CCA-F: Claude Certified Architect — Foundations

By Anthropic · Question Mode

QuestionQ110

Context Management & Reliability

You are creating a structured data-extraction system with Claude. The system extracts information from unstructured documents, validates its output with JavaScript Object Notation (JSON) schemas, and preserves high accuracy. It must gracefully handle edge cases and integrate with downstream systems.

After the daily batch of 10,000 documents finishes, 300 documents (3%) fail with context_length_exceeded errors. The results file identifies every failed item by custom_id.

What is the most cost-effective way to process these failures?

Explanation

A context-length failure is addressed by reducing the size of each affected request, such as by chunking the failed documents and combining their partial extractions. Resubmitting only the requests identified by custom_id avoids reprocessing the 9,700 successful documents. Increasing max_tokens does not shrink the oversized context, while prompt caching only reduces the cost of reusable prompt content and a larger-context model tier may incur additional long-context cost.

Learn more

Community Discussion

No comments yet. Be the first to start the discussion!