Deploy notes contain code changes that are intended to fix bugs, improve performance, and prepare for new features. These deploys take place monthly.
Unless otherwise stated, all features in this release are available in the Production environment as of 2026-July.
Updated Features
Ask Your Data
Large Language Model (LLM) Updates
Location to Enable Feature | NA |
|---|
Name of LTI Tool | Ask Your Data |
Permissions | Intelligent Insights |
Affects User Interface | No |
Affected Areas | LLM |
Summary
- Reliable query results: Canvas now preserves the intended query engine and SQL dialect more consistently across questions, refinements, saved analyses, and replays. This reduces failures caused by a query being interpreted through the wrong data path.
- Stronger recovery when services are busy: Improved retry, backpressure, and fallback behavior helps analyses recover when data services are temporarily constrained or one query engine is slower than expected. Saved results, downloads, and pinboards also retain their original query context more reliably.
- AI response handling more dependable: Updated provider timeout handling reduces avoidable retries and failures on complex questions when an AI provider needs additional time to respond.
- Underlying data stays consistent: Behind-the-scenes improvements to data freshness, synchronization, and processing make Canvas analyses more dependable across production regions.
- Security and maintenance: Routine security, dependency, and reliability updates.
- Pinboards more reliable: Large pinboards now use a more controlled data-loading flow that reduces duplicate requests and retry storms. Temporary retrieval problems are retried more intelligently, and table or chart errors now provide a clear Try again action.
- Accurate Canvas questions: Improved SQL handling, table-name resolution, and query recovery reduce failures when answering questions across different Canvas datasets and regions.
- Faster recovery from temporary service issues: Tighter response limits and safer fallback behavior help Ask Your Data recover more quickly when an AI provider or data service is slow or temporarily unavailable.
- Improved Canvas data processing: Reduced redundant discussion-enrollment processing and improved handling of temporary service pressure, providing more consistent answers during busy periods.
