Implementation:Ucbepic Docetl GatherOperation Execute

Knowledge Sources	DocETL
Domains	NLP, Context_Management
Last Updated	2026-02-08 01:40 GMT

Overview

Concrete operation for enriching document chunks with peripheral context provided by DocETL's operations module.

Description

GatherOperation groups chunked documents by document ID, sorts them by order key, then renders each chunk with configurable peripheral context (previous/next chunks) and optional hierarchy headers. The output adds a _{content_key}_rendered field containing the contextually enriched chunk text.

Usage

Use GatherOperation after SplitOperation and before MapOperation in a chunking pipeline. Configure peripheral_chunks to control how much surrounding context each chunk receives.

Code Reference

Source Location

• Repository: docetl
• File: docetl/operations/gather.py
• Lines: L8-328

Signature

class GatherOperation(BaseOperation):
    class schema(BaseOperation.schema):
        type: str = "gather"
        content_key: str
        doc_id_key: str
        order_key: str
        peripheral_chunks: dict[str, Any] | None = None
        doc_header_key: str | None = None
        main_chunk_start: str | None = None
        main_chunk_end: str | None = None

    def execute(self, input_data: list[dict]) -> tuple[list[dict], float]:
        """Add peripheral context to chunks. Returns (enriched_chunks, cost=0.0)."""

Import

from docetl.operations.gather import GatherOperation

I/O Contract

Inputs

Name	Type	Required	Description
content_key	str	Yes	Chunk content field name
doc_id_key	str	Yes	Document ID field name
order_key	str	Yes	Chunk order field name
peripheral_chunks	dict	No	Config for previous/next chunk inclusion
doc_header_key	str	No	Field containing document section headers
input_data	list[dict]	Yes	Chunked documents from SplitOperation

Outputs

Name	Type	Description
results	list[dict]	Chunks with added {content_key}_rendered field
cost	float	Always 0.0 (no LLM calls)

Usage Examples

operations:
  - name: gather_context
    type: gather
    content_key: content_chunk
    doc_id_key: split_docs_id
    order_key: split_docs_chunk_num
    peripheral_chunks:
      previous:
        head:
          count: 1
      next:
        tail:
          count: 1
    doc_header_key: headers

Related Pages

Implements Principle

• Principle:Ucbepic_Docetl_Peripheral_Context_Gathering

Requires Environment

• Environment:Ucbepic_Docetl_Python_Runtime