[Suggestion / Feature] Auto-continue when generation stops mid-output in chat.z.ai / ZCode
Verified demand: ~51 messages report the model stopping in the middle of a task ("why it always stop writing the html code at the middle and does not continue", "GLM stopping in the middle of a task"). Users must manually prompt "continue" and hope it resumes.
Proposal: Detect when a generation ends prematurely (truncated mid-sentence / no natural stop token) and automatically append a "continue" turn, or show a one-click "Resume generation" button — instead of leaving the user to notice and type it themselves.
Why this matters: Distinguishes from server-saturation drops (issue #142) — this is output truncation, not overload. Auto-resume removes a recurring friction point and makes long outputs (code, docs) reliable.
Regards,
Roman
[Suggestion / Feature] Auto-continue when generation stops mid-output in chat.z.ai / ZCode
Verified demand: ~51 messages report the model stopping in the middle of a task ("why it always stop writing the html code at the middle and does not continue", "GLM stopping in the middle of a task"). Users must manually prompt "continue" and hope it resumes.
Proposal: Detect when a generation ends prematurely (truncated mid-sentence / no natural stop token) and automatically append a "continue" turn, or show a one-click "Resume generation" button — instead of leaving the user to notice and type it themselves.
Why this matters: Distinguishes from server-saturation drops (issue #142) — this is output truncation, not overload. Auto-resume removes a recurring friction point and makes long outputs (code, docs) reliable.
Regards,
Roman