Batched Summarization Chain

Transforms text into a concise summary with intelligent batching

Batched Summarization Chain

AI-generated

Summary

Summarize JSON or binary input data in batches using customizable chunking and summarization methods to produce concise text summaries.

Inputs

  • Data to Summarize (required) — Select how to pass data into the summarization chain: JSON input, binary input, or a document loader sub-node.
  • Chunking Strategy — Choose a chunk splitting strategy: simple (by character count and overlap) or advanced (using a text splitter sub-node). Only applicable if using JSON or binary input.
  • Characters Per Chunk — Set the maximum character length for each chunk when using the simple chunking strategy.
  • Chunk Overlap (Characters) — Specify how many characters overlap between chunks when using the simple chunking strategy.
  • Batch Size — Define how many items to process in parallel to control rate limiting and processing order.
  • Delay Between Batches — Set a delay in milliseconds between batches to manage processing speed and rate limits.
  • Input Data Field Name — (When using binary input) Specify the field name containing the binary data to be summarized.
  • Summarization Method and Prompts — Configure the summarization method (map_reduce, refine, or stuff) and customize prompts used to generate summaries for individual chunks and combined results.

Output shape

A list of JSON items where each item's json contains the summarized text output or error details if the summarization failed for that item.

Processes input data items sequentially or in batches according to batch size and delay settings. If an error occurs, output may contain an error message per item if 'Continue On Fail' is enabled.

Discussion