Repository navigation
Releases: lstack-ai/beta
Releases · lstack-ai/beta
Release list
v0.4.1
- Added guided setup for model providers and workers.
- Added connection, capability and Claude subscription checks during setup.
- Improved connection errors and worker startup status messages.
- Reduced false stops during slow model responses, retries and ongoing work.
- Removed default time limits for new workers while preserving saved limits.
- Improved loop detection and cancellation recovery.
- Fixed repeated rebuilds and false health failures for slow-starting project apps.
- Fixed outdated health status being reused after app restarts.
v0.4.0
- Added conversational ticket planning with read-only project access.
- Added support for planning with workers not assigned to the project.
- Added editable ticket drafts, default draft selection, and individual or bulk creation in Backlog.
- Enlarged the Create Ticket window and improved text-field sizing, navigation, parent-ticket selection and worker selection.
- Increased the planning reply timeout to 30 minutes and improved progress feedback.
- Improved recovery from interrupted planning sessions and restarts.
- Disabled unnecessary Anthropic background requests by default when using other model providers.
- Fixed Claude worker startup failures affecting planning, execution and readiness checks, and corrected misleading credential error messages.
v0.3.1
- Added Continue without an account during initial setup for local use. The choice is remembered across browsers and restarts on the same computer.
- Added a Sign In action in the sidebar when using LStack without an account, so you can sign in or create an account later.
- Once you confirm an account as the installation's owner, sign-in is required and the option to continue without an account is no longer available. Existing account-protected installations keep their current sign-in behavior.
- Improved the account setup flow from network settings: you return there after confirming ownership, with network access still off until you enable it. Local network (LAN) access continues to require the owner's account.
v0.3.0
- Added account creation with email verification and a guided path to setting up your installation.
- Improved password handling on sign-in and sign-up forms.
- Improved compatibility between workers and local model servers such as vLLM, while preserving explicitly configured reasoning preferences.
- Improved text, vision and tool capability checks for reasoning models, with clearer results when a check cannot determine support.
- Applying detected capabilities now preserves existing settings when a check is inconclusive, fails or was not run.
v0.2.6
- Added Test model connection on a worker's Overview to check connectivity for supported providers without creating a project or starting an AI agent session.
- Improved handling of localhost addresses for Docker workers, with clearer guidance when a local model server cannot be reached.
- Updated reviewer guidance to use checks available to each project, including Docker Compose apps, while keeping acceptance decisions with the reviewer.
- Fixed review sessions continuing after a decision, so workers can move on to the next ticket once the session has stopped.
v0.2.5
- Added a dashboard notice when workers need updating, with a link to the existing update controls.
- Kept service logs available after failed startup and improved troubleshooting details for multi-service apps.
- Added command-result capture for Docker Compose projects.
- Improved review messages, recovery summaries, and next-step guidance.
- Improved automated reviews with reviewer-led decisions and complete acceptance-criteria assessments.
- Added automatic follow-up when a review ends without a complete decision, with improved recovery from interrupted sessions.
- Before upgrading, let current work finish. After upgrading, build the updated worker image and recreate idle workers in Settings → Workers before starting new work.
v0.2.4
- Improved reliability when tickets retry or resume work.
- Reduced premature automatic stops during ticket work.
- Added Resume automatic recovery under Runs → Recovery actions, so paused recovery can continue without discarding existing work.
- Made ticket views clearer about whether work is running, blocked, paused, or not queued, and when attention is needed.
- Corrected run history to distinguish automatic stops from user-requested stops.
- Before upgrading, let current work finish. After upgrading, build the updated worker image and recreate idle workers in Settings → Workers before starting new work.
v0.2.3
- Reduced delays when starting tickets after assigning workers to projects.
- Simplified worker settings and improved status updates.
- Fixed Hold and Stop confirmations, with clearer feedback and more reliable ticket updates.
- Fixed View board leaving the investigation dialog open.
- Updated website guidance for sign-in, public downloads, and LAN access.
v0.2.2
- Fixed workers losing access to LStack tools between tickets, which could prevent builders from submitting work and reviewers from recording decisions.
- Stopped runs from continuing when required LStack tools fail to initialize, with clearer failure and recovery messages.
- Enabled Reset ticket attempt for exhausted recovery pauses while preserving protections for unsaved work.
- Existing workers require the updated worker image before accepting new assignments; build the image and recreate workers from the dashboard after upgrading.
- Simplified comments in the public installer without changing installation behavior.
v0.2.1
- Removed GitHub CLI and GitHub sign-in requirements from installation and updates, and simplified installation guidance.
- Fixed LAN port changes leaving the browser on a forbidden error instead of returning to sign-in.
- Made project lists, counts, and board project filters refresh automatically across open windows and devices.
- Added provider readiness checks that verify real worker execution separately from connectivity, with cleanup and recovery for successful and failed checks.
- Protected uncommitted ticket work when Git capture fails, with clear pause reasons and a Retry capture action.
- Improved failed-run diagnostics with the engine's error details and the affected execution context.
- Fixed dispatch of tickets explicitly assigned to an allowed worker.