Context Window Optimisation — Reduce Tokens, Maintain Quality

I will optimise your LLM application's context window usage — identifying redundant content, compressing verbose instructions, implementing rolling conversation summarisation, and redesigning context injection to use tokens more efficiently without degrading output quality. Context window optimisation directly reduces LLM API costs because input tokens contribute to per-request cost — a context window optimisation that reduces average input token count by 30% reduces costs proportionally at your usage volume, while maintaining output quality through careful retention of the information that actually influences output quality.

The optimisation covers system prompt token audit and compression, conversation history rolling summarisation, retrieved context prioritisation and truncation strategy, instruction deduplication, and before/after token count and quality comparison.