Chat on LangChain

1
from langchain_cohere import ChatCohere
2
from langchain_core.messages import AIMessage, HumanMessage
3
4
# Define the Cohere LLM
5
llm = ChatCohere(cohere_api_key="COHERE_API_KEY",              
6
                 model="command-r-plus-08-2024")
7
8
# Send a chat message without chat history
9
current_message = [HumanMessage(content="knock knock")]
10
print(llm.invoke(current_message))
11
12
# Send a chat message with chat history, note the last message is the current user message
13
current_message_and_history = [
14
    HumanMessage(content="knock knock"),
15
    AIMessage(content="Who's there?"),
16
    HumanMessage(content="Tank") ]
17
print(llm.invoke(current_message_and_history))

1
from langchain_cohere import ChatCohere
2
from langchain_cohere.react_multi_hop.agent import create_cohere_react_agent
3
from langchain.agents import AgentExecutor
4
from langchain_community.tools.tavily_search import TavilySearchResults
5
from langchain_core.prompts import ChatPromptTemplate
6
7
# Internet search tool - you can use any tool, and there are lots of community tools in LangChain.
8
# To use the Tavily tool you will need to set an API key in the TAVILY_API_KEY environment variable.
9
os.environ["TAVILY_API_KEY"] = "TAVILY_API_KEY"
10
internet_search = TavilySearchResults()
11
12
# Define the Cohere LLM
13
llm = ChatCohere(cohere_api_key="COHERE_API_KEY",              
14
                 model="command-r-plus-08-2024")
15
16
# Create an agent
17
agent = create_cohere_react_agent(
18
    llm=llm,
19
    tools=[internet_search],
20
    prompt=ChatPromptTemplate.from_template("{question}"),
21
)
22
23
# Create an agent executor
24
agent_executor = AgentExecutor(agent=agent, tools=[internet_search], verbose=True)
25
26
# Generate a response
27
response = agent_executor.invoke(
28
    {
29
        "question": "I want to write an essay. Any tips?",
30
    }
31
)
32
33
# See Cohere's response
34
print(response.get("output"))
35
# Cohere provides exact citations for the sources it used
36
print(response.get("citations"))

1
from langchain_cohere import CohereRagRetriever
2
from langchain.retrievers import WikipediaRetriever
3
from langchain_cohere import ChatCohere
4
5
# User query we will use for the generation
6
user_query = "What is cohere?"
7
# Define the Cohere LLM
8
llm = ChatCohere(cohere_api_key="COHERE_API_KEY",              
9
                 model="command-r-plus-08-2024")
10
# Create the Cohere rag retriever using the chat model
11
rag = CohereRagRetriever(llm=llm, connectors=[])
12
# Create the wikipedia retriever
13
wiki_retriever = WikipediaRetriever()
14
# Get the relevant documents from wikipedia
15
wiki_docs = wiki_retriever.invoke(user_query)
16
# Get the cohere generation from the cohere rag retriever
17
docs = rag.invoke(user_query, documents=wiki_docs)
18
# Print the documents
19
print("Documents:")
20
for doc in docs[:-1]:
21
    print(doc.metadata)
22
    print("\n\n" + doc.page_content)
23
    print("\n\n" + "-" * 30 + "\n\n")
24
# Print the final generation
25
answer = docs[-1].page_content
26
print("Answer:")
27
print(answer)
28
# Print the final citations
29
citations = docs[-1].metadata["citations"]
30
print("Citations:")
31
print(docs[-1].__dict__)

1
from langchain_cohere import CohereRagRetriever
2
from langchain_cohere import ChatCohere
3
from langchain_core.documents import Document
4
5
# Define the Cohere LLM
6
llm = ChatCohere(cohere_api_key="COHERE_API_KEY",              
7
                 model="command-r-plus-08-2024")
8
9
# Create the Cohere rag retriever using the chat model
10
rag = CohereRagRetriever(llm=llm, connectors=[])
11
docs = rag.invoke(
12
    "Does LangChain support cohere RAG?",
13
    documents=[
14
        Document(page_content="LangChain supports cohere RAG!", metadata={"id": "id-1"}),
15
        Document(page_content="The sky is blue!", metadata={"id": "id-2"}),
16
    ],
17
)
18
19
# Print the documents
20
print("Documents:")
21
for doc in docs[:-1]:
22
    print(doc.metadata)
23
    print("\n\n" + doc.page_content)
24
    print("\n\n" + "-" * 30 + "\n\n")
25
# Print the final generation
26
answer = docs[-1].page_content
27
print("Answer:")
28
print(answer)
29
# Print the final citations
30
citations = docs[-1].metadata['citations']
31
print("Citations:")
32
print(citations)

1
from langchain_cohere import CohereRagRetriever
2
from langchain_cohere import ChatCohere
3
from langchain_core.documents import Document
4
5
# Define the Cohere LLM
6
llm = ChatCohere(cohere_api_key="COHERE_API_KEY",              
7
                 model="command-r-plus-08-2024")
8
9
# Create the Cohere rag retriever using the chat model with the web search connector
10
rag = CohereRagRetriever(llm=llm, connectors=[{"id": "web-search"}])
11
docs = rag.invoke("Who founded Cohere?")
12
# Print the documents
13
print("Documents:")
14
for doc in docs[:-1]:
15
    print(doc.metadata)
16
    print("\n\n" + doc.page_content)
17
    print("\n\n" + "-" * 30 + "\n\n")
18
# Print the final generation
19
answer = docs[-1].page_content
20
print("Answer:")
21
print(answer)
22
# Print the final citations
23
citations = docs[-1].metadata['citations']
24
print("Citations:")
25
print(citations)

1
from langchain_cohere import ChatCohere
2
from langchain_core.documents import Document
3
from langchain_core.prompts import ChatPromptTemplate
4
from langchain.chains.combine_documents import create_stuff_documents_chain
5
6
prompt = ChatPromptTemplate.from_messages(
7
    [("human", "What are everyone's favorite colors:\n\n{context}")]
8
)
9
10
# Define the Cohere LLM
11
llm = ChatCohere(cohere_api_key="COHERE_API_KEY",              
12
                 model="command-r-plus-08-2024")
13
14
chain = create_stuff_documents_chain(llm, prompt)
15
16
docs = [
17
    Document(page_content="Jesse loves red but not yellow"),
18
    Document(page_content = "Jamal loves green but not as much as he loves orange")
19
]
20
21
chain.invoke({"context": docs})

1
from langchain_cohere import ChatCohere
2
3
# Define the Cohere LLM
4
llm = ChatCohere(cohere_api_key="COHERE_API_KEY",
5
                 model="command-r-plus-08-2024")
6
7
res = llm.invoke("John is five years old", response_format={
8
            "type": "json_object",
9
            "schema": {
10
                "title": "Person",
11
                "description": "Identifies the age and name of a person",
12
                "type": "object",
13
                "properties": {
14
                    "name": { "type": "string", "description": "Name of the person" },
15
                    "age": { "type": "number", "description": "Age of the person" },
16
                },
17
                "required": [
18
                    "name",
19
                    "age",
20
                ],
21
            }
22
        }
23
)
24
25
print(res)

1
from langchain_cohere import ChatCohere
2
from langchain.chains.summarize import load_summarize_chain
3
from langchain_community.document_loaders import WebBaseLoader
4
5
loader = WebBaseLoader("https://docs.cohere.com/docs/cohere-toolkit")
6
docs = loader.load()
7
8
# Define the Cohere LLM
9
llm = ChatCohere(cohere_api_key="COHERE_API_KEY",              
10
                 model="command-r-plus-08-2024",
11
                 temperature=0)
12
13
chain = load_summarize_chain(llm, chain_type="stuff")
14
15
chain.invoke({"input_documents": docs})

1
llm = ChatCohere(base_url=<YOUR_DEPLOYMENT_URL>,
2
                 cohere_api_key="COHERE_API_KEY",              
3
                 model="MODEL_NAME")