We are thrilled to roll out the very first official product update for UptoCode. This release focuses heavily on accelerating your workflow, enhancing AI accuracy, and delivering a seamless, reliable experience.
Here is a breakdown of what is new, what is faster, and what has been improved.
New features
- Parallel uploads. Upload drawings, regulations, and source files simultaneously. No more waiting for one file to finish before starting the next.
- Precision compliance reports. Reports are now generated directly from your attached regulation documents, complete with exact, matching citations.
- Automated web-sourced audits. No document attached? No problem. UptoCode now uses web-sourced regulations through a multi-step verification process, so you still get an accurate audit.
Speed and performance
- Smarter context caching. Repeat questions within the same chat respond noticeably faster.
- Instant multi-step queries. The AI chains drawings, IFC, and regulations together with zero delay between steps.
- Concurrent CAD ingestion. Large CAD PDFs are processed pages-concurrently, drastically cutting turnaround times.
Reliability and bug fixes
- Robust file handling. Fixed intermittent "file processing failed" errors on large drawing PDFs and resolved occasional document attachment glitches in chats.
- Automatic recovery. Failed uploads automatically retry on transient errors, and stuck processing jobs are instantly detected and recovered.
- Clean AI responses. Internal placeholders and scrubbed text have been eliminated. Responses are now perfectly clean and citation-ready.
- Financial precision. Credit balances in both admin and end-user views are perfectly synced and displayed with precise decimal accuracy.
Quality of life
- Dynamic UI indicators. The chat input unlocks the exact moment your files finish processing, and independent file-type upload states mean uploading a drawing will not lock the regulation upload button.
- Longer AI outputs. We extended the allowed length for AI responses, with a clear "truncated" indicator in the rare event a model hits a limit.
- Enhanced analytics and logs. Failed conversation turns are recorded with partial costs for precise usage analytics, backed by richer diagnostic logging for our support team.
Thank you to our early users whose invaluable feedback shaped this release. Sign in to try it.