LLMPromptConfigObject

class council.prompt.LLMPromptConfigObject(kind: str, version: str, metadata: DataObjectMetadata, spec: T)

Bases: DataObject[LLMPromptConfigSpec]

Helper class to instantiate a LLMPrompt from a YAML file

get_system_prompt_template(model: str) → str

Return system prompt template for a given model.

get_user_prompt_template(model: str) → str

Return user prompt template for a given model. Raises ValueError if no user prompt template was provided.

property has_user_prompt_template: bool

Return True, if user prompt template was specified in yaml file.

Code Example

The following code illustrates the way to load prompt from a YAML file.

from council.prompt import LLMPromptConfigObject

prompt = LLMPromptConfigObject.from_yaml("data/prompts/llm-prompt-sql-template.yaml")
system_prompt = prompt.get_system_prompt_template("default")
user_prompt = prompt.get_user_prompt_template("default")

Sample yaml file:

kind: LLMPrompt
version: 0.1
metadata:
  name: "SQL_template"
  description: "Prompt template used for SQL generation"
  labels:
    abc: xyz
spec:
  system:
    - model: default
      template: |
        You are a sql expert solving the `Task` 
        leveraging the database schema in the `DATASET` section.

        # Instructions
        - Assess whether the `Task` is reasonable and possible 
          to solve given the database schema
        - Keep your explanation concise with only important details and assumptions
        
        {dataset_description}
        
        # Response formatting
        
        Your entire response must be inside the following code blocks. 
        All code blocks are mandatory.
        
        ```solved
        True/False, indicating whether the task is solved
        ```
        
        ```explanation
        String, concise explanation of the solution if solved or reasoning if not solved
        ```
        
        ```sql
        String, the sql query if the task is solved, otherwise empty
        ```
  user:
    - model: default
      template: |
        {question}