New document recognition engine
Feature Improvement The engine that extracts text from PDFs and images was replaced with a newer one. Extraction accuracy is up across scanned documents and photographed ones. You can check the result in the content area of View details on an item. When a key value is missing, replacing the document with a cleaner copy works better than rewriting the policy.Better document search
Feature Improvement We improved how the agent finds its evidence inside documents. It locates the relevant passage more reliably when a document is long or when there are several of them. Verification accuracy improves across every supported format — adding an item.Token usage per step
New Feature The tokens an analysis consumed can now be seen broken down per work item. It appears in the Info tab and the Audit tab totals, and in the API response assystemMetadata.tokenUsagePerWorkId.
Use it to see which step costs the most when you are tuning a policy — GET /analyses/:analysisId.
Trial expiry calculation
Feature Improvement We tidied up how the expiry date of a trial project is calculated. The days remaining and the expiry date are shown on the Project Info tab — project management.Related pages
- Profiles and items — adding documents and checking the extraction
- Analysis results — the Info and Audit tabs