Skip to main content
The sequential() strategy processes document chunks one at a time, passing the accumulated results to each subsequent extraction. This allows the model to build context as it processes the document.

​
Usage

import { extract, sequential } from 'struktur';
import { openai } from '@ai-sdk/openai';

const result = await extract({
  artifacts,
  schema,
  strategy: sequential({
    model: openai('gpt-4o'),
    chunkSize: 100000,
  }),
});

​
Configuration

​
model
LanguageModel
required
The AI SDK language model to use for extraction.
​
chunkSize
number
required
Maximum tokens per chunk. Documents are split into batches that fit within this limit.
​
maxImages
number
Maximum number of images per chunk. Useful for controlling vision API costs.
​
outputInstructions
string
Additional instructions to guide the model’s output format or behavior.
​
execute
function
Custom retry executor function. Defaults to runWithRetries.
​
strict
boolean
Enable strict mode for structured output validation. Defaults to false.

​
When to use

  • You have large documents that exceed context limits
  • Sequential context is important (e.g., narratives, meeting minutes)
  • You want to avoid a separate merge step
  • You can tolerate slower processing than parallel strategies

​
Trade-offs

Advantages:
  • No separate merge step needed
  • Model sees accumulated context from previous chunks
  • Lower token usage than parallel + merge
  • Deterministic ordering
Limitations:
  • Slower than parallel strategies (no concurrency)
  • Later chunks see accumulated data, increasing context size
  • Cannot leverage parallelization

​
Performance characteristics

The strategy estimates batches.length + 2 steps:
  1. Prepare
  2. Extract from batch 1 through N (sequential)
  3. Complete
Processing is sequential, so total time = sum of all chunk processing times.

​
Example with image limits

import { extract, sequential } from 'struktur';
import { anthropic } from '@ai-sdk/anthropic';

const result = await extract({
  artifacts: documentWithImages,
  schema: reportSchema,
  strategy: sequential({
    model: anthropic('claude-3-5-sonnet-20241022'),
    chunkSize: 150000,
    maxImages: 5, // Limit to 5 images per chunk
    outputInstructions: 'Extract data sequentially, maintaining chronological order',
  }),
  events: {
    onStep: ({ step, total, label }) => {
      console.log(`Processing: ${step}/${total} - ${label}`);
    },
  },
});

Build docs developers (and LLMs) love

Get started for freeTalk to us