Glyph 1.1.2: Long chats stop failing mid-stream

If a long conversation started refusing to answer, this is the release that fixes it. Glyph had been estimating how much of a conversation would fit into the model rather than measuring it, and past a certain length the estimate was wrong in the direction that fails. What changed Fixed streaming failures on long chats: prompt sizing now follows the model's actual context window and uses exact tokenizer counts, so requests no longer fail with 'exceeds the available context size'. Glyph 1.1.2 is available from fortaify.com, and installed copies update themselves.