Demo Skill

Demonstrates UltraRAG-style execution tracing with LangGraph + YAML pipelines.

Tools

hello

Simple demo command for testing hot reload.

Parameters:

• name (string, optional): Name to greet (default: “Guest”)

echo

Echo back the input message.

Parameters:

• message (string, optional): Message to echo (default: “Hello!”)

test_yaml_pipeline

Test YAML pipeline compilation and execution with omni.tracer.

Parameters:

• pipeline_type (string, optional): Type of pipeline to test.
  • "simple": Sequential pipeline (analyze → draft → finalize)
  • "loop": Pipeline with iteration loop (analyze → evaluate → reflect)
  • "branch": Pipeline with conditional branching
  • "rag": Full RAG pipeline with retrieval

run_langgraph

Execute packaged graphflow runtime with trace output and quality gates.

Parameters:

• scenario (string, optional): Scenario name ("simple" or "complex")
• quality_threshold (string, optional): Draft/finalize quality threshold
• quality_gate_novelty_threshold (string, optional): Gate novelty threshold
• quality_gate_coverage_threshold (string, optional): Gate coverage threshold
• quality_gate_min_evidence_count (string, optional): Minimum evidence count
• quality_gate_require_tradeoff (string, optional): Require trade-off in output
• quality_gate_max_fail_streak (string, optional): Max gate-fail streak before force-draft

list_pipeline_examples

List available YAML pipeline test examples.

Returns:

• Available pipeline types and their descriptions
• Usage instructions

Pipeline Examples

The demo skill includes YAML pipeline files in pipelines/ directory:

YAML Pipeline Structure

servers:
  generator: builtin

parameters:
  topic: "$topic"

pipeline:
  - generator.analyze
  - generator.draft
  - generator.finalize

runtime:
  checkpointer:
    kind: memory
  tracer:
    callback_dispatch_mode: inline

UltraRAG Features

This demo showcases:

• YAML pipelines – Declarative workflow definitions
• Loop control – Iterative reflection with max_iterations
• Conditional branching – Router-based routing
• Memory tracking – Variables saved with history
• Rich output – Colored step-by-step tracing

Usage

# Simple greeting
omni skill run demo.hello --name "World"

# Echo test
omni skill run demo.echo --message "Test message"

# List available pipeline examples
omni skill run demo.list_pipeline_examples

# Test YAML pipelines
omni skill run demo.test_yaml_pipeline --pipeline_type "simple"
omni skill run demo.test_yaml_pipeline --pipeline_type "loop"
omni skill run demo.test_yaml_pipeline --pipeline_type "branch"
omni skill run demo.test_yaml_pipeline --pipeline_type "rag"

# Run packaged graphflow runtime
omni skill run demo.run_langgraph --scenario "complex"