Long chapters on my local model stopped mid-sentence around 4,000 words, or the generation sometimes died on my large (30B+) model. (fixed in v1.6.2)

Two defects, both fixed. First, when a model hit its output-length cap, BookyAI silently accepted the cut-off text — now it detects the cap and automatically asks the

model to continue exactly where it stopped until the chapter is complete. Second, slow models could exceed an internal 5-minute network timeout — generation is now

streamed continuously, so there is no timeout to hit regardless of model size or hardware speed. No settings needed; both fixes are automatic.