Salesforce

Tools for interacting with Salesforce.

Overview

This notebook provides examples of interacting with Salesforce using LangChain.

Setup

1. Install the required dependencies:

   pip install langchain-salesforce

2. Set up your Salesforce credentials as environment variables:

   export SALESFORCE_USERNAME="your-username"
   export SALESFORCE_PASSWORD="your-password" 
   export SALESFORCE_SECURITY_TOKEN="your-security-token"
   export SALESFORCE_DOMAIN="test" # Use 'test' for sandbox, remove for production

These environment variables will be automatically picked up by the integration.

Getting Your Security Token

If you need a security token:

1. Log into Salesforce
2. Go to Settings
3. Click on "Reset My Security Token" under "My Personal Information"
4. Check your email for the new token

Instantiation

import os

from langchain_salesforce import SalesforceTool

username = os.getenv("SALESFORCE_USERNAME", "your-username")
password = os.getenv("SALESFORCE_PASSWORD", "your-password")
security_token = os.getenv("SALESFORCE_SECURITY_TOKEN", "your-security-token")
domain = os.getenv("SALESFORCE_DOMAIN", "login")

tool = SalesforceTool(
    username=username, password=password, security_token=security_token, domain=domain
)

Invocation

def execute_salesforce_operation(
    operation, object_name=None, query=None, record_data=None, record_id=None
):
    """Executes a given Salesforce operation."""
    request = {"operation": operation}
    if object_name:
        request["object_name"] = object_name
    if query:
        request["query"] = query
    if record_data:
        request["record_data"] = record_data
    if record_id:
        request["record_id"] = record_id
    result = tool.run(request)
    return result

Query

This example queries Salesforce for 5 contacts.

query_result = execute_salesforce_operation(
    "query", query="SELECT Id, Name, Email FROM Contact LIMIT 5"
)

Describe an Object

Fetches metadata for a specific Salesforce object.

describe_result = execute_salesforce_operation("describe", object_name="Account")

List Available Objects

Retrieves all objects available in the Salesforce instance.

list_objects_result = execute_salesforce_operation("list_objects")

Create a New Contact

Creates a new contact record in Salesforce.

create_result = execute_salesforce_operation(
    "create",
    object_name="Contact",
    record_data={"LastName": "Doe", "Email": "doe@example.com"},
)

Update a Contact

Updates an existing contact record.

update_result = execute_salesforce_operation(
    "update",
    object_name="Contact",
    record_id="003XXXXXXXXXXXXXXX",
    record_data={"Email": "updated@example.com"},
)

Delete a Contact

Deletes a contact record from Salesforce.

delete_result = execute_salesforce_operation(
    "delete", object_name="Contact", record_id="003XXXXXXXXXXXXXXX"
)

Chaining

from langchain.prompts import PromptTemplate
from langchain_openai import ChatOpenAI
from langchain_salesforce import SalesforceTool

tool = SalesforceTool(
    username=username, password=password, security_token=security_token, domain=domain
)

llm = ChatOpenAI(model="gpt-4o-mini")

prompt = PromptTemplate.from_template(
    "What is the name of the contact with the id {contact_id}?"
)

chain = prompt | tool.invoke | llm

result = chain.invoke({"contact_id": "003XXXXXXXXXXXXXXX"})

API Reference:PromptTemplate | ChatOpenAI

API reference

langchain-salesforce README

Related

• Tool conceptual guide
• Tool how-to guides