List view

• Sprint 26: Eigent v1.0.2 Model Control & Task UX

  This milestone tracks the new features planned for Eigent v1.0.2. The main focus is to give users and agents better control over model behavior, task runtime, human-in-the-loop permissions, feedback signals, and workspace-level model setup. ## Focus Areas ### 1. Model Thinking Effort Add support for configurable model thinking effort levels, such as low, medium, high, extra high, and max. This gives users more control over the tradeoff between speed, cost, and reasoning depth. ### 2. Task Timeout Design Improve task timeout behavior and align it with competitor expectations. Long-running tasks should not be stopped too early by a fixed timeout when the model or agent still needs more time. Timeout behavior should better support complex tasks, main-agent/sub-agent coordination, and user-visible task status. ### 3. Human Toolkit Controls Add user-facing controls for Human Toolkit availability and dangerous mode. Users should have clearer control over when human-in-the-loop tools are enabled, and when higher-risk actions require stronger permission or explicit opt-in. ### 4. Reward Signal Add thumbs up / thumbs down feedback signals. This gives users a lightweight way to provide feedback on agent outputs, task results, and interaction quality. ### 5. Workspace Model Setup Improve Workspace model setup so the selected model better matches each user-selected use case. The goal is to make model configuration feel more contextual, reliable, and aligned with what the user is trying to do. ## Tracking This milestone should contain: - Requirement issues for the v1.0.2 new features - Design discussion issues where behavior is not finalized - Implementation PRs for Sprint 26 - Follow-up fixes found while validating these features Implementation details can be refined during development.

  Due by July 10, 2026

  •1/22 issues closed
  4% complete21 open1 closed

• Linux Focus

  No due date

  •0/5 issues closed
  0% complete5 open0 closed