I3eka commented on PR #44492: URL: https://github.com/apache/superset/pull/44492#issuecomment-5758496682
Live validation is complete on an existing Native AI integration with this focused fix applied: - At a 2,048-token cap, the provider-limited reply retained all received text and ended with the error outcome and `done.ok=false`. The incomplete notice survives reopening, docking and page reload. - With the deployment-only allowance raised to 4,096, an 8,855-character synthetic analytical answer completed in one model turn, with no tools, retries or continuation. The SSE final text, persisted message and whole rendered answer agree. Reload, docking and actual HTTP clipboard copy/paste also preserve the complete answer. - The integrated checkout passes 742 backend AI tests, 135 targeted frontend tests and all applicable changed-file checks, including TypeScript. These integration results do not replace the older parent-branch validation limitations documented above. The initial repetitive-text probe received a short model refusal, so it was not counted as a long-answer pass; the analytical check reused that test conversation. No deployment settings, private data or additional fixes were added to this PR. The draft/dependency hold remains unchanged. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
