community:qianfan endpoint support init params & remove useless params definietion (#15381)

- **Description:**
- support custom kwargs in object initialization. For instantance, QPS
differs from multiple object(chat/completion/embedding with diverse
models), for which global env is not a good choice for configuration.
  - **Issue:** no
  - **Dependencies:** no
  - **Twitter handle:** no

@baskaryan PTAL

libs/community/tests/integration_tests/chat_models/test_qianfan_endpoint.py

@@ -217,3 +217,18 @@ def test_functions_call() -> None:
     chain = prompt | chat.bind(functions=_FUNCTIONS)
     resp = chain.invoke({})
     assert isinstance(resp, AIMessage)
+
+
+def test_rate_limit() -> None:
+    chat = QianfanChatEndpoint(model="ERNIE-Bot", init_kwargs={"query_per_second": 2})
+    assert chat.client._client._rate_limiter._sync_limiter._query_per_second == 2
+    responses = chat.batch(
+        [
+            [HumanMessage(content="Hello")],
+            [HumanMessage(content="who are you")],
+            [HumanMessage(content="what is baidu")],
+        ]
+    )
+    for res in responses:
+        assert isinstance(res, BaseMessage)
+        assert isinstance(res.content, str)

libs/community/tests/integration_tests/embeddings/test_qianfan_endpoint.py

@@ -25,3 +25,15 @@ def test_model() -> None:
     embedding = QianfanEmbeddingsEndpoint(model="Embedding-V1")
     output = embedding.embed_documents(documents)
     assert len(output) == 2
+
+
+def test_rate_limit() -> None:
+    llm = QianfanEmbeddingsEndpoint(
+        model="Embedding-V1", init_kwargs={"query_per_second": 2}
+    )
+    assert llm.client._client._rate_limiter._sync_limiter._query_per_second == 2
+    documents = ["foo", "bar"]
+    output = llm.embed_documents(documents)
+    assert len(output) == 2
+    assert len(output[0]) == 384
+    assert len(output[1]) == 384

libs/community/tests/integration_tests/llms/test_qianfan_endpoint.py

@@ -33,3 +33,11 @@ async def test_qianfan_aio() -> None:
 
     async for token in llm.astream("hi qianfan."):
         assert isinstance(token, str)
+
+
+def test_rate_limit() -> None:
+    llm = QianfanLLMEndpoint(model="ERNIE-Bot", init_kwargs={"query_per_second": 2})
+    assert llm.client._client._rate_limiter._sync_limiter._query_per_second == 2
+    output = llm.generate(["write a joke"])
+    assert isinstance(output, LLMResult)
+    assert isinstance(output.generations, list)