Skip to main content
Open In ColabOpen on GitHub

Salesforce

This notebook shows how to use the Salesforce tools in LangChain.

The Salesforce integration provides tools for:

  1. Querying Salesforce data using SOQL
  2. Getting information about Salesforce objects
  3. Listing available Salesforce objects

Installation​

First, you need to install the simple-salesforce package:

pip install simple-salesforce
from simple_salesforce import Salesforce
from langchain_community.tools.salesforce import (
    QuerySalesforceTool,
    InfoSalesforceTool,
    ListSalesforceTool
)

Setup​

You can initialize the Salesforce tools in two ways:

  1. Using direct credentials
  2. Using an existing Salesforce instance

Option 1: Using Direct Credentials​

# Initialize Salesforce with credentials
sf = Salesforce(
    username='your_username',
    password='your_password',
    security_token='your_token',
    domain='test'  # Use 'test' for sandbox, remove for production
)

# Create tools
tools = [
    QuerySalesforceTool(sfdc_instance=sf),
    InfoSalesforceTool(sfdc_instance=sf),
    ListSalesforceTool(sfdc_instance=sf)
]

Using the Tools​

1. List Available Objects​

# List all queryable objects in your Salesforce instance
result = list_tool.run("")
print(result)

2. Get Object Information​

# Get information about specific objects
result = info_tool.run("Account,Contact")
print(result)

3. Query Salesforce Data​

# Execute a SOQL query
query = "SELECT Id, Name, Industry FROM Account LIMIT 5"
result = query_tool.run(query)
print(result)

Using with an Agent​

The Salesforce tools can be used with LangChain agents to enable natural language interactions with your Salesforce data:

from langchain_core.prompts import ChatPromptTemplate
from langchain_openai import ChatOpenAI
from langchain.agents import create_openai_functions_agent
from langchain.agents import AgentExecutor

# Initialize the language model
llm = ChatOpenAI(temperature=0)

# Create a prompt template
prompt = ChatPromptTemplate.from_messages([
    ("system", "You are a helpful assistant that can query Salesforce data. "
              "Use the available tools to help answer questions about Salesforce data."),
    ("user", "{input}")
])

# Create the agent
agent = create_openai_functions_agent(llm, tools, prompt)
agent_executor = AgentExecutor(agent=agent, tools=tools, verbose=True)
# Example: Ask the agent to get information about accounts
result = agent_executor.invoke(
    {"input": "What are the top 5 accounts by revenue?"}
)
print(result["output"])

Error Handling​

The tools are designed to handle errors gracefully:

# Example: Invalid SOQL query
result = query_tool.run("SELECT Invalid FROM Account")
print(result)

# Example: Invalid object name
result = info_tool.run("InvalidObject")
print(result)

Was this page helpful?