ORCH + Fabric — orchestrate Fabric AI patterns as a multi-agent pipeline with state machine

[oxgeneral]

Hey Fabric community!

I built ORCH — a CLI runtime that turns multiple AI agent processes into a coordinated team. It pairs naturally with Fabric's pattern-based AI workflows.

The use case: Chain multiple Fabric patterns across specialized agents. Instead of running patterns sequentially in a script, ORCH distributes them across workers with automatic retry, state tracking, and dependency management.

# Add Fabric pattern agents to ORCH
orch agent add "summarizer" --adapter shell \
  --cmd "fabric --pattern summarize"
orch agent add "extractor" --adapter shell \
  --cmd "fabric --pattern extract_wisdom"
orch agent add "analyst" --adapter shell \
  --cmd "fabric --pattern analyze_claims"

# Task routing with dependencies
orch task add "Summarize research paper" --assignee summarizer
orch task add "Extract key insights" --assignee extractor --depends-on <summarize-task>
orch task add "Validate claims" --assignee analyst --depends-on <extract-task>

# Real-time pipeline dashboard
orch tui

State machine enforcement: todo → in_progress → review → done — each pattern result goes through a review gate before the next stage starts. Failed patterns auto-retry with backoff.

Fabric handles pattern execution. ORCH handles the team coordination, sequencing, and resilience layer.

→ GitHub: https://github.com/oxgeneral/ORCH
→ Install: npm install -g @oxgeneral/orch

Would love to hear how Fabric users are currently chaining patterns across multiple processes!