Back to All changelogs
September 15, 2025

Page classification now includes reasoning explanations

Page classification results now include the model's reasoning for each decision to help with debugging and prompt engineering.

Key Highlights

  • Detailed explanations for why pages received specific classifications
  • Helps identify prompt engineering opportunities and debug classification errors
  • Automatically included in all classification results with no performance impact

What's new

Page classification results now include the model's reasoning for each classification decision. When pages are classified, you'll receive both the classification labels and a text explanation of why those labels were chosen.

Why it matters

  • Debug classification errors - understand why pages got unexpected labels
  • Improve prompt engineering - see how the model interprets your classification prompts
  • Build confidence - verify the model's decision-making process
  • Iterate faster - refine prompts based on reasoning insights

Highlights

  • Detailed reasoning for each page classification
  • Helps identify prompt engineering opportunities  
  • Useful for complex documents where classification logic isn't obvious
  • No performance impact - reasoning is generated alongside classification

How to use

Reasoning is automatically included in classification results:

1[.code-block-title]Code[.code-block-title]doc_ai = DocumentAI()
2
3result = doc_ai.parse_and_wait(
4  file="complex_document.pdf",
5  page_classification=PageClassificationConfig(
6    page_classes=["Terms", "Transactions", "Account_Info"]
7  )
8)
9
10for page in result.pages:
11  print(f"Page {page.page_number}:")
12  print(f"Classifications: {page.classifications}")
13  print(f"Reasoning: {page.classification_reasoning}")

Example output:

1[.code-block-title]Example Output[.code-block-title]{
2  "page_class": "Terms",
3  "page_numbers": [
4    2
5  ],
6  "classification_reasons": {
7    "2": "The page provides instructions on how to balance an account, which involves managing account information and balances. It also details how to track and reconcile deposits and withdrawals, directly relating to transactions. Furthermore, it outlines terms and procedures for resolving errors, understanding billing rights, and calculating finance charges, which fall under terms and conditions."
8  }
9},

Use cases

  • Prompt debugging - See why certain pages aren't getting expected classifications
  • Model validation - Verify the classification logic makes sense for your use case  
  • Documentation - Generate explanations for downstream users about document structure
  • Quality assurance - Spot-check model reasoning on complex documents

Status

✅ Live now. Classification reasoning included automatically in all page classification results.

Get server-less runtime for agents and data ingestion

Data ingestion like never before.
TRUSTED BY PRO DEVS GLOBALLY

Tensorlake is the Agentic Compute Runtime the durable serverless platform that runs Agents at scale.

"At SIXT, we're building AI-powered experiences for millions of customers while managing the complexity of enterprise-scale data. TensorLake gives us the foundation we need—reliable document ingestion that runs securely in our VPC to power our generative AI initiatives."

Boyan Dimitrov
CTO, Sixt

“Tensorlake enabled us to avoid building and operating an in-house OCR pipeline by providing a robust, scalable OCR and document ingestion layer with excellent accuracy and feature coverage. Ongoing improvements to the platform, combined with strong technical support, make it a dependable foundation for our scientific document workflows.”

Yaroslav Sklabinskyi
CEO, Reliant AI

"For BindHQ customers, the integration with Tensorlake represents a shift from manual data handling to intelligent automation, helping insurance businesses operate with greater precision, and responsiveness across a variety of transactions"

Cristian Joe
CEO @ BindHQ

“Tensorlake let us ship faster and stay reliable from day one. Complex stateful AI workloads that used to require serious infra engineering are now just long-running functions. As we scale, that means we can stay lean—building product, not managing infrastructure.”

Arpan Bhattacharya
Founder & CEO @ The Intelligent Search Company