Skip to content
- Improved Workday RaaS connector efficiency and reliability with
operation-specific report loading, bounded retries for throttling and
transient server failures, and duplicate membership handling.
- Added a tested users-only Workday setup, copyable mock payloads, optional
group examples, field-mapping guidance, and a client acceptance checklist.
- Added OpenRouter support for Kimi K3 and DeepSeek V4 Pro.
- Improved the reliability and performance of filtered LDAP group-member
searches.
- Fixed Google Workspace refreshes so group membership synchronization no
longer replaces updated user names with email addresses.
- Hardened the local PostgreSQL development workflow against accidentally
inherited database credentials.
- Added Enterprise observability through standard OpenTelemetry traces and
metrics, with an admin System health view and matching REST and MCP status
checks.
- Added visibility into database health, active Roster instances, worker
backlog, telemetry delivery, and dropped batches.
- Resolve activity now creates Model Run records for administrator
troubleshooting, including input and output by default. Administrators can
redact Model Run input/output, tool payloads, actor details, and errors in
Settings. Configured Model Run retention is enforced automatically, and tool
payload redaction also covers function calls embedded in stored output.
- Added trace links to Model Runs and model-provider token usage. External
telemetry does not include prompts, model output, HTTP bodies, credentials,
or user-provided content.
- Added guidance and a local reference stack for connecting Roster to a
customer-managed observability platform. Telemetry storage, dashboards, SLO
evaluation, and alerting remain under the customer’s control.
- Structured logs, health probes, and trace correlation remain available in
every edition. Free and Pro do not send external telemetry or expose advanced
observability status.
- Hardened observability and MCP failure handling so exporter or status-query
failures do not interrupt Roster, untrusted request IDs are not exported, and
oversized MCP requests are rejected safely.
- Strengthened safety controls for SAML sign-in redirects, audit-event queries,
platform usage information, and MCP update operations.
- Improved LDAP directory-group handling and protected interactive association
from excessively large or slow membership lookups.
- Corrected project-owner updates and clarified Enterprise high-availability
availability in deployment documentation.
- Hardened human-readable terminal output when Roster data contains terminal
control characters. JSON output remains unchanged for automation.
- Added PostgreSQL 17 and 18 support for Enterprise deployments, while keeping
SQLite as the default for standalone installations.
- Added high-availability deployment support with multiple Roster instances,
PostgreSQL failover, read replicas, and load-balancer health checks.
- Added verified online SQLite backups, streamed hosted backups, and complete
backup, verification, and restore procedures.
- Improved multi-instance request limits, background-job recovery, database
connection handling, and safer release-time PostgreSQL migrations.
- Added clearer database, hosting, backup, restore, and high-availability setup
guidance for production operators.
- Existing SQLite deployments remain supported with the same
/data volume.
This release does not include an automated SQLite-to-PostgreSQL data migration.
- Aligned the public CLI package with Roster 1.1.0.
- Existing CLI commands, OAuth profiles, and API-key automation remain
compatible. Database selection and migration are managed by the Roster
server and require no CLI configuration change.
- Added configurable first-login password-change behavior. The default reminder
can be postponed for the current session and is shown again in a later
session until the password is changed.
- Added
required and off options for deployments that need mandatory
rotation or no reminder.
- Existing Docker installations can reuse their configuration and
/data
volume without a manual migration.
- Strengthened authentication, authorization, initial administrator setup, and
connector safety.
- Enforced the required first-login password change before protected actions.
- Retired the temporary Preview edition and preview Docker image; production
remains available as Free, Pro, and Enterprise from the stable Roster image.
- Existing Docker installations can reuse their current configuration and
/data volume without a manual data migration. Keep ROSTER_AUTH_SECRET
unchanged and back up /data before upgrading. Review the
1.0.3 upgrade notes
before changing the image version.
- Improved Resolve model response handling for more reliable participant routing.
- Simplified model provider setup guidance.
- Added broader OpenRouter gateway support for Resolve model providers.
- Updated model provider documentation with supported model and gateway
guidance.
- Updated the public CLI package and setup documentation.
- Published the first stable Roster release.
- Supports Free, Pro, and Enterprise editions.
- Published the stable Roster CLI package.