Prompt for scoring each prescription (procedural instruction) individually against its specific intent and the trajectory evidence.
Per the PlugMem paper, each prescription is evaluated on a 1-10 scale assessing whether the intent was achieved and how well the prescription was executed. Scores are normalized to [0.0, 1.0] for internal use.
Returns structured output via chat_structured/3 using a Zoi schema.
Summary
Functions
Builds the system and user messages for the return-scoring prompt.
Parses and normalizes the scored prescriptions from the LLM response.
Returns the Zoi schema for structured LLM output validation.
Functions
Builds the system and user messages for the return-scoring prompt.
@spec parse_response(map()) :: {:ok, [map()]} | {:error, Mnemosyne.Errors.Invalid.PromptError.t()}
Parses and normalizes the scored prescriptions from the LLM response.
@spec schema() :: Zoi.Type.t()
Returns the Zoi schema for structured LLM output validation.