Configure Aspose.LLM for .NET for 128K-262K contexts — pick the right preset, enable flash attention, quantize the KV cache, and manage memory.... Vision + text with many image turns. Prerequisites A preset...Without flash attention, KV reads scale quadratically with context size...