langchain-ai / langchain-cohere Goto Github PK

Hello. I need help.

I need google_search_tool only as in the context of the message there are launches like "search", "find". How to call agent_executor with such questions, everything works fine, but how to ask a normal question without search, then I get the error directly_answer is not a valid tool, try one of [google_search_tool]., latencies are higher than 6 times, but the model gave a normal response.

class Chat:
 
    def __init__(self):
        model = ChatCohere()
        agent = create_cohere_react_agent(llm=model, tools=[], prompt=prompt)
        self.agent_executor = AgentExecutor(agent=agent, tools=[google_search_tool], verbose=True)
        self.chain = prompt | model
 
    def prediction(self, question, history):
        return self.agent_executor.astream({
            "question": question,
            "chat_history": history
        }})
 
    def invoke(self, question):
        return self.chain.invoke({"question": question})
 
 
cohere = Chat()

from langchain_google_community import GoogleSearchAPIWrapper
from langchain_core.tools import Tool

from config import GOOGLE_API_KEY

search = GoogleSearchAPIWrapper(google_api_key=GOOGLE_API_KEY)


google_search_tool = Tool(
    name="google_search_tool",
    description="Search Google for recent results.",
    func=search.run
)

Below the example to reproduce it. Happens the same with command-r or command-r-plus. Also it happens regardless of the structure of the tool.

from langchain_cohere import ChatCohere
from langchain_core.messages import HumanMessage, AIMessage, AnyMessage, SystemMessage
from pydantic.v1 import BaseModel


class SelectIceCream(BaseModel):
    """Use this function only when the user wants to buy ice cream"""

    ice_cream_type: str


cohere_api_key = "<...>"
llm = ChatCohere(model="command-r", temperature=0.6, cohere_api_key=cohere_api_key)
llm_with_tools = llm.bind_tools([SelectIceCream])


chat = [
    SystemMessage(content="You are a helpful assistant"),
    HumanMessage(content="Where is the city of Paris? Be concise."),
]


def main():
    response = llm.invoke(chat)
    print(f"Without tools:\t{response=} \n\n---\n\n")

    response = llm_with_tools.invoke(chat)
    print(f"With tools:\t{response=}")


main()

Outputs:

Without tools:  response=AIMessage(content='The city of Paris is in the north of France, on the River Seine.', additional_kwargs={'documents': None, 'citations': None, 'search_results': None, 'search_queries': None, 'is_search_required': None, 'generation_id': '86f2f8b4-a99d-4246-8d72-6b7843ff55d0'}, response_metadata={'documents': None, 'citations': None, 'search_results': None, 'search_queries': None, 'is_search_required': None, 'generation_id': '86f2f8b4-a99d-4246-8d72-6b7843ff55d0'}, id='run-ede8f9b2-9e9a-4086-af6a-0274c02c315e-0') 

---


With tools:     response=AIMessage(content='', additional_kwargs={'documents': None, 'citations': None, 'search_results': None, 'search_queries': None, 'is_search_required': None, 'generation_id': 'f4ed5db6-7ea7-4495-a06a-f364cd53b417'}, response_metadata={'documents': None, 'citations': None, 'search_results': None, 'search_queries': None, 'is_search_required': None, 'generation_id': 'f4ed5db6-7ea7-4495-a06a-f364cd53b417'}, id='run-ef87eae8-7408-4f36-9f07-778ebd49f612-0')

So without tools gives a normal reply but the response content of the llm bound to tools is always an empty string.

I'm using langchain-core==0.1.16 and langchain-cohere==0.1.3

According to the Document in LangChain - Cohere
The usage for LLM is

from langchain_cohere import Cohere

It is same as stated in the source code's comment.
However, if use accordingly, the error will be

# ImportError: cannot import name 'Cohere' from 'langchain_cohere' (/Users/user-cloaked/miniforge3/envs/langchain/lib/python3.11/site-packages/langchain_cohere/__init__.py)

According to the structure of the source code, the actual working method seems to be:

from langchain_cohere.llms import Cohere

Assume that this usage is the actual way.
There is a bug inside the llms.py

If calling the Cohere according to manual, should be something like this

model = Cohere(model="command", max_tokens=256, temperature=0.75)

However, it will generate error like this

# TypeError: langchain_cohere.llms.completion_with_retry() got multiple values for keyword argument 'model'

Because when running completion_with_retry(), self.model is provided instead of model
i.e. it didn't allow to specify any model in other words
I am not sure if it is intended to do so or just a copy and paste error, the async also have to same problem by reading the source code without running.

So the final working writing at the moment is

from langchain_cohere.llms import Cohere
model = Cohere(max_tokens=256, temperature=0.75)
# ...

llm.with_structured_output method does not support Pydantic models with enum fields. Getting following error.

    return master_prompts | self.get_default_model().with_structured_output(MessageRouter)
                            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "<path>.venv/lib/python3.11/site-packages/langchain_core/_api/beta_decorator.py", line 109, in warning_emitting_wrapper
    return wrapped(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^^^^
  File "<path>.venv/lib/python3.11/site-packages/langchain_cohere/chat_models.py", line 210, in with_structured_output
    llm = self.bind_tools([schema], **kwargs)
          ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "<path>.venv/lib/python3.11/site-packages/langchain_cohere/chat_models.py", line 189, in bind_tools
    formatted_tools = _format_to_cohere_tools(tools)
                      ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "<path>.venv/lib/python3.11/site-packages/langchain_cohere/cohere_agent.py", line 61, in _format_to_cohere_tools
    return [_convert_to_cohere_tool(tool) for tool in tools]
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "<path>.venv/lib/python3.11/site-packages/langchain_cohere/cohere_agent.py", line 61, in <listcomp>
    return [_convert_to_cohere_tool(tool) for tool in tools]
            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "<path>.venv/lib/python3.11/site-packages/langchain_cohere/cohere_agent.py", line 155, in _convert_to_cohere_tool
    parameter_definitions={
                          ^
  File "<path>.venv/lib/python3.11/site-packages/langchain_cohere/cohere_agent.py", line 156, in <dictcomp>
    param_name: ToolParameterDefinitionsValue(
                ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "<path>.venv/lib/python3.11/site-packages/pydantic/v1/main.py", line 341, in __init__
    raise validation_error
pydantic.v1.error_wrappers.ValidationError: 1 validation error for ToolParameterDefinitionsValue
type
  none is not an allowed value (type=type_error.none.not_allowed)

Tried this code from the documentation example

from langchain_cohere.embeddings import CohereEmbeddings

cohere_embeddings = CohereEmbeddings(model="embed-english-light-v3.0")
text = "This is a test document."

query_result = cohere_embeddings.embed_query(text)
print(query_result)

doc_result = cohere_embeddings.embed_documents([text])
print(doc_result)

And got the following error

Traceback (most recent call last):
File "", line 6, in
File "/home/kolyan288/anaconda3/envs/Global/lib/python3.10/site-packages/langchain_cohere/embeddings.py", line 173, in embed_query
return self.embed([text], input_type="search_query")[0]
File "/home/kolyan288/anaconda3/envs/Global/lib/python3.10/site-packages/langchain_cohere/embeddings.py", line 118, in embed
embeddings = self.embed_with_retry(
File "/home/kolyan288/anaconda3/envs/Global/lib/python3.10/site-packages/langchain_cohere/embeddings.py", line 100, in embed_with_retry
return _embed_with_retry(**kwargs)
File "/home/kolyan288/anaconda3/envs/Global/lib/python3.10/site-packages/tenacity/init.py", line 330, in wrapped_f
return self(f, *args, **kw)
File "/home/kolyan288/anaconda3/envs/Global/lib/python3.10/site-packages/tenacity/init.py", line 467, in call
do = self.iter(retry_state=retry_state)
File "/home/kolyan288/anaconda3/envs/Global/lib/python3.10/site-packages/tenacity/init.py", line 368, in iter
result = action(retry_state)
File "/home/kolyan288/anaconda3/envs/Global/lib/python3.10/site-packages/tenacity/init.py", line 410, in exc_check
raise retry_exc.reraise()
File "/home/kolyan288/anaconda3/envs/Global/lib/python3.10/site-packages/tenacity/init.py", line 183, in reraise
raise self.last_attempt.result()
File "/home/kolyan288/anaconda3/envs/Global/lib/python3.10/concurrent/futures/_base.py", line 451, in result
return self.__get_result()
File "/home/kolyan288/anaconda3/envs/Global/lib/python3.10/concurrent/futures/_base.py", line 403, in __get_result
raise self._exception
File "/home/kolyan288/anaconda3/envs/Global/lib/python3.10/site-packages/tenacity/init.py", line 470, in call
result = fn(*args, **kwargs)
File "/home/kolyan288/anaconda3/envs/Global/lib/python3.10/site-packages/langchain_cohere/embeddings.py", line 98, in _embed_with_retry
return self.client.embed(**kwargs)
File "/home/kolyan288/anaconda3/envs/Global/lib/python3.10/site-packages/cohere/client.py", line 137, in embed
responses = [
File "/home/kolyan288/anaconda3/envs/Global/lib/python3.10/site-packages/cohere/client.py", line 137, in
responses = [
File "/home/kolyan288/anaconda3/envs/Global/lib/python3.10/concurrent/futures/_base.py", line 621, in result_iterator
yield _result_or_cancel(fs.pop())
File "/home/kolyan288/anaconda3/envs/Global/lib/python3.10/concurrent/futures/_base.py", line 319, in _result_or_cancel
return fut.result(timeout)
File "/home/kolyan288/anaconda3/envs/Global/lib/python3.10/concurrent/futures/_base.py", line 458, in result
return self.__get_result()
File "/home/kolyan288/anaconda3/envs/Global/lib/python3.10/concurrent/futures/_base.py", line 403, in _get_result
raise self.exception
File "/home/kolyan288/anaconda3/envs/Global/lib/python3.10/concurrent/futures/thread.py", line 58, in run
result = self.fn(*self.args, **self.kwargs)
File "/home/kolyan288/anaconda3/envs/Global/lib/python3.10/site-packages/cohere/client.py", line 140, in
lambda text_batch: BaseCohere.embed(
File "/home/kolyan288/anaconda3/envs/Global/lib/python3.10/site-packages/cohere/base_client.py", line 1446, in embed
typing.cast(typing.Any, construct_type(type=typing.Any, object=_response.json())) # type: ignore
File "/home/kolyan288/anaconda3/envs/Global/lib/python3.10/site-packages/httpx/_models.py", line 764, in json
return jsonlib.loads(self.content, **kwargs)
File "/home/kolyan288/anaconda3/envs/Global/lib/python3.10/json/init.py", line 346, in loads
return _default_decoder.decode(s)
File "/home/kolyan288/anaconda3/envs/Global/lib/python3.10/json/decoder.py", line 337, in decode
obj, end = self.raw_decode(s, idx=_w(s, 0).end())
File "/home/kolyan288/anaconda3/envs/Global/lib/python3.10/json/decoder.py", line 355, in raw_decode
raise JSONDecodeError("Expecting value", s, err.value) from None
json.decoder.JSONDecodeError: Expecting value: line 1 column 1 (char 0)

My dependencies

langchain==0.2.1
langchain-chroma==0.1.1
langchain-cli==0.0.24
langchain-cohere==0.1.8
langchain-community==0.2.1
langchain-core==0.2.3
langchain-experimental==0.0.59
langchain-huggingface==0.0.1
langchain-openai==0.1.8
langchain-text-splitters==0.2.0

platform: linux mint 21.1

python version: 3.10.14

Using a LangChain ConversationalRetrievalChain, we fail on a call to tokenize because of a missing model parameter:

> Entering new ConversationalRetrievalChain chain...
INFO:httpx:HTTP Request: POST https://api.cohere.ai/v1/embed "HTTP/1.1 200 OK"
INFO:httpx:HTTP Request: POST http://localhost:6333/collections/my.collection/points/search "HTTP/1.1 200 OK"
ERROR:root:Exception raised by task = <Task finished name='Task-522' coro=<Chain.ainvoke() done, defined at /home/zachary/.cache/pypoetry/virtualenvs/MY_VENV/lib/python3.10/site-packages/langchain/chains/base.py:170> exception=TypeError("Client.tokenize() missing 1 required keyword-only argument: 'model'")>
Traceback (most recent call last):
  File "/home/zachary/PycharmProjects/my_project/somewhere/in/my/code/utils.py", line 108, in _handle_task_result
    task.result()
  File "/home/zachary/.cache/pypoetry/virtualenvs/MY_VENV/lib/python3.10/site-packages/langchain/chains/base.py", line 212, in ainvoke
    raise e
  File "/home/zachary/.cache/pypoetry/virtualenvs/MY_VENV/lib/python3.10/site-packages/langchain/chains/base.py", line 203, in ainvoke
    await self._acall(inputs, run_manager=run_manager)
  File "/home/zachary/.cache/pypoetry/virtualenvs/MY_VENV/lib/python3.10/site-packages/langchain/chains/conversational_retrieval/base.py", line 207, in _acall
    docs = await self._aget_docs(new_question, inputs, run_manager=_run_manager)
  File "/home/zachary/.cache/pypoetry/virtualenvs/MY_VENV/lib/python3.10/site-packages/langchain/chains/conversational_retrieval/base.py", line 333, in _aget_docs
    return self._reduce_tokens_below_limit(docs)
  File "/home/zachary/.cache/pypoetry/virtualenvs/MY_VENV/lib/python3.10/site-packages/langchain/chains/conversational_retrieval/base.py", line 298, in _reduce_tokens_below_limit
    tokens = [
  File "/home/zachary/.cache/pypoetry/virtualenvs/MY_VENV/lib/python3.10/site-packages/langchain/chains/conversational_retrieval/base.py", line 299, in <listcomp>
    self.combine_docs_chain.llm_chain._get_num_tokens(doc.page_content)
  File "/home/zachary/.cache/pypoetry/virtualenvs/MY_VENV/lib/python3.10/site-packages/langchain/chains/llm.py", line 384, in _get_num_tokens
    return _get_language_model(self.llm).get_num_tokens(text)
  File "/home/zachary/.cache/pypoetry/virtualenvs/MY_VENV/lib/python3.10/site-packages/langchain_cohere/chat_models.py", line 366, in get_num_tokens
    return len(self.client.tokenize(text=text).tokens)
TypeError: Client.tokenize() missing 1 required keyword-only argument: 'model'

Background

This PR made the model: str parameter to tokenize() mandatory, and is included starting in cohere-python v5.2.0.

I was using cohere-python v5.2.5, langchain-core v0.1.42, langchain-community v0.0.32.

langchain-cohere is calling the function without the argument (link), which then throws an error:

    def get_num_tokens(self, text: str) -> int:
        """Calculate number of tokens."""
        return len(self.client.tokenize(text=text).tokens)

Suggested Fix

The Cohere documentation still says that this argument is optional, even though it isn't. If it is indeed mandatory, then the fix is probably to choose a reasonable default while allowing the user to override.

Thanks, I'm happy to provide more info as needed.

I was trying to run an example code, on Cohere's website at https://docs.cohere.com/docs/cohere-and-langchain

##This is the code

from pprint import pprint

from langchain_cohere import ChatCohere, CohereRagRetriever
from langchain_core.documents import Document

user_query = "Does LangChain support Cohere RAG?"
llm = ChatCohere()
rag = CohereRagRetriever(llm=llm, connectors=[])
docs = rag.get_relevant_documents(
user_query,
documents=[
Document(page_content="LangChain supports Cohere RAG!"),
Document(page_content="The sky is blue!"),
],
)
answer = docs.pop()

pprint("Relevant documents:")
pprint(docs)

pprint(f"Question: {user_query}")
pprint("Answer:")
pprint(answer.page_content)
pprint(answer.metadata["citations"])

##And this is the error I am getting

Traceback (most recent call last):
File "c:\Users\samar\Desktop\Assistant\COHERETRYOUTSS.py", line 57, in
docs = rag.get_relevant_documents(
File "C:\Users\samar\Desktop\Assistant\RA\lib\site-packages\langchain_core\retrievers.py", line 321, in get_relevant_documents
raise e
File "C:\Users\samar\Desktop\Assistant\RA\lib\site-packages\langchain_core\retrievers.py", line 314, in get_relevant_documents
result = self._get_relevant_documents(
File "C:\Users\samar\Desktop\Assistant\RA\lib\site-packages\langchain_cohere\rag_retrievers.py", line 82, in _get_relevant_documents
return _get_docs(res)
File "C:\Users\samar\Desktop\Assistant\RA\lib\site-packages\langchain_cohere\rag_retrievers.py", line 38, in _get_docs
"token_count": response.generation_info["token_count"],
KeyError: 'token_count'
PS C:\Users\samar\Desktop\Assistant>

Please help. Thank you

ImportError Traceback (most recent call last)
in <cell line: 1>()
----> 1 from langchain_cohere import ChatCohere
2 from langchain_core.messages import HumanMessage
3 from langgraph.graph import END, MessageGraph
4
5 model = ChatCohere(temperature=0)

2 frames
/usr/local/lib/python3.10/dist-packages/langchain_cohere/cohere_agent.py in
2 from typing import Any, Callable, Dict, List, Sequence, Tuple, Type, Union
3
----> 4 from cohere.types import (
5 ChatRequestToolResultsItem,
6 Tool,

ImportError: cannot import name 'ChatRequestToolResultsItem' from 'cohere.types' (/usr/local/lib/python3.10/dist-packages/cohere/types/init.py)

NOTE: If your import is failing due to a missing package, you can
manually install dependencies using either !pip or !apt.

from langchain_cohere import CohereEmbeddings
embeddings = CohereEmbeddings(model="embed-multilingual-v3.0")
emb = embeddings.embed_query("Hello, world!")

This script works in versions before 0.2.0. In 0.2.0 it produces the error:

---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
Cell In[4], [line 4](vscode-notebook-cell:?execution_count=4&line=4)
      [1](vscode-notebook-cell:?execution_count=4&line=1) from langchain_cohere import CohereEmbeddings
      [2](vscode-notebook-cell:?execution_count=4&line=2) embeddings = CohereEmbeddings(model="embed-multilingual-v3.0")
----> [4](vscode-notebook-cell:?execution_count=4&line=4) emb = embeddings.embed_query("Hello, world!")

File ~/.local/lib/python3.10/site-packages/langchain_cohere/embeddings.py:193, in CohereEmbeddings.embed_query(self, text)
    [184](https://vscode-remote+ssh-002dremote-002bthyme.vscode-resource.vscode-cdn.net/home/gavin/jupyter/~/.local/lib/python3.10/site-packages/langchain_cohere/embeddings.py:184) def embed_query(self, text: str) -> List[float]:
    [185](https://vscode-remote+ssh-002dremote-002bthyme.vscode-resource.vscode-cdn.net/home/gavin/jupyter/~/.local/lib/python3.10/site-packages/langchain_cohere/embeddings.py:185)     """Call out to Cohere's embedding endpoint.
    [186](https://vscode-remote+ssh-002dremote-002bthyme.vscode-resource.vscode-cdn.net/home/gavin/jupyter/~/.local/lib/python3.10/site-packages/langchain_cohere/embeddings.py:186) 
    [187](https://vscode-remote+ssh-002dremote-002bthyme.vscode-resource.vscode-cdn.net/home/gavin/jupyter/~/.local/lib/python3.10/site-packages/langchain_cohere/embeddings.py:187)     Args:
   (...)
    [191](https://vscode-remote+ssh-002dremote-002bthyme.vscode-resource.vscode-cdn.net/home/gavin/jupyter/~/.local/lib/python3.10/site-packages/langchain_cohere/embeddings.py:191)         Embeddings for the text.
    [192](https://vscode-remote+ssh-002dremote-002bthyme.vscode-resource.vscode-cdn.net/home/gavin/jupyter/~/.local/lib/python3.10/site-packages/langchain_cohere/embeddings.py:192)     """
--> [193](https://vscode-remote+ssh-002dremote-002bthyme.vscode-resource.vscode-cdn.net/home/gavin/jupyter/~/.local/lib/python3.10/site-packages/langchain_cohere/embeddings.py:193)     return self.embed([text], input_type="search_query")[0]

File ~/.local/lib/python3.10/site-packages/langchain_cohere/embeddings.py:143, in CohereEmbeddings.embed(self, texts, input_type)
    [130](https://vscode-remote+ssh-002dremote-002bthyme.vscode-resource.vscode-cdn.net/home/gavin/jupyter/~/.local/lib/python3.10/site-packages/langchain_cohere/embeddings.py:130) def embed(
    [131](https://vscode-remote+ssh-002dremote-002bthyme.vscode-resource.vscode-cdn.net/home/gavin/jupyter/~/.local/lib/python3.10/site-packages/langchain_cohere/embeddings.py:131)     self,
    [132](https://vscode-remote+ssh-002dremote-002bthyme.vscode-resource.vscode-cdn.net/home/gavin/jupyter/~/.local/lib/python3.10/site-packages/langchain_cohere/embeddings.py:132)     texts: List[str],
    [133](https://vscode-remote+ssh-002dremote-002bthyme.vscode-resource.vscode-cdn.net/home/gavin/jupyter/~/.local/lib/python3.10/site-packages/langchain_cohere/embeddings.py:133)     *,
    [134](https://vscode-remote+ssh-002dremote-002bthyme.vscode-resource.vscode-cdn.net/home/gavin/jupyter/~/.local/lib/python3.10/site-packages/langchain_cohere/embeddings.py:134)     input_type: typing.Optional[cohere.EmbedInputType] = None,
    [135](https://vscode-remote+ssh-002dremote-002bthyme.vscode-resource.vscode-cdn.net/home/gavin/jupyter/~/.local/lib/python3.10/site-packages/langchain_cohere/embeddings.py:135) ) -> List[List[float]]:
    [136](https://vscode-remote+ssh-002dremote-002bthyme.vscode-resource.vscode-cdn.net/home/gavin/jupyter/~/.local/lib/python3.10/site-packages/langchain_cohere/embeddings.py:136)     embeddings = self.embed_with_retry(
    [137](https://vscode-remote+ssh-002dremote-002bthyme.vscode-resource.vscode-cdn.net/home/gavin/jupyter/~/.local/lib/python3.10/site-packages/langchain_cohere/embeddings.py:137)         model=self.model,
    [138](https://vscode-remote+ssh-002dremote-002bthyme.vscode-resource.vscode-cdn.net/home/gavin/jupyter/~/.local/lib/python3.10/site-packages/langchain_cohere/embeddings.py:138)         texts=texts,
   (...)
    [141](https://vscode-remote+ssh-002dremote-002bthyme.vscode-resource.vscode-cdn.net/home/gavin/jupyter/~/.local/lib/python3.10/site-packages/langchain_cohere/embeddings.py:141)         embedding_types=self.embedding_types,
    [142](https://vscode-remote+ssh-002dremote-002bthyme.vscode-resource.vscode-cdn.net/home/gavin/jupyter/~/.local/lib/python3.10/site-packages/langchain_cohere/embeddings.py:142)     ).embeddings
--> [143](https://vscode-remote+ssh-002dremote-002bthyme.vscode-resource.vscode-cdn.net/home/gavin/jupyter/~/.local/lib/python3.10/site-packages/langchain_cohere/embeddings.py:143)     return [list(map(float, e)) for e in embeddings]

File ~/.local/lib/python3.10/site-packages/langchain_cohere/embeddings.py:143, in <listcomp>(.0)
    [130](https://vscode-remote+ssh-002dremote-002bthyme.vscode-resource.vscode-cdn.net/home/gavin/jupyter/~/.local/lib/python3.10/site-packages/langchain_cohere/embeddings.py:130) def embed(
    [131](https://vscode-remote+ssh-002dremote-002bthyme.vscode-resource.vscode-cdn.net/home/gavin/jupyter/~/.local/lib/python3.10/site-packages/langchain_cohere/embeddings.py:131)     self,
    [132](https://vscode-remote+ssh-002dremote-002bthyme.vscode-resource.vscode-cdn.net/home/gavin/jupyter/~/.local/lib/python3.10/site-packages/langchain_cohere/embeddings.py:132)     texts: List[str],
    [133](https://vscode-remote+ssh-002dremote-002bthyme.vscode-resource.vscode-cdn.net/home/gavin/jupyter/~/.local/lib/python3.10/site-packages/langchain_cohere/embeddings.py:133)     *,
    [134](https://vscode-remote+ssh-002dremote-002bthyme.vscode-resource.vscode-cdn.net/home/gavin/jupyter/~/.local/lib/python3.10/site-packages/langchain_cohere/embeddings.py:134)     input_type: typing.Optional[cohere.EmbedInputType] = None,
    [135](https://vscode-remote+ssh-002dremote-002bthyme.vscode-resource.vscode-cdn.net/home/gavin/jupyter/~/.local/lib/python3.10/site-packages/langchain_cohere/embeddings.py:135) ) -> List[List[float]]:
    [136](https://vscode-remote+ssh-002dremote-002bthyme.vscode-resource.vscode-cdn.net/home/gavin/jupyter/~/.local/lib/python3.10/site-packages/langchain_cohere/embeddings.py:136)     embeddings = self.embed_with_retry(
    [137](https://vscode-remote+ssh-002dremote-002bthyme.vscode-resource.vscode-cdn.net/home/gavin/jupyter/~/.local/lib/python3.10/site-packages/langchain_cohere/embeddings.py:137)         model=self.model,
    [138](https://vscode-remote+ssh-002dremote-002bthyme.vscode-resource.vscode-cdn.net/home/gavin/jupyter/~/.local/lib/python3.10/site-packages/langchain_cohere/embeddings.py:138)         texts=texts,
   (...)
    [141](https://vscode-remote+ssh-002dremote-002bthyme.vscode-resource.vscode-cdn.net/home/gavin/jupyter/~/.local/lib/python3.10/site-packages/langchain_cohere/embeddings.py:141)         embedding_types=self.embedding_types,
    [142](https://vscode-remote+ssh-002dremote-002bthyme.vscode-resource.vscode-cdn.net/home/gavin/jupyter/~/.local/lib/python3.10/site-packages/langchain_cohere/embeddings.py:142)     ).embeddings
--> [143](https://vscode-remote+ssh-002dremote-002bthyme.vscode-resource.vscode-cdn.net/home/gavin/jupyter/~/.local/lib/python3.10/site-packages/langchain_cohere/embeddings.py:143)     return [list(map(float, e)) for e in embeddings]

ValueError: could not convert string to float: 'float_'

I was trying to use this library to implement Cohere RAG, but have encountered two errors.

Error when no relevant documents are found

One occurs when no relevant documents are found:

rag = CohereRagRetriever(llm=ChatCohere(model='command-r-plus'))
docs = rag.get_relevant_documents(
    'Give me a list of leap years that have not yet happened.',
    documents=[Document('These documents'), Document('are not relevant')],
)

Gives the error:

  File ".../langchain_core/retrievers.py", line 321, in get_relevant_documents
    raise e
  File ".../langchain_core/retrievers.py", line 314, in get_relevant_documents
    result = self._get_relevant_documents(
  File ".../langchain_cohere/rag_retrievers.py", line 82, in _get_relevant_documents
    return _get_docs(res)
  File ".../langchain_cohere/rag_retrievers.py", line 23, in _get_docs
    and len(response.generation_info["documents"]) > 0
TypeError: object of type 'NoneType' has no len()

This line fails because documents is defaulted to None, not an empty list, and therefore doesn't support the len() method.

Changing to a simple boolean check of

    if (
        "documents" in response.generation_info
        and response.generation_info["documents"]
    ):

could be a way to fix this.

Error when documents are found

Running with relevant documents lets it find something and skip the above error:

rag = CohereRagRetriever(llm=ChatCohere(model='command-r-plus'))
docs = rag.get_relevant_documents(
    'Give me a list of leap years that have not yet happened.',
    documents=[Document('Past Leap Years: 2016, 2020, 2024'), Document('Upcoming Leap Years: 2028, 2032, 2036')],
)

but it gives a later error:

  File ".../langchain_core/retrievers.py", line 384, in aget_relevant_documents
    raise e
  File ".../langchain_core/retrievers.py", line 377, in aget_relevant_documents
    result = await self._aget_relevant_documents(
  File ".../langchain_cohere/rag_retrievers.py", line 102, in _aget_relevant_documents
    return _get_docs(res)
  File ".../langchain_cohere/rag_retrievers.py", line 38, in _get_docs
    "token_count": response.generation_info["token_count"],

This seems to be because the response returned by the ChatModel now stores token count in response.meta rather than in response.token_count so it's not set here and then crashes in rag_retrievers at the line specified above.

This was tested all tested with langchain-cohere 0.1.3 (latest) and langchain-core versions 0.1.43 and 0.1.44.

In the current released version, your asynchronous operation _agenerate() is a blocking call.

However, someone has fixed this already - 038cded

But you have yet to release a new package with the fix.

When will the new package be released so that ChatCohere can be used asynchronously?