Core API

cotlab.core.base.BasePromptStrategy

Bases: ABC

Abstract base class for prompt construction strategies.

name abstractmethod property

Strategy name for logging.

build_prompt(input_data) abstractmethod

Build a prompt from input data.

Parameters:

Name	Type	Description	Default
input_data	Dict[str, Any]	Dictionary with at least 'question' key	required

Returns:

Type	Description
str	Formatted prompt string

get_compatible_datasets()

Return list of compatible dataset names, or None if compatible with all.

Override this in specialized prompts to restrict usage.

get_system_message()

Return system message if applicable.

parse_response(response) abstractmethod

Parse model response to extract answer and reasoning.

Parameters:

Name	Type	Description	Default
response	str	Raw model output	required

Returns:

Type	Description
Dict[str, Any]	Dictionary with 'answer', 'reasoning', and any other extracted fields

cotlab.core.base.StructuredOutputMixin

Mixin to add multi-format structured output capability to any prompt strategy.

Supports: PLAIN, JSON, TOON, TOML, XML, YAML, MARKDOWN

Usage

class MyStrategy(StructuredOutputMixin, BasePromptStrategy): def init(self, output_format="plain", ...): self.output_format = output_format self.cot_format = False # Set True for CoT inside structure

cotlab.core.base.ExperimentResult dataclass

JSON-serializable experiment result.

load(path) classmethod

Load from JSON file.

save(path)

Save to JSON file.

to_json()

Serialize to JSON string.

cotlab.core.base.OutputFormat

Supported output formats for structured responses.