02. Runnable structure (graph) review

LCEL runnable After creating it, you will often want to check it to better understand what is going on.

This laptop covers some ways to do this.

Copy

# Configuration file for managing API keys as environment variables
from dotenv import load_dotenv

# Load API key information
load_dotenv()

Copy

 True 

Copy

# LangSmith Set up tracking. https://smith.langchain.com
# !pip install langchain-teddynote
from langchain_teddynote import logging

# Enter a project name.
logging.langsmith("LCEL-Advanced")

Copy

 Start tracking LangSmith. 
[Project name] 
LCEL-Advanced 

Copy

# !pip install -qU faiss-cpu tiktoken

# Installing libraries for drawing graphs
# !pip install -qU grandalf

Copy

from langchain.prompts import ChatPromptTemplate
from langchain.vectorstores import FAISS
from langchain_core.output_parsers import StrOutputParser
from langchain_core.runnables import RunnableLambda, RunnablePassthrough
from langchain_openai import ChatOpenAI, OpenAIEmbeddings

vectorstore = FAISS.from_texts(
    # Create FAISS vector stores from text data.
    ["Teddy is an AI engineer who loves programming!"],
    embedding=OpenAIEmbeddings(),
)

# Create a retriever based on a vector store.
retriever = vectorstore.as_retriever()

template = """Answer the question based only on the following context:
{context}  

Question: {question}"""

prompt = ChatPromptTemplate.from_template(
    template
)  # Create a ChatPromptTemplate based on a template.

model = ChatOpenAI(model="gpt-4o-mini")  # Initialize the ChatOpenAI model.

# Create a chain.
chain = (
    # Specify the search context and question.
    {"context": retriever, "question": RunnablePassthrough()}
    | prompt  # Generates a prompt.
    | model  # Run the language model.
    | StrOutputParser()  # Parse the output result into a string.
)

Check graph configuration

You can get a graph of runnable.

chain.get_graph() The method returns the chain's execution graph.

• This method returns a graph object representing the connection between each node and node in the chain.

• The node in the graph represents each step in the chain, and the edge represents the data flow between the steps.

Copy

# Get a node from a graph of chains.
chain.get_graph().nodes

Copy

 {'42e8cf39cdc04753b462eadf704b90dd': Node(id='42e8cf39cdc04753b462eadf704b90dd', name='Parallel  Input', data=  Input'>, metadata=None),'c2838bfce6a9401ba2807cb5cb178fc7': Node(id='c2838bfce6a9401ba2807cb5cb178fc7', name<  Output', data=  Output'>, metadata=None),'e67d152153c6484b76a3746f60e3463': Node(id='e67d152153c6484b76a3746  '144b3154fa'44912ada6a839e0467': Node (id='e744b3154fa44912n{context} \n\nQuestion: {question}')], metadata=None), '6c6d616c02bd4c81b43810b64f6fc': Node(id=6  , async_client=  , root_client=  , root_async_client=  , model_name='gpt-4o-mini', openai_api_key=SecretStr('********'), openai_proxy=','dc2909a51b59  , metadata=None)} 

Copy

# Get an edge from the graph of a chain.
chain.get_graph().edges

Copy

 [Edge ='b253fc2bd96d4501b3e7bfcd4538e968', target=aab4632cd59, data<G1 data=None, conditional=False), Edge(source='e71b4fd10a1641fdbdf193d26560aa5e', target='266d15455db4fc99b1f0102  361740ae4a444bb949ed995e03139ae', target='11880a33c9124dacb537c57e47d1f5c7', data=None, conditional=False), Edge 

Graph output

Outputing the graph allows you to express it in an easy-to-understand form.

Although the output results are not very easy to read, the output allows you to check the graph in a more understandable form.

Copy

# Prints the graph of the chain in ASCII format.
chain.get_graph().print_ascii()

Copy

           +---------------------------------+         
           | Parallel  Input |         
           +---------------------------------+         
                    ** **                
                 *** ***             
               ** **           
+----------------------+ +-------------+  
| VectorStoreRetriever | | Passthrough |  
+----------------------+ +-------------+  
                    ** **                
                      *** ***                  
                         ** **                     
           +----------------------------------+        
           | Parallel  Output |        
           +----------------------------------+        
                             *                         
                             *                         
                             *                         
                  +--------------------+               
                  | ChatPromptTemplate |               
                  +--------------------+               
                             *                         
                             *                         
                             *                         
                      +------------+                   
                      | ChatOpenAI |                   
                      +------------+                   
                             *                         
                             *                         
                             *                         
                   +-----------------+                 
                   | StrOutputParser |                 
                   +-----------------+                 
                             *                         
                             *                         
                             *                         
                +-----------------------+              
                | StrOutputParserOutput |              
                +-----------------------+              

Bring prompt

An important part of the chain is the prompt used.

chain.get_prompts() The method returns a list of prompt objects used in the chain.

Copy

chain.get_prompts()  # Get the prompt used in the chain.

Copy

 [ChatPromptTemplate (input_variables=['context','question'], messages=[HumanMessagePromptTemplate(prompt=PromptTemplate>['context','question