Implementation:Ucbepic Docetl OpDescriptions

Knowledge Sources	Ucbepic_Docetl
Domains	Data_Processing, Optimization, Documentation
Last Updated	2026-02-08 00:00 GMT

Overview

Concrete tool for structured descriptions of all DocETL operators used in LLM prompts during optimization provided by DocETL.

Description

The op_descriptions module defines the Operator Pydantic model and instantiates structured descriptions for every DocETL operator type. Each Operator instance includes the operator name, whether it is LLM-powered, a description, usage guidance, required/optional parameters, return specification, and a minimal YAML example. The module provides pre-built instances including op_map, op_extract, op_parallel_map, op_filter, op_reduce, and others. These descriptions are serialized via the to_string() method and injected into LLM prompts during MCTS expansion and directive instantiation.

Usage

Use this module when constructing LLM prompts that need to describe available DocETL operators, such as during the MOAR optimizer's expansion phase or when generating pipeline configurations from natural language.

Code Reference

Source Location

• Repository: Ucbepic_Docetl
• File: docetl/reasoning_optimizer/op_descriptions.py
• Lines: 1-638

Signature

class Operator(BaseModel):
    name: str
    type_llm_or_not: str
    description: str
    when_to_use: str
    required_parameters: str
    optional_parameters: Optional[str] = None
    returns: str
    minimal_example_configuration: str

    def to_string(self) -> str: ...

# Pre-built operator instances:
op_map: Operator
op_extract: Operator
op_parallel_map: Operator
op_filter: Operator
op_reduce: Operator
# ... additional operator instances

Import

from docetl.reasoning_optimizer.op_descriptions import Operator, op_map, op_reduce, op_filter

I/O Contract

Inputs

Name	Type	Required	Description
name	str	Yes	The name of the operator (e.g., "Map", "Reduce", "Filter")
type_llm_or_not	str	Yes	Whether the operator is "LLM-powered" or "Not LLM-powered"
description	str	Yes	Multi-sentence description of what the operator does
when_to_use	str	Yes	Guidance on when to select this operator
required_parameters	str	Yes	Formatted string of required YAML parameters
optional_parameters	Optional[str]	No	Formatted string of optional YAML parameters
returns	str	Yes	Description of what the operator returns
minimal_example_configuration	str	Yes	YAML snippet showing minimal operator usage

Outputs

Name	Type	Description
formatted_string	str	Markdown-formatted operator description ready for LLM prompt injection

Usage Examples

from docetl.reasoning_optimizer.op_descriptions import op_map, op_reduce

# Get formatted description for use in an LLM prompt
map_description = op_map.to_string()
reduce_description = op_reduce.to_string()

# Inject into a system prompt
system_prompt = f"""
Available operators:
{map_description}
{reduce_description}
"""

Related Pages

• Environment:Ucbepic_Docetl_Python_Runtime