community: update documentation and model IDs for FriendliAI provider (#28984)

### Description  

- In the example, remove `llama-2-13b-chat`,
`mixtral-8x7b-instruct-v0-1`.
- Fix llm friendli streaming implementation.
- Update examples in documentation and remove duplicates.

### Issue  
N/A  

### Dependencies  
None  

### Twitter handle  
`@friendliai`

libs/community/langchain_community/chat_models/friendli.py

@@ -75,12 +75,12 @@ class ChatFriendli(BaseChatModel, BaseFriendli):
             from langchain_community.chat_models import FriendliChat
 
             chat = Friendli(
-                model="llama-2-13b-chat", friendli_token="YOUR FRIENDLI TOKEN"
+                model="meta-llama-3.1-8b-instruct", friendli_token="YOUR FRIENDLI TOKEN"
             )
             chat.invoke("What is generative AI?")
     """
 
-    model: str = "llama-2-13b-chat"
+    model: str = "meta-llama-3.1-8b-instruct"
 
     @property
     def lc_secrets(self) -> Dict[str, str]:

libs/community/langchain_community/llms/friendli.py

@@ -16,14 +16,19 @@ from langchain_core.utils.utils import convert_to_secret_str
 from pydantic import Field, SecretStr
 
 
-def _stream_response_to_generation_chunk(stream_response: Any) -> GenerationChunk:
+def _stream_response_to_generation_chunk(
+    stream_response: Any,
+) -> GenerationChunk:
     """Convert a stream response to a generation chunk."""
-    if stream_response.event == "token_sampled":
-        return GenerationChunk(
-            text=stream_response.text,
-            generation_info={"token": str(stream_response.token)},
-        )
-    return GenerationChunk(text="")
+    if not stream_response.get("choices", None):
+        return GenerationChunk(text="")
+    return GenerationChunk(
+        text=stream_response.choices[0].text,
+        # generation_info=dict(
+        #     finish_reason=stream_response.choices[0].get("finish_reason", None),
+        #     logprobs=stream_response.choices[0].get("logprobs", None),
+        # ),
+    )
 
 
 class BaseFriendli(Serializable):
@@ -34,7 +39,7 @@ class BaseFriendli(Serializable):
     # Friendli Async client.
     async_client: Any = Field(default=None, exclude=True)
     # Model name to use.
-    model: str = "mixtral-8x7b-instruct-v0-1"
+    model: str = "meta-llama-3.1-8b-instruct"
     # Friendli personal access token to run as.
     friendli_token: Optional[SecretStr] = None
     # Friendli team ID to run as.
@@ -107,7 +112,7 @@ class Friendli(LLM, BaseFriendli):
             from langchain_community.llms import Friendli
 
             friendli = Friendli(
-                model="mixtral-8x7b-instruct-v0-1", friendli_token="YOUR FRIENDLI TOKEN"
+                model="meta-llama-3.1-8b-instruct", friendli_token="YOUR FRIENDLI TOKEN"
             )
     """
 

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

@@ -114,14 +114,14 @@ async def test_friendli_ainvoke(
 @pytest.mark.requires("friendli")
 def test_friendli_stream(mock_friendli_client: Mock, friendli_llm: Friendli) -> None:
     """Test stream with friendli."""
+    mock_choice_0 = Mock()
+    mock_choice_0.text = "Hello "
+    mock_choice_1 = Mock()
+    mock_choice_1.text = "langchain"
     mock_chunk_0 = Mock()
-    mock_chunk_0.event = "token_sampled"
-    mock_chunk_0.text = "Hello "
-    mock_chunk_0.token = 0
+    mock_chunk_0.choices = [mock_choice_0]
     mock_chunk_1 = Mock()
-    mock_chunk_1.event = "token_sampled"
-    mock_chunk_1.text = "Friendli"
-    mock_chunk_1.token = 1
+    mock_chunk_1.choices = [mock_choice_1]
     mock_stream = MagicMock()
     mock_chunks = [mock_chunk_0, mock_chunk_1]
     mock_stream.__iter__.return_value = mock_chunks
@@ -129,7 +129,7 @@ def test_friendli_stream(mock_friendli_client: Mock, friendli_llm: Friendli) ->
     mock_friendli_client.completions.create.return_value = mock_stream
     stream = friendli_llm.stream("Hello langchain")
     for i, chunk in enumerate(stream):
-        assert chunk == mock_chunks[i].text
+        assert chunk == mock_chunks[i].choices[0].text
 
     mock_friendli_client.completions.create.assert_called_once_with(
         model=friendli_llm.model,
@@ -149,22 +149,22 @@ async def test_friendli_astream(
     mock_friendli_async_client: AsyncMock, friendli_llm: Friendli
 ) -> None:
     """Test async stream with friendli."""
+    mock_choice_0 = Mock()
+    mock_choice_0.text = "Hello "
+    mock_choice_1 = Mock()
+    mock_choice_1.text = "langchain"
     mock_chunk_0 = Mock()
-    mock_chunk_0.event = "token_sampled"
-    mock_chunk_0.text = "Hello "
-    mock_chunk_0.token = 0
+    mock_chunk_0.choices = [mock_choice_0]
     mock_chunk_1 = Mock()
-    mock_chunk_1.event = "token_sampled"
-    mock_chunk_1.text = "Friendli"
-    mock_chunk_1.token = 1
+    mock_chunk_1.choices = [mock_choice_1]
     mock_stream = AsyncMock()
     mock_chunks = [mock_chunk_0, mock_chunk_1]
-    mock_stream.__aiter__.return_value = mock_chunks
+    mock_stream.__aiter__.return_value = iter(mock_chunks)
 
     mock_friendli_async_client.completions.create.return_value = mock_stream
     stream = friendli_llm.astream("Hello langchain")
     async for i, chunk in aenumerate(stream):
-        assert chunk == mock_chunks[i].text
+        assert chunk == mock_chunks[i].choices[0].text
 
     mock_friendli_async_client.completions.create.assert_awaited_once_with(
         model=friendli_llm.model,