- Add comprehensive documentation in botbook/ with 12 chapters - Add botapp/ Tauri desktop application - Add botdevice/ IoT device support - Add botlib/ shared library crate - Add botmodels/ Python ML models service - Add botplugin/ browser extension - Add botserver/ reorganized server code - Add bottemplates/ bot templates - Add bottest/ integration tests - Add botui/ web UI server - Add CI/CD workflows in .forgejo/workflows/ - Add AGENTS.md and PROD.md documentation - Add dependency management scripts (DEPENDENCIES.sh/ps1) - Remove legacy src/ structure and migrations - Clean up temporary and backup files
1.4 KiB
1.4 KiB
Multi-Agent Workflows Guide
Creating Workflows
Basic Workflow Structure
ORCHESTRATE WORKFLOW "workflow-name"
STEP 1: BOT "analyzer" "process input"
STEP 2: BOT "validator" "check results"
END WORKFLOW
Human Approval Integration
STEP 3: HUMAN APPROVAL FROM "manager@company.com"
TIMEOUT 1800 ' 30 minutes
ON TIMEOUT: ESCALATE TO "director@company.com"
Parallel Processing
STEP 4: PARALLEL
BRANCH A: BOT "processor-1" "handle batch-a"
BRANCH B: BOT "processor-2" "handle batch-b"
END PARALLEL
Event-Driven Coordination
ON EVENT "data-ready" DO
CONTINUE WORKFLOW AT STEP 5
END ON
PUBLISH EVENT "processing-complete"
Cross-Bot Memory Sharing
BOT SHARE MEMORY "successful-patterns" WITH "learning-bot"
BOT SYNC MEMORY FROM "master-knowledge-bot"
Best Practices
- Keep workflows focused - Max 10 steps per workflow
- Use meaningful names - Clear bot and step names
- Add timeouts - Always set timeouts for human approvals
- Share knowledge - Use memory sharing for bot learning
- Handle events - Use event system for loose coupling
Workflow Persistence
Workflows automatically survive server restarts. State is stored in PostgreSQL and recovered on startup.
Visual Designer
Use the drag-and-drop designer at /designer/workflow to create workflows visually. The designer generates BASIC code automatically.