Changelog

All notable changes to Launchverse are documented here.

2026 — Q3

July 2026

Helix — Your Autonomous AI Engineer

Helix is Launchverse's built-in AI cloud engineer that deploys, diagnoses, and fixes infrastructure autonomously. Just tell it what you need in plain English.

• Autonomy levels: L0–L4 control how much Helix can execute without approval, from read-only advisor to fully autonomous operator
• Autonomous deployments: Push a repo and Helix detects the stack, builds containers, and deploys — no YAML required
• Self-healing infrastructure: When builds fail, Helix reads logs, diagnoses root cause, and applies fixes automatically
• Build-fix bank: Helix recalls canonical fixes and stores successful repairs for similar future errors
• Live web search: Helix can look up current package versions, docs, and error solutions
• Live reasoning chain: Watch Helix think in real-time — see the analysis, the plan, tool calls, and results streamed to your dashboard
• Direct SSH execution: Helix connects to your infrastructure via secure SSH to run diagnostics, restart services, and fix config issues directly in containers
• GitHub PR automation: Helix analyzes failures, writes code fixes, opens pull requests, and merges when CI passes
• Infrastructure management: Scale instances, manage databases (PostgreSQL, MySQL, Redis, MongoDB), configure domains with automatic SSL, and set up cron jobs — all through natural language
• Environment secrets: Add, update, and rotate environment variables with automatic redeployment
• Watchdog mode: Helix runs continuously in the background — when issues arise, it springs into action autonomously based on your configured trust level

Helix Benchmark Results

• Unit benchmark: 18/18 passed
• Real-world benchmark: 29/29 passed
• Heavy autonomous engineering benchmark: 386/405 passed (95.3%)
• Flagship greenfield run: passed — Helix scaffolded, fixed, published, and deployed the helix-benchmark-flagship fixture, and /api/health returned {"status":"ok"}

Helix Reliability Fixes

• publish_sandbox_to_repo now publishes the sandbox working tree instead of the committed HEAD
• sandbox_read_file no longer redacts file contents, and secret redaction no longer corrupts source code like const secret = searchParams.get('secret')
• validate_workspace correctly handles preview URLs that already include https://
• Hallucinated tool names are reported back to the model as failures instead of crashing the turn

Helix OpenAI-Compatible Gateway

• Expose Helix through POST /api/v1/chat/completions and GET /api/v1/models
• Server-Sent Events streaming for real-time tool execution output
• Per-project API key scoping, rate limits, and usage metering
• Official TypeScript SDK: @launchverse/helix-gateway
• OpenAPI 3.1 specification and integration guides for Vercel AI Gateway, Cencori, OpenAI SDK, and IDE plugins

Smart Project Groups

• Transform isolated apps into cohesive microservice ecosystems
• Zero-config internal networking with auto-generated DNS aliases
• Automated group-level secret injection
• Dependency-aware deployments (databases boot before apps)
• Visual project canvas support for group architectures

2026 — Q2

May 2026

PR Preview Enhancements

• Preview environments now support per-preview environment variable overrides
• Teardown grace period extended to 24 hours
• Auto-deploy on PR sync is now configurable per project

Database Provisioning

• Added support for DragonFly and KeyDB
• Database credential rotation via API
• Point-in-time recovery for PostgreSQL instances

Platform Stability

• Improved build queue fairness — long-waiting builds now get priority
• Reduced cold-start times by 40% for Node.js applications
• Enhanced auto-rollback health probe with configurable thresholds

April 2026

Initial API v1 Release

• Stable REST API at /api/v1 with authentication, projects, deployments, and domains endpoints
• OpenAPI 3.1 specification available at /api/v1/openapi.json
• Cursor-based pagination for all list endpoints
• Rate limiting with per-token tracking

Managed Databases

• PostgreSQL, MySQL, MariaDB, MongoDB, and Redis support
• Automated daily backups with 7-day retention
• Public access toggle for external tooling

March 2026

• GitHub App integration for auto-deployments
• Custom domain support with automatic Let's Encrypt TLS
• Team management with role-based access control
• Two-factor authentication (TOTP)