# Test Plan: Long-Context Prompt Chunking

## Objective
Validate that chunking long prompts preserves semantic coherence, retrieval accuracy, and answer quality compared to full-context processing.

## Scope
- Input documents/prompts exceeding context limits (e.g., 10K–1M tokens)
- Chunking strategies: fixed-size, sentence/paragraph-based, recursive, semantic, sliding window with overlap

## Test Scenarios

**1. Chunking correctness**
- Verify chunk sizes stay within token limits (test edge cases: exactly at limit, 1 token over)
- Confirm no content loss or duplication across chunks
- Validate