vikingowl
49531864f1
Phase 1 of Conductor integration - DAG-based plan execution support:
Types:
- Add StepExecutionResult enum (Success, Failed, Skipped, Pending, InProgress)
- Add SkipReason enum for tracking why steps were skipped
- Add PauseReason enum for execution pause states
- Add DagValidationResult enum for validation outcomes
PlanStep extensions:
- Add depends_on: Vec<String> for explicit step dependencies
- Add execution_result: Option<StepExecutionResult> for tracking
- Add retry_count: u8 for failure retry tracking
AccumulatedPlanStatus extensions:
- Add ExecutingParallel { active_count, remaining } variant
- Add Paused(PauseReason) variant
- Add Aborted { reason } variant
DAG validation (Kahn's algorithm):
- validate_dag() returns execution order with parallelism levels
- Detects cycles and missing dependencies
- dependencies_satisfied() checks if step can execute
- get_ready_steps() returns executable steps
- skip_with_dependents() cascades failure through graph
Tests: 12 new DAG validation tests covering chains, parallelism,
diamonds, cycles, missing deps, and skip cascading.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
Owlen UI
A Terminal User Interface (TUI) for the Owlen AI agent, built with Ratatui.
Features
- Rich Text Rendering: Markdown support with syntax highlighting for code blocks.
- Interactive Components: Intuitive panels for chat, tool execution, and session status.
- Real-time Streaming: Smooth display of agent output as it's generated.
- Task Visualization: Dedicated view for tracking the agent's progress through a task list.
Architecture
The UI is built using an event-driven architecture integrated with the agent-core event stream. It leverages ratatui for terminal rendering and crossterm for event handling.
Components
ChatPanel: Displays the conversation history.TaskPanel: Shows the current implementation plan and task status.ToolPanel: Visualizes active tool executions and their output.ModelPicker: Allows selecting between available LLM providers and models.