AI Response Improvements

• Extended reasoning for final answers provides more thorough explanations
• Previous year answer rewrites now stay closer to the original content, improving consistency
• Better table formatting with clearer shape explanations

API Pagination

• New pagination utility with configurable default and maximum limits for list endpoints

Previous-Year Answer Feature

• You can now leverage answers from previous reporting periods when completing questionnaires. The system intelligently matches questions across years and filters relevant historical answers from your knowledge hub.

Knowledge Records Enhancements

• Date fields are now available when creating knowledge records, enabling better temporal tracking of your data.
• A new method to list knowledge records provides easier access to your stored information.

Table Processing Improvements

• Improved handling of tables with fixed-value row labels for more accurate "not applicable" assessments.
• Enhanced table reduction logic ensures all findings and positive selections are properly included in results.
• Better error handling for table processing with graceful fallbacks when issues occur.

Answer Scoring Improvements

• Answer analysis now incorporates screening guidance when scoring responses, providing more accurate and context-aware evaluations based on your questionnaire requirements.

Document Processing

• Document indexing now handles large pages more reliably by capping page length, preventing processing failures on documents with unusually long content.

Bug Fixes

• Fixed an issue where screening could fail to run when using shared templates due to workspace mismatches.

Attribute-Based Access Control (ABAC) You can now configure fine-grained permissions across your workspace using attribute-based access control. This new system lets you set access rules at the entity level for screenings, questionnaires, folders, files, tags, URLs, and companies—giving you precise control over who can view or modify specific resources.

New Contributor Role A new CONTRIBUTOR role is now available in workspaces, allowing you to grant users limited permissions for collaboration without full workspace access.

Questionnaire Response Improvements New payload builders for short answer, date, and metric question types improve how questionnaire responses are structured and processed.

Questionnaire Sharing

• You can now share questionnaires between workspaces using new sharing endpoints
• Share, list, and revoke access to questionnaires across your organization

Company Search Performance

• Company lookups by name, homepage, and country are now faster thanks to optimized database indexing.

Batch Screenings

• Improved task reference handling for more reliable batch screening operations and better traceability of screening results.

Evidence Enhancements

• Evidence responses now include source URLs, making it easier to trace findings back to their original documents.

Screening Improvements

• Fixed the order of validation checks during screening creation for more reliable processing.
• Improved handling of selections when no evidence is found, preventing unnecessary reversals.

New Object Type

• You can now use BENCHMARK_REPORT as an object type when working with the API, enabling better categorization of benchmark-related documents.

Benchmark Feedback

• You can now submit feedback on benchmark results, helping to improve analysis quality over time. This extends the existing feedback system to support benchmark-specific use cases.

Manual Answer Mode

• You can now set answer_mode on screenings to control how responses are generated. In manual mode, the system waits for human input rather than auto-filling answers, giving you full control over response content.

Screening Assignees

• Assign team members to screenings using the new assignee_id field. Assignees receive email notifications when assigned, making it easier to delegate and track screening tasks across your team.

Answer Scoring

• A new scoring endpoint analyzes responses against scoring criteria from your templates. This helps evaluate answer quality and compliance with your defined standards.

Additional Improvements

• Table-level reasoning is now included in screening exports for better audit trails
• Improved handling of tables with no evidence, reducing false negatives

Webhook payloads now include the main question ID, making it easier to correlate webhook events with specific questions in your system. This allows you to directly reference the parent question when processing webhook callbacks without additional API lookups.