Preuzmi Eigent 1.0.6 za Windows

Eigent v1.0.6 introduces redesigned Automations and model settings, adds a work-role step to onboarding, and makes Sessions more reliable when switching views, reloading, or resuming interrupted work.
Highlights
- Redesigned Automations — Start from six example workflows, see upcoming runs, and inspect results in a dedicated run dashboard. Pause, resume, or request a run from the automation list. Schedule editing now provides clearer local-time previews and validation, while preserving existing schedules the editor cannot represent. Scheduled runs require Eigent to remain open on this computer.
- Unified model configuration — Search configured models, manage provider configurations, and choose a default for new Sessions from Settings → Models. Add or edit models in dedicated dialogs, discover available models or enter a model name manually, and control which Eigent models appear in the selector.
- Work roles in your profile — Choose your role during onboarding or skip the step, then update it later in General settings.
- Clearer task outcomes — See why a task completed, stopped, failed, timed out, or was interrupted. When a task fails without a final reply, inspect its recorded actions and any unconfirmed outcomes.
- Better continuity between Sessions — Keep unsent text and review feedback when switching Sessions, preserve terminal scrollback, and retain line breaks when pasting multiline text.
Session and runtime reliability
- Start new tasks in restored or previously failed Sessions, and create new Sessions while Cloud history is still restoring.
- Send follow-up messages reliably after a window reload or a lost chat stream.
- Restore actionable approvals after reloads and Resume, preserve accepted decisions when requests are interrupted, and keep tasks waiting when other approvals remain open.
- Keep task timers steady through approvals and Session reloads, and synchronize completed states across chat, navigation, and file summaries.
- Improve live task updates and release finished or idle event streams.
- Recover workspace access held by lost task attempts, explain which Session is blocking file changes, and provide clearer guidance when a task cannot safely resume.
- Release Single Agent resources after failed or stopped startup, and complete isolated-workspace finalization when its caller is cancelled.
Model compatibility
- Let models without registered effort capabilities run at their provider default instead of blocking them because of a selected thinking effort. Models using legacy inferred capabilities fall back to the nearest supported effort; explicitly registered models still enforce their declared capabilities.
- Add GPT-6 Sol effort declarations for OpenAI and Azure, fix startup with aliased model providers, and explain unsupported effort selections and provider outages more clearly.
Files, Spaces, and usability
- Fix blank PDF previews in Files and offer an external-open action when an embedded PDF fails to load.
- Open local files from Session summaries, report preview loading states accurately, and keep browser previews associated with the correct task.
- Show direct task checkpoints in version history.
- Improve Bundle import recovery, add resource-reference checks before publishing, and make sharing selections easier to read.
- Clean up deleted Session data, with an explicit option to remove an eligible Session working directory and warnings for unpublished changes.
- Avoid saving one-off requests as long-term memory merely because they mention a Space.
- Correct Home relative timestamps, default Session names, trial usage messages, browser-port validation, and explanations of what happens when you quit with tasks running.
- Strengthen authorization checks on legacy local-backend routes and keep Bundle installation proposals tied to their originating Space.
Full changelog: v1.0.5 → v1.0.6