January 2025 Release Notes

Welcome to the January 2025 release notes. This month's release focuses on stability, performance, and several highly requested quality-of-life improvements. We have addressed 47 bug reports, shipped 12 new features, and improved API response times across the board. Here is a detailed breakdown of the most impactful changes.

API Performance Improvements

We have completed a major optimization pass on our core API endpoints. The /agents/query endpoint now responds 35% faster thanks to improved query planning and connection pooling. The /conversations/history endpoint has been optimized for large conversation threads, reducing response times from an average of 800ms to under 200ms for threads with over 500 messages.

Rate limiting has been refined to use a sliding window algorithm instead of the previous fixed window approach. This provides a smoother experience for applications that make bursty API calls and eliminates the thundering herd problem that some customers experienced at window boundaries.

Performance is not a feature - it is a fundamental requirement. Every millisecond of latency erodes user trust and engagement.

Dashboard Updates

The analytics dashboard has received a visual refresh with improved data visualization components. New chart types include funnel analysis for conversation flows, cohort retention views for user engagement tracking, and real-time heatmaps showing agent activity patterns. All dashboard components now support dark mode and are fully responsive for tablet and mobile viewing.

We have also added a new custom report builder that allows teams to create and save their own dashboard views with custom date ranges, filters, and metric combinations. Reports can be scheduled for automated email delivery on a daily, weekly, or monthly cadence.

Bug Fixes

A critical bug affecting webhook delivery reliability has been resolved. In certain edge cases, webhooks were being delivered out of order when multiple events occurred within the same millisecond. The fix introduces a sequence number that guarantees ordered delivery regardless of timing conditions.

We have also fixed an issue where the SDK's automatic reconnection logic could enter an infinite retry loop when the server returned a 403 status during token refresh. The reconnection logic now properly delegates to the authentication refresh flow before retrying the connection.

Security Patches

This release includes important security updates. We have upgraded our TLS configuration to require TLS 1.3 for all API connections, deprecated support for two legacy cipher suites, and implemented certificate transparency logging for all our domains. Additionally, API tokens now support configurable IP allowlists for an extra layer of access control.

A vulnerability in our Markdown rendering pipeline that could allow limited HTML injection in conversation previews has been patched. We have confirmed that this vulnerability was not exploited in production. All affected library dependencies have been updated to their latest secure versions.

New Features

Conversation branching is now available in beta. Users can fork a conversation at any point to explore alternative dialogue paths without losing the original thread. This feature has been particularly popular with teams using Helix for content generation and creative workflows where exploring multiple directions is essential.

We have also introduced a new batch processing API that allows developers to submit multiple agent tasks as a single request. The batch API handles parallelization, retry logic, and result aggregation automatically, making it significantly easier to build workflows that process large volumes of data through Helix agents.