SinglePass and Summarized Extraction → Unstract.com
Complete Extraction, Optimized Execution
SinglePass and Summarized Extraction introduces two approaches to save token costs: prompt consolidation and document summarization.
Current extraction patterns don't scale. They multiply
Every new field you extract increases your API calls—and your costs.
The math is brutal
More fields × More documents = Exponentially higher costs.
You’re stuck between two choices
Limit the data you extract or watch your API costs spiral out of control.
Neither option works when you need comprehensive extraction at scale.
Minimum tokens. Maximum extraction. Zero refactoring.
SinglePass Extraction
Consolidate your prompts into one optimized version and execute it once against the full document. Speed up processing and experience dramatic cost savings.
Summarized Extraction
Summarized Extraction creates a concise version of your documents with only vital information. It then runs your prompts against this optimized version. Process lesser text to speed up your extractions, while using fewer tokens.
SinglePass Extraction
Consolidate your prompts into one optimized version and execute it once against the full document. Speed up processing and experience dramatic cost savings.
Summarized Extraction
Transformative results
SinglePass and Summarized Extraction delivers tangible returns across your entire flow of operations.
7x
Token Cost Savings
80%
Reduction in Latency
100%
Document Context Maintained
Lenny Hartman
CTO, Tokenstreet
Unstract lets us turn a wide range of document formats into clean, structured data with low integration effort and without compromising on enterprise-grade controls. Its high extraction accuracy paired with clear highlight-based validation speeds up our process a lot!
FAQs
How do SinglePass and Summarized Extraction reduce token costs?
SinglePass Extraction combines all your field extraction prompts into one optimized request. Instead of sending your document to the LLM 20 times to extract 20 fields, it gets sent once. This reduces token consumption and API calls.
Summarized Extraction first creates a summary of your document containing only relevant information, and then performs extraction on this compressed version. The original document may be 50 pages, but the summarized version could be just 2-3 pages with all the necessary data preserved.
Will these features compromise extraction accuracy?
No. Both features are designed to maintain the same accuracy levels as traditional extraction methods.
- SinglePass reduces the chance of conflicting extractions between multiple calls.
- Summarized Extraction removes the irrelevant information that can confuse LLMs.
Which feature should I use for my documents?
Use SinglePass when:
- Processing standardized documents (invoices, forms, claims)
- Extracting many fields from shorter documents (under 10 pages)
- Need the absolute fastest processing time
- Working with structured or semi-structured documents
Use Summarized Extraction when:
- Processing long, complex documents
- Document length varies
- Context and nuance are critical for accurate extraction
- Documents contain lots of irrelevant information