":
         return _import_yuan2()
+    elif name == "You":
+        return _import_you()
     elif name == "VolcEngineMaasLLM":
         return _import_volcengine_maas()
     elif name == "type_to_cls_dict":
@@ -959,6 +967,7 @@ __all__ = [
     "Writer",
     "Xinference",
     "YandexGPT",
+    "You",
     "Yuan2",
 ]
 
@@ -1056,6 +1065,7 @@ def get_type_to_cls_dict() -> Dict[str, Callable[[], Type[BaseLLM]]]:
         "qianfan_endpoint": _import_baidu_qianfan_endpoint,
         "yandex_gpt": _import_yandex_gpt,
         "yuan2": _import_yuan2,
+        "you": _import_you,
         "VolcEngineMaasLLM": _import_volcengine_maas,
         "SparkLLM": _import_sparkllm,
     }

libs/community/langchain_community/llms/you.py

@@ -0,0 +1,140 @@
+import os
+from typing import Any, Dict, Generator, Iterator, List, Literal, Optional
+
+import requests
+from langchain_core.callbacks.manager import CallbackManagerForLLMRun
+from langchain_core.language_models.llms import LLM
+from langchain_core.outputs import GenerationChunk
+from langchain_core.pydantic_v1 import Field
+
+SMART_ENDPOINT = "https://chat-api.you.com/smart"
+RESEARCH_ENDPOINT = "https://chat-api.you.com/research"
+
+
+def _request(base_url: str, api_key: str, **kwargs: Any) -> Dict[str, Any]:
+    """
+    NOTE: This function can be replaced by a OpenAPI-generated Python SDK in the future,
+    for better input/output typing support.
+    """
+    headers = {"x-api-key": api_key}
+    response = requests.post(base_url, headers=headers, json=kwargs)
+    response.raise_for_status()
+    return response.json()
+
+
+def _request_stream(
+    base_url: str, api_key: str, **kwargs: Any
+) -> Generator[str, None, None]:
+    headers = {"x-api-key": api_key}
+    params = dict(**kwargs, stream=True)
+    response = requests.post(base_url, headers=headers, stream=True, json=params)
+    response.raise_for_status()
+
+    # Explicitly coercing the response to a generator to satisfy mypy
+    event_source = (bytestring for bytestring in response)
+
+    try:
+        import sseclient
+
+        client = sseclient.SSEClient(event_source)
+    except ImportError:
+        raise ImportError(
+            (
+                "Could not import `sseclient`. "
+                "Please install it with `pip install sseclient-py`."
+            )
+        )
+
+    for event in client.events():
+        if event.event in ("search_results", "done"):
+            pass
+        elif event.event == "token":
+            yield event.data
+        elif event.event == "error":
+            raise ValueError(f"Error in response: {event.data}")
+        else:
+            raise NotImplementedError(f"Unknown event type {event.event}")
+
+
+class You(LLM):
+    """Wrapper around You.com's conversational Smart and Research APIs.
+
+    Each API endpoint is designed to generate conversational
+    responses to a variety of query types, including inline citations
+    and web results when relevant.
+
+    Smart Endpoint:
+    - Quick, reliable answers for a variety of questions
+    - Cites the entire web page URL
+
+    Research Endpoint:
+    - In-depth answers with extensive citations for a variety of questions
+    - Cites the specific web page snippet relevant to the claim
+
+    To connect to the You.com api requires an API key which
+    you can get at https://api.you.com.
+
+    For more information, check out the documentations at
+    https://documentation.you.com/api-reference/.
+
+    Args:
+        endpoint: You.com conversational endpoints. Choose from "smart" or "research"
+        ydc_api_key: You.com API key, if `YDC_API_KEY` is not set in the environment
+    """
+
+    endpoint: Literal["smart", "research"] = Field(
+        "smart",
+        description=(
+            'You.com conversational endpoints. Choose from "smart" or "research"'
+        ),
+    )
+    ydc_api_key: Optional[str] = Field(
+        None,
+        description="You.com API key, if `YDC_API_KEY` is not set in the envrioment",
+    )
+
+    def _call(
+        self,
+        prompt: str,
+        stop: Optional[List[str]] = None,
+        run_manager: Optional[CallbackManagerForLLMRun] = None,
+        **kwargs: Any,
+    ) -> str:
+        if stop:
+            raise NotImplementedError(
+                "Stop words are not implemented for You.com endpoints."
+            )
+        params = {"query": prompt}
+        response = _request(self._request_endpoint, api_key=self._api_key, **params)
+        return response["answer"]
+
+    def _stream(
+        self,
+        prompt: str,
+        stop: Optional[List[str]] = None,
+        run_manager: Optional[CallbackManagerForLLMRun] = None,
+        **kwargs: Any,
+    ) -> Iterator[GenerationChunk]:
+        if stop:
+            raise NotImplementedError(
+                "Stop words are not implemented for You.com endpoints."
+            )
+        params = {"query": prompt}
+        for token in _request_stream(
+            self._request_endpoint, api_key=self._api_key, **params
+        ):
+            yield GenerationChunk(text=token)
+
+    @property
+    def _request_endpoint(self) -> str:
+        if self.endpoint == "smart":
+            return SMART_ENDPOINT
+        return RESEARCH_ENDPOINT
+
+    @property
+    def _api_key(self) -> str:
+        return self.ydc_api_key or os.environ["YDC_API_KEY"]
+
+    @property
+    def _llm_type(self) -> str:
+        return "you.com"

libs/community/tests/unit_tests/llms/test_imports.py

@@ -98,6 +98,7 @@ EXPECT_ALL = [
     "QianfanLLMEndpoint",
     "YandexGPT",
     "Yuan2",
+    "You",
     "VolcEngineMaasLLM",
     "WatsonxLLM",
     "SparkLLM",

libs/community/tests/unit_tests/llms/test_you.py

@@ -0,0 +1,37 @@
+import pytest
+import requests_mock
+
+
+@pytest.mark.parametrize("endpoint", ("smart", "research"))
+@pytest.mark.requires("sseclient")
+def test_invoke(
+    endpoint: str, requests_mock: requests_mock.Mocker, monkeypatch: pytest.MonkeyPatch
+) -> None:
+    from langchain_community.llms import You
+    from langchain_community.llms.you import RESEARCH_ENDPOINT, SMART_ENDPOINT
+
+    json = {
+        "answer": (
+            "A solar eclipse occurs when the Moon passes between the Sun and Earth, "
+            "casting a shadow on Earth and ..."
+        ),
+        "search_results": [
+            {
+                "url": "https://en.wikipedia.org/wiki/Solar_eclipse",
+                "name": "Solar eclipse - Wikipedia",
+                "snippet": (
+                    "A solar eclipse occurs when the Moon passes "
+                    "between Earth and the Sun, thereby obscuring the view of the Sun "
+                    "from a small part of Earth, totally or partially. "
+                ),
+            }
+        ],
+    }
+    request_endpoint = SMART_ENDPOINT if endpoint == "smart" else RESEARCH_ENDPOINT
+    requests_mock.post(request_endpoint, json=json)
+
+    monkeypatch.setenv("YDC_API_KEY", "...")
+
+    llm = You(endpoint=endpoint)
+    output = llm.invoke("What is a solar eclipse?")
+    assert output == json["answer"]