[AI Agents] Added ProjectId and OrganizationId to all OpenAI requests
[AI Agents] Added Temperature setting for OpenAI and Azure OpenAI
[AI Agents] Added support for Azure OpenAI
[AI Agents] Added support for OpenRouter via the OpenAI compatible API
[AI Agents] Implemented server-side streaming for AI Agent responses
[AI Agents] Utilized a new request handler for queries within AI Agents
[Audit Log] Fixed audit log entry for DeleteByQuery and PatchByQuery operations to correctly log the query
[Audit Log] Fixed logging on connection closed
[Backup] Fixed an issue that caused restoring an encrypted backup to fail if audit logging was enabled
[Backup] Fixed listing of AWS S3 restore points
[Backup] Improved next backup time calculation for missed schedules
[Backup] Refactored backup status reporting to be stored locally on each node and consolidated by the cluster observer
[Backup/Restore] Added Max Read Operations Per Second option to backup, restore, import, and export operations
[Certificates] Added AcmeRenewalThresholdPercentage configuration option to control the automatic server certificate renewal threshold (default 30%)
[Certificates] Added support for server certificates with only a Server Authentication EKU
[Certificates] Generated client certificates for pull replication with client EKU
[Cluster] Added Cluster.FullReportIntervalInMin configuration option to force sending a full cluster node status report periodically (default 5 minutes)
[Corax] Enhanced the binary match query builder for more complex query combinations
[Corax] Improved indexing performance by processing posting lists and lookup trees separately
[Counters] Fixed an issue in the counters repair task related to context disposal
[Debug] Added periodic backup information to the debug package
[Debug] Added the total allocated to the allocations event listener event
[Debug] Fixed Debug Package information output for sharded databases
[Documents] Improved handling of document IDs with non-ASCII and escaped characters, including a tool to clean corrupted IDs
[Embeddings Generation] Added OverlapTokens option to paragraph-based text chunking for embeddings generation
[Embeddings Generation] Improved reliability of the embeddings generation
[GenAI] Fixed exception handling in Gen AI tasks
[GenAI] Refactored the message format for Generative AI chat
[Indexes] Changed index idling to be based on time since last query instead of an in-memory value
[Indexes] Corax search engine indexes are now skipped during a database compact operation
[Indexes] Fixed index name replacement in the 'Test Index' feature to avoid substring conflicts
[Licensing] Fixed subscription count calculation for licensing when database keys are missing
[Memory] Improved unmanaged memory management for stream operations to reduce allocations
[Operations] Added an endpoint to restart a single database
[PAL] Improved accuracy of storage and memory calculations on macOS
[Performance] Optimized XXHash implementation using AVX-512 instructions where available
[Performance] Optimized load document handler to reduce async calls and memory allocations
[Querying] Allowed vector search on fields populated by an Embeddings Generation task
[Querying] Ensured read transactions for streaming queries are disposed of explicitly to prevent leaks
[Setup] Using just a file name in Security.Certificate.Path during the Setup Wizard
[Smuggler] Increased the maximum number of statements allowed when validating smuggler transformation scripts
[SNMP] Added SNMP endpoints for machine load average (1m, 5m, 15m)
[Time Series] Fixed restoring of incremental time series with dead values
[Tools] Allowed trusting server certificates with a single server authentication EKU via the rvn tool
[Voron] Fixed a race condition where the oldest active transaction could report a non-zero value when no transactions were active
[Voron] Optimized compression buffer handling for encrypted databases to reduce disk I/O