HumanLayer integration

[TerminallyLazy]

🤖➡️👤 HumanLayer Human-in-the-Loop Integration

Summary

This PR adds complete HumanLayer integration to Agent Zero, enabling human-in-the-loop workflows for approvals, consultations, and multi-channel communication. The integration follows Agent Zero's
architectural patterns and provides a seamless user experience through the Settings UI.

Features Added

🔧 Core Integration

• HumanLayer SDK Integration: Added humanlayer>=0.3.0 dependency with AsyncHumanLayer client support
• HumanLayerClient Wrapper: Comprehensive client wrapper handling configuration, validation, and error management
• Conditional Loading: Extension system only loads HumanLayer tools when humanlayer_enabled=True

🛠️ Tools Implementation

• Approval Tool (humanlayer_approval.py): Request human approval for potentially risky operations
• Consultation Tool (humanlayer_consultation.py): Ask humans questions with context for decision-making
• Contact Management Tool (humanlayer_contact.py): Validate and manage communication channels
• Enhanced Code Execution: Automatic approval requests for high-risk code operations

⚙️ Settings UI Integration

• Complete Settings Panel: 11 configuration fields in External Services tab
• User-Friendly Interface: Toggles, dropdowns, password fields, and number inputs
• Multi-Channel Support: Slack and Email configuration with validation
• API Endpoints: Configuration validation and testing endpoints
• Visual Elements: Custom HumanLayer icon and proper UI integration

🏗️ Architecture Compliance

• Agent Zero Patterns: All tools extend Tool base class and use Response objects
• Async/Await Support: Full async compatibility throughout the integration
• Extension System: Uses Agent Zero's lifecycle hooks for conditional functionality
• Configuration Flow: Settings automatically flow from UI to agent's additional config
• Graceful Fallbacks: Proper error handling when HumanLayer is disabled or misconfigured

Configuration

Users can configure HumanLayer through Settings → External Services with:

Core Settings:

• Enable/disable integration toggle
• API key (password field)
• Default contact channel selection
• Approval timeout configuration
• Verbose logging toggle

Channel Configuration:

• Slack: Channel ID, context description, Block Kit support
• Email: Address, subject line, context description

Usage Examples

Once configured, HumanLayer automatically integrates into Agent Zero workflows:

# Automatic approval for risky operations
agent.execute_code("rm -rf /tmp/old_files")  # → Requests human approval

# Manual consultation requests
agent.consult_human("Should I proceed with database migration?")

# Approval workflows
agent.request_approval("Delete 1000 old log entries", context="Disk cleanup")

Testing

- ✅ Comprehensive unit test suite for all tools and integrations
- ✅ Integration tests validating end-to-end functionality
- ✅ Configuration validation and error handling tests
- ✅ Syntax and type checking validation
- ✅ API endpoint testing for validation and testing features

Files Changed

Core Integration:
- requirements.txt - Added HumanLayer dependency
- python/helpers/humanlayer_client.py - HumanLayer SDK wrapper
- python/tools/humanlayer_*.py - Three main tools (approval, consultation, contact)
- python/tools/code_execution_tool.py - Enhanced with auto-approval

Extensions:
- python/extensions/system_prompt/_30_humanlayer_prompts.py - Conditional prompt loading
- python/extensions/before_main_llm_call/_15_humanlayer_check.py - Pre-execution monitoring

Settings Integration:
- python/helpers/settings.py - Added HumanLayer settings schema and UI fields
- initialize.py - Pass HumanLayer settings to agent configuration
- python/api/humanlayer_validate.py - Configuration validation endpoint
- python/api/humanlayer_test.py - Tool testing endpoint

UI Components:
- webui/public/humanlayer.svg - HumanLayer icon
- agents/agent0/prompts/agent.system.tool.humanlayer_*.md - Tool prompt files

Testing:
- tests/test_humanlayer_*.py - Comprehensive test suite

Breaking Changes

None. This is a purely additive feature that doesn't affect existing functionality when disabled.

Migration Notes

No migration required. The integration is disabled by default and can be enabled through the Settings UI.

---
🎉 Ready for human-in-the-loop AI workflows! This integration brings human oversight and collaboration directly into Agent Zero's autonomous operations.