Repository navigation
Veynrel 2.3.1
Markdown generation in Studio now waits up to 90 seconds per request. Stop releases the operation immediately; a provider may still complete and charge for an issued request. There are no automatic retries or continuations.
When a valid nonempty Markdown response reaches the output limit, Studio shows Incomplete response: output limit reached, its literal text and Copy fragment. This fragment cannot replace a source or create a result note. Complete rows in mixed batches keep their usual guarded Review, Apply and Undo. History retains the incomplete status; preview text is not restored after restart.
Other API callers, diagrams, prompts, models and dependencies retain their existing contracts. This patch does not establish model quality. Offline replay reconstructs envelopes from 14 saved historical Qwen responses and their delays: 13 complete results and one incomplete preview. No new inference was performed; the original 2.3.0 experiment remains unchanged.
Update through Obsidian when offered, or use the separate main.js, manifest.json, and styles.css assets. Preserve data.json, indexes, Recall state and Studio journals. The plugin ID remains ai-knowledge-hub and minimum Obsidian version remains 1.8.7.