LLM Chain

​

Node Details

• Name: llmChain
• Type: LLMChain
• Category: [[Chains]]
• Version: 3.0

​

Parameters

1. Language Model (Required)
   • Type: BaseLanguageModel
   • Description: The language model to be used for generating responses.
2. Prompt (Required)
   • Type: BasePromptTemplate
   • Description: The prompt template to be used for constructing queries to the LLM.
3. Output Parser (Optional)
   • Type: BaseLLMOutputParser
   • Description: Parser to process and structure the output from the LLM.
4. Input Moderation (Optional)
   • Type: Moderation[]
   • Description: Moderation tools to detect and prevent harmful input.
   • List: true
5. Chain Name (Optional)
   • Type: string
   • Description: A name for the chain, useful for identification in complex workflows.
   • Placeholder: “Name Your Chain”

​

Input

• Depends on the prompt template, but typically a string or an object containing variables for the prompt.

​

Output

• If Output Prediction is selected:
  • The processed response from the LLM, potentially parsed if an output parser is specified.
• If LLM Chain is selected:
  • The LLMChain object itself, which can be used in other components.

​

How It Works

1. The chain receives input, which is used to populate the prompt template.
2. If input moderation is enabled, it checks the input for potential harmful content.
3. The populated prompt is sent to the language model for processing.
4. The LLM generates a response based on the prompt.
5. If an output parser is specified, it processes the LLM’s response.
6. The final output (either the processed response or the chain object) is returned.

​

Use Cases

• Generating creative content based on specific prompts
• Answering questions or providing explanations on various topics
• Translating or paraphrasing text
• Analyzing and summarizing documents
• Generating code or technical documentation
• Creating conversational AI components

​

Special Features

• Flexible Prompt Engineering: Supports various prompt templates for complex query construction.
• Output Parsing: Can structure LLM outputs into specific formats or extract particular information.
• Input Moderation: Optional safeguards against inappropriate or harmful inputs.
• Customizable: Can be tailored for specific use cases through prompt design and output parsing.
• Streaming Support: Compatible with streaming responses for real-time interaction.
• Multi-Modal Capability: Can handle text and image inputs if the language model supports it.

​

Notes

• The effectiveness of the chain heavily depends on the quality of the prompt template and the capabilities of the chosen language model.
• Custom prompt templates can significantly influence the behavior and output of the chain.
• When using the chain in production, implement proper error handling, especially for potential API failures or moderation issues.
• Consider the token limits of the chosen LLM when designing prompts and processing outputs.
• The chain supports both text-only and multi-modal (text + image) inputs, depending on the language model’s capabilities.
• Output parsing can be crucial for integrating LLM responses into structured workflows or databases.