LangSmith/client/Client/pull_prompt/include_model

• Client.pull_prompt() — 🦜️🛠️ LangSmith documentation

  • pull_prompt(prompt_identifier: str, *, include_model: bool | None = False) → Any

  • Overview

    • The pull_prompt method is a core functionality of the LangSmith Client that allows you to retrieve prompts stored in LangSmith PromptHub.
    • This method enables seamless integration between your code and prompts managed in LangSmith, facilitating prompt versioning, sharing, and reuse.

  • Parameters

    • prompt_identifier: str

      • A string that uniquely identifies the prompt to retrieve
      • Can be in several formats:
        • Simple name for your own private prompts: "joke-generator"
        • With author handle for public prompts: "username/prompt-name"
        • With version tag or commit hash: "joke-generator:v1" or "joke-generator:12344e88"

    • include_model: bool | None = False

      • When set to True, returns the prompt bundled with its associated model configuration
      • When False (default), returns only the prompt template

  • Model Identification Implications

    • How include_model=True Works

      • When a prompt is stored with an associated model in PromptHub, setting include_model=True will:
        • Return a RunnableSequence that combines both the prompt template and the model
        • Automatically configure the appropriate model with the settings stored in PromptHub
        • Enable direct invocation without needing to manually set up the model
      • Example:
        • python

          from langsmith import Client
          
          client = Client()
          # Returns a RunnableSequence of prompt + model
          chain = client.pull_prompt("structured-output-prompt", include_model=True)
          # Can be directly invoked
          result = chain.invoke({"input": "data to process"})

    • Environment Requirements

      • When using include_model=True, you must have:
        • The appropriate model provider package installed (e.g., langchain_openai)
        • Required environment variables set (e.g., OPENAI_API_KEY)
        • Access permissions to the model specified in the prompt

  • Relationship to LangSmith PromptHub

    • PromptHub Integration

      • PromptHub serves as a central repository for storing, versioning, and sharing prompts
      • pull_prompt is the primary method to retrieve prompts from PromptHub programmatically
      • Enables collaboration across teams by providing access to shared prompt resources

    • Versioning Capabilities

      • Supports retrieving specific versions of prompts using tags or commit hashes
      • Facilitates reproducibility by allowing exact prompt versions to be referenced
      • Enables A/B testing by comparing different prompt versions

  • Structured Output Applications

    • Optimizing for Structured Responses

      • Particularly valuable when working with prompts designed for structured output
      • When combined with include_model=True, can pull prompts specifically tuned for:
        • JSON output formatting
        • Schema validation
        • Consistent response structures

    • Example with Structured Output

      • python

        from langsmith import Client
        
        client = Client()
        # Pull a prompt designed for structured JSON output with its model
        structured_chain = client.pull_prompt("json-extractor", include_model=True)
        
        # The model is already configured with the right parameters for structured output
        # such as response_format={"type": "json_object"} for OpenAI models
        result = structured_chain.invoke({"text": "Extract entities from this text..."})
        # result will be properly structured according to the prompt's design

  • Best Practices

    • When to Use include_model=True

      • When you want to use the exact model configuration stored with the prompt
      • When sharing prompts across a team to ensure consistent model settings
      • When the prompt has been optimized for a specific model

    • When to Use include_model=False

      • When you want to use your own model configuration
      • When you need to customize model parameters for your specific use case
      • When you want to use the prompt with a different model than originally specified

  • Related Methods

    • Other PromptHub Interactions

      • push_prompt() - Store a prompt in PromptHub
      • list_prompts() - List available prompts
      • delete_prompt() - Remove a prompt from PromptHub
      • like_prompt() / unlike_prompt() - Interact with community prompts