🌀Changelog

What's New in Helix

Track our journey as we build the future of intelligent multi-agent AI. Every update brings us closer to the singularity. 🚀

v21.0.0Major Release2026-03-11

Security Hardening, AI Governance & Knowledge Graph

Highlights

Cross-conversation Knowledge Graph with LLM-powered fact extraction
Safety & Governance dashboard — first platform to surface AI ethics controls to users
Agent Reputation leaderboard with per-agent satisfaction tracking
Comprehensive security audit: tier gates, rate limits, auth hardening across 50+ files

Added

•Knowledge Graph: LLM extracts (subject, predicate, object) facts from every conversation, deduplicates, indexes for semantic retrieval, and injects into copilot context automatically
•/memory page with Facts, Timeline, and Search tabs — browse, search, and delete extracted knowledge
•/safety page: Pinned Safety Constraints (6 immutable guardrails), AI Safety Framework (4 ethical principles), and Ethics Audit Log
•Agent Reputation tab on /agents: sortable leaderboard with satisfaction %, avg rating, response time, and summary stats
•Voice input in chat: Web Speech API primary, MediaRecorder-to-backend fallback
•TTS playback on agent messages via edge-tts self-hosted synthesis
•Provenance-linked responses: every agent message shows "based on" sources (memory, search, UCF)
•Code execution sandbox on /agentic: Monaco editor, Python/JS/Shell, Ctrl+Enter execution, inline output
•Agentic LLM loop: "Run Autonomously" button starts multi-step code generation with progress tracking
•Webhook management UI on /spirals: register, list, copy URL, test webhooks
•Background tasks dashboard widget with task stats and recent task list
•Integrations Hub on /settings showing platform integration connection status
•Agent Builder wizard on /agents/builder: 3-step template-based agent creation
•Message rating UI: thumbs up/down on every agent response in chat
•Spiral trigger scheduler wiring: cron, interval, and UCF-threshold triggers activate immediately on save
•"What Makes Helix Different" feature-discovery section on dashboard

Improved

•credentials:include added to all fetch() calls across 20+ frontend files — fixes auth cookie propagation
•Silent catch blocks eliminated: 24+ backend except:pass and 12+ frontend empty catch blocks now log properly
•Fake metrics replaced with zeros + _simulated/_default flags across 4 backend files
•In-memory user state migrated to Redis: wellness, templates, MCP gateway, OAuth CSRF, dash widgets, forum AI, moderation, Discord linking, integrations hub
•Agent performance logging rewritten to async (asyncpg) and wired into both copilot /message and /message/stream endpoints
•GDPR erasure path now covers knowledge facts, all memory tiers, and conversation data
•Gradient text AI-slop cleanup: removed decorative gradient headings from 9+ pages per design system rules
•AgentSelector shows live satisfaction % and rated count per agent card

Fixed

•Anonymous tier bypass: inverted logic in Council and Chain endpoints let unauthenticated users skip tier checks
•MCP Gateway plan escalation: client-submitted plan field was trusted — now forced to "free" server-side
•Zapier auth bypass: ImportError fallback returned None instead of raising HTTP 503
•LLM routes (/api/llm/generate, /api/llm/train) were completely unauthenticated — now gated with require_starter
•Unbounded LLM routing cache: _active_profiles_cache now capped at 5,000 with FIFO eviction
•TypeScript optional chaining errors in spirals/page.tsx (existing.config → existing?.config)
•Duplicate credentials headers in 6 frontend files from linter overlap

v20.0.0Major Release2026-03-10

Tier Gates, Rate Limits & Auth Hardening

Highlights

Every API endpoint now has proper authentication and subscription tier enforcement
Rate limiting deployed across all execution and write endpoints
Enterprise-grade security posture — no unauthenticated endpoints remain

Added

•Tier gates: Code execution (Starter+), Workflow engine (Pro+), Agent orchestrator (Pro+), Agent builder (Starter+), GitHub IDE writes (Starter+), Multimedia/Meme (Hobby+)
•Rate limiting decorators on 30+ endpoints: execution at 5-10/min, reads at 30/min
•Webhook registry persistence: moved from in-memory dict to Redis hash (helix:webhook_registry)
•Background tasks persistence: task records now persisted to Redis with 7-day TTL

Improved

•Router registry reliability with _safe_include() pattern for graceful fallbacks
•CLAUDE.md expanded with architecture rules: no in-memory user state, persistence hierarchy, error handling standards

Fixed

•Copilot performance logging wired into both message endpoints with timing

v2.0.0Major Release2026-01-20

Production Ready Release

Highlights

100% production-ready milestone
Agent Capabilities System with secure sandboxed execution
12 MCP Servers compiled and production-ready
Discord Voice Features with TTS support

Added

•Agent Capabilities System: Execution Engine, Tool Framework, Memory System, Multimodal Processing
•Discord Bot Enhancement: 20+ slash commands, voice channel integration, TTS support
•Helix Spirals Workflow Automation: 16+ node types, Google Workspace integration
•Health Dashboard with real-time system metrics
•12 MCP Servers: memory, coordination, infrastructure, code, filesystem, and more

Improved

•Frontend accessibility with ARIA attributes
•Error handling with exponential backoff
•TypeScript type safety across components

Fixed

•Python syntax error in system_coordination_core.py
•WebSocket connection stability

v1.9.0Feature Update2026-01-14

Code Quality & Modularization

Highlights

Main.py reduced by 51% (5,582 → 2,749 lines)
33 endpoints extracted into 5 modular route files
Zero breaking changes with all 524+ routes maintained

Added

•stripe_routes.py - 7 Stripe billing endpoints (320 lines)
•arjuna_routes.py - 7 Arjuna Portal portal endpoints (500 lines)
•coordination_api_routes.py - 10 coordination streaming endpoints (685 lines)
•mandelbrot_routes.py - 5 Mandelbrot UCF generation endpoints (180 lines)
•legacy_standalone_routes.py - 6 legacy endpoints (507 lines)

Improved

•Code organization and maintainability
•Testability (can test modules independently)
•Graceful import fallbacks for optional dependencies

v1.8.0Major Release2024-12-22

Multi-Agent Chat & Analytics Dashboard

Highlights

Real-time WebSocket-based chat interface
8-qubit system state visualization
Discord multi-agent integration with 10+ commands

Added

•Frontend Chat System (1,200+ lines): Agent selector, performance indicators, emoji picker
•System Dashboard (1,500+ lines): Bloch spheres, entanglement matrix, routine monitor
•Discord Integration: Interactive agent selection, conversation threading, 16 languages

Improved

•UsageDashboard with authentication headers and retry logic
•Component reusability and accessibility
•Rendering performance optimization

Fixed

•Fixed Python syntax error in system_coordination_core.py
•Enhanced WebSocket connection stability

v1.7.0Feature Update2024-12-15

Voice Patrol & LLM-Powered Agents

Highlights

Discord voice channel patrol with themed agent personalities
All 24 agents now have intelligent, contextual LLM responses
Supports 4 LLM providers: Ollama, Anthropic, OpenAI, Custom

Added

•Voice Patrol System: Auto-join, patrol loop, 5 agent voice profiles
•LLM Agent Engine (700+ lines): Unique system prompts per agent
•Commands: !voice-join, !voice-leave, !voice-announce, !voice-status

Improved

•Agent personality consistency across platforms
•Conversation history per session

Stay Updated

Get notified about new features and improvements.

Create Free Account ⭐ Star on GitHub

What's New in Helix

Track our journey as we build the future of intelligent multi-agent AI. Every update brings us closer to the singularity. 🚀

v21.0.0Major Release2026-03-11

Security Hardening, AI Governance & Knowledge Graph

Highlights

Cross-conversation Knowledge Graph with LLM-powered fact extraction
Safety & Governance dashboard — first platform to surface AI ethics controls to users
Agent Reputation leaderboard with per-agent satisfaction tracking
Comprehensive security audit: tier gates, rate limits, auth hardening across 50+ files

Added

•Knowledge Graph: LLM extracts (subject, predicate, object) facts from every conversation, deduplicates, indexes for semantic retrieval, and injects into copilot context automatically
•/memory page with Facts, Timeline, and Search tabs — browse, search, and delete extracted knowledge
•/safety page: Pinned Safety Constraints (6 immutable guardrails), AI Safety Framework (4 ethical principles), and Ethics Audit Log
•Agent Reputation tab on /agents: sortable leaderboard with satisfaction %, avg rating, response time, and summary stats
•Voice input in chat: Web Speech API primary, MediaRecorder-to-backend fallback
•TTS playback on agent messages via edge-tts self-hosted synthesis
•Provenance-linked responses: every agent message shows "based on" sources (memory, search, UCF)
•Code execution sandbox on /agentic: Monaco editor, Python/JS/Shell, Ctrl+Enter execution, inline output
•Agentic LLM loop: "Run Autonomously" button starts multi-step code generation with progress tracking
•Webhook management UI on /spirals: register, list, copy URL, test webhooks
•Background tasks dashboard widget with task stats and recent task list
•Integrations Hub on /settings showing platform integration connection status
•Agent Builder wizard on /agents/builder: 3-step template-based agent creation
•Message rating UI: thumbs up/down on every agent response in chat
•Spiral trigger scheduler wiring: cron, interval, and UCF-threshold triggers activate immediately on save
•"What Makes Helix Different" feature-discovery section on dashboard

Improved

•credentials:include added to all fetch() calls across 20+ frontend files — fixes auth cookie propagation
•Silent catch blocks eliminated: 24+ backend except:pass and 12+ frontend empty catch blocks now log properly
•Fake metrics replaced with zeros + _simulated/_default flags across 4 backend files
•In-memory user state migrated to Redis: wellness, templates, MCP gateway, OAuth CSRF, dash widgets, forum AI, moderation, Discord linking, integrations hub
•Agent performance logging rewritten to async (asyncpg) and wired into both copilot /message and /message/stream endpoints
•GDPR erasure path now covers knowledge facts, all memory tiers, and conversation data
•Gradient text AI-slop cleanup: removed decorative gradient headings from 9+ pages per design system rules
•AgentSelector shows live satisfaction % and rated count per agent card

Fixed

•Anonymous tier bypass: inverted logic in Council and Chain endpoints let unauthenticated users skip tier checks
•MCP Gateway plan escalation: client-submitted plan field was trusted — now forced to "free" server-side
•Zapier auth bypass: ImportError fallback returned None instead of raising HTTP 503
•LLM routes (/api/llm/generate, /api/llm/train) were completely unauthenticated — now gated with require_starter
•Unbounded LLM routing cache: _active_profiles_cache now capped at 5,000 with FIFO eviction
•TypeScript optional chaining errors in spirals/page.tsx (existing.config → existing?.config)
•Duplicate credentials headers in 6 frontend files from linter overlap

v20.0.0Major Release2026-03-10

Tier Gates, Rate Limits & Auth Hardening

Highlights

Every API endpoint now has proper authentication and subscription tier enforcement
Rate limiting deployed across all execution and write endpoints
Enterprise-grade security posture — no unauthenticated endpoints remain

Added

•Tier gates: Code execution (Starter+), Workflow engine (Pro+), Agent orchestrator (Pro+), Agent builder (Starter+), GitHub IDE writes (Starter+), Multimedia/Meme (Hobby+)
•Rate limiting decorators on 30+ endpoints: execution at 5-10/min, reads at 30/min
•Webhook registry persistence: moved from in-memory dict to Redis hash (helix:webhook_registry)
•Background tasks persistence: task records now persisted to Redis with 7-day TTL

Improved

•Router registry reliability with _safe_include() pattern for graceful fallbacks
•CLAUDE.md expanded with architecture rules: no in-memory user state, persistence hierarchy, error handling standards

Fixed

•Copilot performance logging wired into both message endpoints with timing

v2.0.0Major Release2026-01-20

Production Ready Release

Highlights

100% production-ready milestone
Agent Capabilities System with secure sandboxed execution
12 MCP Servers compiled and production-ready
Discord Voice Features with TTS support

Added

•Agent Capabilities System: Execution Engine, Tool Framework, Memory System, Multimodal Processing
•Discord Bot Enhancement: 20+ slash commands, voice channel integration, TTS support
•Helix Spirals Workflow Automation: 16+ node types, Google Workspace integration
•Health Dashboard with real-time system metrics
•12 MCP Servers: memory, coordination, infrastructure, code, filesystem, and more

Improved

•Frontend accessibility with ARIA attributes
•Error handling with exponential backoff
•TypeScript type safety across components

Fixed

•Python syntax error in system_coordination_core.py
•WebSocket connection stability

v1.9.0Feature Update2026-01-14

Code Quality & Modularization

Highlights

Main.py reduced by 51% (5,582 → 2,749 lines)
33 endpoints extracted into 5 modular route files
Zero breaking changes with all 524+ routes maintained

Added

•stripe_routes.py - 7 Stripe billing endpoints (320 lines)
•arjuna_routes.py - 7 Arjuna Portal portal endpoints (500 lines)
•coordination_api_routes.py - 10 coordination streaming endpoints (685 lines)
•mandelbrot_routes.py - 5 Mandelbrot UCF generation endpoints (180 lines)
•legacy_standalone_routes.py - 6 legacy endpoints (507 lines)

Improved

•Code organization and maintainability
•Testability (can test modules independently)
•Graceful import fallbacks for optional dependencies

v1.8.0Major Release2024-12-22

Multi-Agent Chat & Analytics Dashboard

Highlights

Real-time WebSocket-based chat interface
8-qubit system state visualization
Discord multi-agent integration with 10+ commands

Added

•Frontend Chat System (1,200+ lines): Agent selector, performance indicators, emoji picker
•System Dashboard (1,500+ lines): Bloch spheres, entanglement matrix, routine monitor
•Discord Integration: Interactive agent selection, conversation threading, 16 languages

Improved

•UsageDashboard with authentication headers and retry logic
•Component reusability and accessibility
•Rendering performance optimization

Fixed

•Fixed Python syntax error in system_coordination_core.py
•Enhanced WebSocket connection stability

v1.7.0Feature Update2024-12-15

Voice Patrol & LLM-Powered Agents

Highlights

Discord voice channel patrol with themed agent personalities
All 24 agents now have intelligent, contextual LLM responses
Supports 4 LLM providers: Ollama, Anthropic, OpenAI, Custom

Added

•Voice Patrol System: Auto-join, patrol loop, 5 agent voice profiles
•LLM Agent Engine (700+ lines): Unique system prompts per agent
•Commands: !voice-join, !voice-leave, !voice-announce, !voice-status

Improved

•Agent personality consistency across platforms
•Conversation history per session

Stay Updated

Get notified about new features and improvements.

Create Free Account ⭐ Star on GitHub