Chat Prompt Template

A specialized node for creating structured chat prompts with system messages, human messages, and optional message history, enabling dynamic and context-aware conversations with Language Models.

Node Details

• Name: chatPromptTemplate

• Type: ChatPromptTemplate

• Version: 2.0

• Category: Prompts

Base Classes

• ChatPromptTemplate

• BasePromptTemplate

Parameters

Inputs

1. System Message (Required)

   • Type: string

   • Description: Initial system message that sets the context or role for the AI

   • Example: "You are a helpful assistant that translates {input_language} to {output_language}."

2. Human Message (Required)

   • Type: string

   • Description: Human message prompt added at the end of the message sequence

   • Example: “{text}”

3. Format Prompt Values (Optional)

   • Type: JSON

   • Description: Variables specification for use in prompts

   • Example:

     {
       "input_language": "English",
       "output_language": "Spanish"
     }
     

4. Messages History (Optional)

   • Type: Tabs

   • Default: messageHistoryCode

   • Description: Additional messages after System Message for few-shot examples

   • Tabs:

     • Add Messages (Code)

       • Type: code

       • Description: Custom message history using JavaScript code

Functionality

1. Message Management

   • System message handling

   • Human message formatting

   • History integration

   • Variable substitution

2. Template Processing

   • Dynamic execution

   • Code sandboxing

   • Context building

   • Error handling

Use Cases

1. Conversational AI

   • Customer service

   • Language translation

   • Task assistance

   • Knowledge base queries

2. Context Management

   • Role definition

   • Few-shot learning

   • Conversation history

   • Dynamic responses

Integration Notes

• Uses @langchain/core/prompts

• Supports JavaScript execution

• Handles JSON parsing

• Implements sandboxed environments

Best Practices

1. System Message Design

   • Clear instructions

   • Specific role definition

   • Consistent context

   • Appropriate constraints

2. Message History

   • Relevant examples

   • Progressive complexity

   • Clear structure

   • Safe code execution

3. Variable Management

   • Clear naming

   • Type consistency

   • Error handling

   • Default values

Input/Output

Input

• The node takes in the defined parameters (system message, human message, prompt values, and optional message history).

• If message history code is provided, it is executed in a sandboxed environment.

Output

• The node outputs a ChatPromptTemplate object that includes:

  • A system message

  • Optional message history (if provided and valid)

  • A human message

Usage

This node is used to create structured chat prompts for language models. It’s particularly useful for:

1. Setting up consistent system instructions across multiple interactions.

2. Defining a standard format for human inputs.

3. Incorporating few-shot examples or specific conversation context through the message history feature.

4. Allowing for dynamic prompt creation by using variables in the system and human messages.

Implementation Details

• The node uses the @langchain/core/prompts library to create the chat prompt template.

• It supports dynamic execution of JavaScript code for creating custom message histories.

• The node handles JSON parsing for prompt values, allowing for variable substitution in the prompts.

• Error handling is implemented for invalid JSON or message history code execution.

Notes

• The message history feature uses a sandboxed VM for executing user-provided code, ensuring safety and isolation.

• The node is designed to be flexible, allowing for both simple prompt creation and more complex setups with custom message histories.

• It integrates with a database (likely for storing or retrieving additional context), as indicated by the use of DataSource and IDatabaseEntity.