Skip to content

Changelog

The public changelog now tracks the cleaned production release line only. Older v4 and v3 history is intentionally excluded from the published docs so the site matches the current GitHub release train.

Version v2.5.1

Release Date: 2026-05-15

Platform Release: Deterministic AI Runtime Routing, Clearer Failures, and Runtime Health Accuracy

This release tightens AI runtime selection, removes silent same-request failover, and makes Runtime Health match the actual credential source used by super-admin contexts.

Features

  • Second-attempt provider rotation — repeated identical asks can now move Muse and shared LLM agents to the next healthy provider through a retry-hint flow instead of retrying invisibly inside the same request.
  • Clearer first-failure reporting — Muse now returns the real provider or model failure reason first, so operators can see what broke before deciding to retry.

Improvements

  • Deterministic runtime precedence — AI Settings defaults are always applied first, agent-card settings second, and reachable local Ollama third when no saved provider can run.
  • Runtime Health accuracy — provider cards now show when availability comes from a shared or environment-backed key rather than a tenant-saved credential.
  • LinkedIn connection UX — the primary connection action now uses cleaner copy and opens the inline setup flow directly.

Bug Fixes

  • Removed silent same-request provider fallback paths that could override the configured default runtime.
  • Fixed misleading ShareAI availability states in super-admin Runtime Health views.
  • Updated release metadata surfaces to report version 2.5.1 consistently.

Version v2.4.9

Release Date: 2026-05-10

Platform Release: Multi-Agent Orchestration, Isolated Demo Environments, and AI Settings UX

This release expands the platform's agent orchestration layer, hardens tenant-aware demo flows, and smooths the AI Settings experience for local and self-hosted model providers.

Features

  • Multi-agent execution foundation — added delegated agent execution building blocks, expanded marketing-agent orchestration, and supporting AI tool/runtime updates for more structured specialist-agent handoffs.
  • Isolated demo lifecycle support — improved tenant-aware onboarding and demo-environment handling so user trials and demo-style environments can be provisioned more safely.
  • Broader connection-management coverage — refined admin flows for Facebook, Google Drive, Google Analytics, Telegram, LinkedIn MCP, and Postiz integrations.

Improvements

  • Agent Studio and AI admin refresh — updated Agent Studio, AI Settings, Workflow Builder, and Assistant Dashboard flows around cleaner registry-driven routing, stronger MCP coordination, and clearer customization copy.
  • Tenancy bootstrapping refinement — improved subdomain initialization and brand/tenant wiring for safer isolated environments and smoother tenant-aware execution.
  • Release pipeline updates — improved GitHub snapshot publishing and production synchronization scripts for faster release visibility and safer rollout coordination.
  • Agent naming and UI polish — refreshed tenant-facing copy and renamed the legacy mirror-mode label to Ambasador Agent in agent-management surfaces.

Bug Fixes

  • Ollama local-model UX — fixed auto-detect/test behavior so detected endpoints normalize correctly, immediate test actions work before saving, and stale “API Key Required” warnings clear once Ollama is detected.
  • Agent Studio customization alignment — corrected the two-column footer layout so helper text and input areas align more consistently.
  • Provider/runtime fallback handling — reduced false configuration failures tied to local endpoint normalization and runtime key selection.
  • Production deployment verification — revalidated production deploy flows and post-sync checks on both aiohm.org and aiohm.app.

Version v2.4.8

Release Date: 2026-05-08

Platform Release: Agent Studio UX, Custom Agent Operations, and Production Rollout

This release improves the Agent Studio experience, stabilizes custom-agent persistence, and ships the changes to both production environments.

Features

  • Agent Studio create panel is now action-driven and only appears after clicking Create Agent.
  • Added custom-agent delete action (custom agents only) with confirmation.
  • Added Agent Studio analytics info box with total agents and custom-agent counts.

Improvements

  • Simplified Agent Studio header actions by removing the extra Open Muse shortcut and tightening navigation labels.
  • Improved custom-agent creation path behavior in tenant context.

Bug Fixes

  • Fixed Agent Studio runtime error by correcting AIAgents class namespace usage.
  • Confirmed central custom_agents migration rollout and successful deploy verification on aiohm.org and aiohm.app.

Version v2.4.3

Release Date: 2026-04-27

Platform Release: Tenancy Hardening, Auditing & Deploy Safety

This release focuses on production safety and multi-tenant correctness. It strengthens tenant resolution across auth/session pathways, introduces first-party auditing foundations, and closes a deployment gap that could remove the docs bundle during release sync.

Features

  • Auditing foundation — integrated Laravel Auditing (owen-it/laravel-auditing) with baseline audit configuration and initial audits migration.
  • Tenant-aware model and bootstrap improvements — tenancy boot logic and model concerns were expanded to improve consistency across shared admin and tenant-bound execution paths.
  • Expanded Social Visual Text tests — additional unit coverage for text generation and formatting flows.

Improvements

  • Tenancy middleware hardening — updated tenant initialization and isolation middleware to reduce cross-tenant leakage risk in authenticated/session-driven flows.
  • Provisioning flow updates — setup and tenancy DB manager improvements for safer brand bootstrap and tenant database lifecycle handling.
  • Homepage refresh — updated public home page messaging and structure.

Bug Fixes

  • Deploy docs safety guard — release deploy validation now requires public/docs/index.html before sync in scripts/github-aiohm-last-version.sh, preventing accidental docs removal from production.

Version v2.4.2

Release Date: 2026-04-24

Platform Release: Agent Customization, Quality Rubrics & Reliability

This release introduces per-tenant agent behaviour overrides, a full Quality Rubrics UI, multi-tenant shared admin portal, and a round of targeted reliability fixes.

Features

  • Agent Customization UI — New settings page lets each tenant override hardcoded agent behaviour without code changes: add extra copywriting instructions, extend forbidden-word lists, add mode trigger keywords, and replace mode instruction text per marketing mode (strategy, copy, social, email, SEO, PPC, growth, creative).
  • Quality Rubrics CRUD — Super admins can now create, edit, and clone quality rubric definitions directly from the admin panel. Includes repeater for criterion name/description/weight, auto-synced weights map, Clone action for tenant-specific overrides, and active/version controls.
  • Shared Admin Portal (admin.aiohm.app) — Multi-tenant shared admin portal now works. InitializeTenancyForSubdomains middleware detects authenticated users on shared portals and initialises tenancy from their stored tenant_id rather than domain mapping.
  • Check All API Keys — New header button on the AI Control Center page bulk-tests all configured provider API keys and reports status in one click.
  • Agent Customization DB Layer — New agent_customizations table (central DB) and model with forTenant() resolver, mergeKeywords(), and resolveInstruction() helpers. Marketing agent and mode detector load per-tenant overrides at runtime.

Improvements

  • Muse Rate-Limit Failover — MuseModeAgent now falls back to a lighter model and then an alternate provider via ApiProviderFailsafe when a rate-limit or quota error is detected. Eliminates "couldn't reach model" errors from exhausted Gemini keys.
  • .env Key Priority — AppServiceProvider now ensures .env API keys take priority over stale values in the Settings table. Super-admin dev tenant skips Brand override entirely and uses .env keys directly.
  • Deploy Script Reliability — Replaced fragile inline PHP heredoc in the deploy script with a standalone scripts/patch-vendor-mcp.php file. Groq MCP schema patch and tsconfig moduleResolution fix applied on clean deploy.
  • Vector / Knowledge Graph UX — Aligned vector entry and knowledge graph UI interactions; fixed scan ownership attribution.

Bug Fixes

  • Image Studio Title Truncation — ExecuteWorkflowJob was passing titles up to 120 chars to the poster generator whose validator rejects anything over 45. Fixed with mb_substr($title, 0, 45) and mb_substr($subtitle, 0, 90).
  • Memory Write Queue Slack Settings 500 — PHP 8.4 Redirector crash on /memory-write-queues/slack-settings caused by the wildcard /{record} route swallowing the literal path. Fixed by declaring the slack-settings page route before view in getPages().
  • RSS Job Type Error — Corrected job type resolution in RSS processing pipeline.
  • ListDocumentationFiles MCP Tool — Fixed parameter handling to prevent schema validation errors with Groq provider.

Version v2.4.1

Release Date: 2026-04-19

Hotfix: Production Snapshot

  • release: Prepare v2.4.1 production snapshot — stable baseline cut from local-dev-pro branch after workflow builder and seeder stabilisation.
  • feat: Added Tantramovement WorkflowTemplate seeder and verification scripts.
  • fix: Restored workflow builder node card styling from DEV backup.

Version v2.4.0

Release Date: 2026-04-15

Platform Release: Muse Memory & Performance

This release brings a major memory management update for Muse, along with significant performance improvements, new memory systems, and connectivity features.

Features

  • Telegram Multi-Instance — Connect multiple Telegram bots per brand with the new UserConnection table. Each bot now has its own webhook and settings.
  • AI Key Diagnostics — Image generation now shows clear tenant key resolution status (tenant/global/env/missing) so you can debug failures faster.
  • Muse Learnings System — Automated extraction of strategic insights from conversations with quality-controlled storage in learnings.md and approval rate tracking.
  • Muse Belief Coaching — Silent belief signal extraction that informs Muse's tone and framing without exposing profile data to users.
  • Skill Compounding — Automatic synthesis of recurring approved insights into reusable agent skills (via MuseSkillCandidateService).
  • Mode-Aware Prompts — Separate prompt injection for Muse (creative partner) vs Tech/Tech Partner (mirror) modes with distinct context bundles.

Improvements

  • Adaptive Prompt Fidelity — Jaccard-scored scratch note selection replaces simple "last 5 notes" with importance × recency × relevance scoring.
  • Importance Scoring — Scratch notes and vector entries now have importance metadata (1-10 scale) with high-importance (8+) archiving protection.
  • Belief Map Persistence — beliefs.json now syncs to Google Drive alongside brand_bible.md and learnings.md for cloud backup.
  • Brand Bible Injection — MirrorModeAgent now receives brand voice context directly from brand_bible.md for public-facing responses.
  • Dynamic Site Context — MirrorModeAgent includes real-time site name, URL, and date in responses.
  • Memory Pruning — PruneMemoryJob runs 4 AM sync with deduplication, summarization, and graceful archiving.

Bug Fixes

  • Brand Book Export — PDF export now includes quick-start discovery answers and expanded wizard data.
  • Scheduled Workflows — Cron expressions now work correctly for scheduled workflow triggers.
  • OAuth Redirect — Google OAuth redirect now uses correct tenant-specific domains.
  • Mode Migration — Internal mode key "mirror" gracefully renamed to "tech" with alias support for /mirror and /tech commands.

Version v2.2.5

Release Date: 2026-04-14

Hotfix: Production Snapshot

  • release(version): Production snapshot — project version baseline reset to v2.2.5.

Version v2.1.0

Release Date: 2026-04-01

Platform Release: Version Reset and Deploy Workflow

  • release(version): Reset to v2.1.0 — platform version and release counter explicitly reset for new stable release line.
  • chore(release): Deployment flow — added scripts/github-aiohm-last-version.sh run for aiohm.app production update.
  • fix(ci): GitHub release linkage — ensure the latest tag is created/pushed and referenced in deployment artifact fetch.

Version v2.0.1

Release Date: 2026-03-26

Hotfix: Founders Edition Card Reveal

  • fix(home): Founders Edition hover reveal — the overlay now fully opens on hover using a 3D door-swing animation (rotateY) matching the sales page behaviour. Previously the clip-path animation stopped at 50%, leaving half the card obscured.

Version v2.0.0

Release Date: 2026-03-26

AIOHM Becomes Your AI Marketing Agency

Version 2.0.0 is a major milestone that repositions AIOHM as a full-service AI marketing agency platform. The entire user journey, documentation, and agent layer have been redesigned around the concept of an agency team — each AI agent now fills a specialist role in your marketing department.

Agency Team Roles

  • feat(agents): Creative Director (Muse) — the master strategist orchestrating your entire agency.
  • feat(agents): Campaign Strategist — dedicated marketing campaign planning and execution.
  • feat(agents): SEO Specialist — search optimization, keyword research, and content strategy.
  • feat(agents): Content Publisher — WordPress content management, publishing, and site operations.
  • feat(agents): Commerce Manager — WooCommerce product operations, inventory, and sales.
  • feat(agents): Outreach Manager — email marketing, lead nurturing, and Mautic automation.
  • feat(agents): Data Analyst — Google Analytics insights, reporting, and performance tracking.
  • feat(agents): Web Designer — landing pages, microsites, and visual web experiences.
  • feat(agents): Content Curator — RSS feed monitoring, content discovery, and autoblogging.
  • feat(agents): Brand Ambassador — public-facing AI for visitor engagement and support.
  • feat(agents): Research Analyst — entity extraction, market research, and knowledge mapping.

Platform Evolution

  • feat(docs): Agency-First Documentation — complete VitePress rewrite with agency onboarding flow, team-based navigation, and client-focused feature descriptions.
  • feat(ux): Client Journey Redesign — new getting-started flow designed for marketing teams and agency clients, not developers.
  • feat(brand): Marketing Agency Positioning — updated branding, hero messaging, and feature descriptions across all public-facing pages.
  • refactor(version): Version Reset to 2.0.0 — clean break from the previous release train to mark the agency pivot.

Version v1.0.10

Release Date: 2026-03-22

Production Flow Hardening

This release tightened the production pipeline after the new clean-main strategy went live. The goal was to make filtered releases, built docs, and VPS deploys more reliable.

  • fix(release): Filtered Snapshot Reliability - preserve Git worktree metadata correctly while GitHub assembles the production-only branch.
  • fix(validation): Required Checks Alignment - keep production validation and secret scanning consistent with the protected main branch.
  • docs(site): Public Docs Build Stability - rebuild and publish the production docs bundle from the updated source tree.

Version v1.0.9

Release Date: 2026-03-22

Clean Install And Release Pipeline

This release established the new packaging model: dev branches keep the full workspace, while main ships a production-only application with installer tooling and built docs.

  • feat(install): First-Time Installer - add the install/ package and php artisan aiohm:install --production flow for fresh deployments.
  • feat(release): Production Mirror Branch - make main the filtered production branch instead of a full dev checkout.
  • feat(docs): Built Docs Contract - serve production documentation from public/docs/ while keeping raw VitePress source on dev branches only.
  • fix(security): Secret-Safe Production Snapshot - exclude dev-only folders, runtime storage noise, and local-only files from release packages.

Version v1.0.5

Release Date: 2026-03-20

Demo Asset Curation

This release cleaned up which seeded files belong in source control and which should remain runtime-only.

  • fix(storage): Runtime Noise Filtering - refine ignore rules so generated storage artifacts stop leaking into the repository.
  • feat(install): Demo Asset Preservation - keep intentional demo templates, gallery assets, and starter content available for installs.

Version v1.0.4

Release Date: 2026-03-20

Production Validation Tightening

  • fix(ci): Production PR Validation - strengthen GitHub checks that guard the filtered production branch.
  • fix(ci): Release Workflow Safety - reduce the risk of dev-only files reaching main.

Version v1.0.3

Release Date: 2026-03-20

Admin UI Polish

  • feat(ui): Filament View Overrides - publish branded component overrides for the admin interface so the production panel keeps the intended presentation.

Version v1.0.2

Release Date: 2026-03-20

Changelog Pipeline Follow-Up

  • fix(docs): Changelog Automation Workflow - refine the docs automation setup after the first production-only release pass.

Version v1.0.1

Release Date: 2026-03-20

Changelog Pipeline Setup

  • feat(docs): Public Changelog Build - introduce the first GitHub automation pass for publishing the documentation output.

Version v1.0.0

Release Date: 2026-03-20

Production Baseline Reset

This is the reset point for the new public release line. The repository moved to a production-only main, a curated installer package, and a cleaner deploy story.

  • release(platform): Version Reset - restart the public release history at v1.0.0.
  • feat(release): Production-Only Main Baseline - establish the clean branch and release package strategy as the new default.