Community Support
Meechi is an Open Core project maintained by a Studio of Two.
Private Support
Have a question or need help with your setup? Email us directly. We read every message.
No marketing lists. No selling data.
Feature Requests
Found a bug? Want to suggest a feature? Help us shape the future of Meechi by tracking it publicly.
Requires a GitHub account.
Have a quick question?
Check the FAQ on the Home PageLatest Updates
A critical structural patch to the open source dependencies and publication lifecycle.
- •Standardized all @tiptap/* dependencies to ^3.14.0, eliminating the fatal "Twin Editor" TypeScript collisions.
- •Rewired the NPM Prepbulish hook to directly leverage local Yarn binaries, eliminating the need for global Corepack.
A specialized infrastructure update hardening the monorepo for Yarn 4 and completing the extraction of the Semantic Links ecosystem.
- •Migrated to Yarn 4.6.0 (Berry) for improved monorepo performance and workspace protocol safety.
- •Standardized all release and publication scripts to use local binaries, ensuring environment-independent orchestration.
- •Resolved a precision bug in the AI chunking engine to eliminate trailing phantom segments.
- •Fully synchronized the extracted @meechi-ai/core logic with the open-source repository.
A major philosophical expansion and intelligence upgrade, shifting to a unified LlamaIndex hub and compressed contexts.
- •Formally implemented the duality of Lore (history) and Echoes (present).
- •Implemented LLMLingua-2 prompt compression and Forensic DNA Mirror for vastly expanded effective context.
- •Pivoted to a unified LlamaIndex Hub for professional RAG and parsing, with Karpathy-inspired Lean RAG for offline modes.
- •Hardened settings, MCP tools, and system prompts for robust local search UI and configuration persistence.
- •Ensured core build script ESM compatibility and hardened deployment path-length sanitization.
Introducing granular icon styling and resolving chat UI regressions.
- •Added a new "Thin" (0.75px) icon style option for a lighter aesthetic.
- •Fixed a regression where agent icons were missing in chat history logs.
- •Resolved TypeScript build blockers to unblock Next.js Fast Refresh.
Major architectural refactoring to eliminate "God Files" and harden system boundaries.
- •Decomposed JiyuInput, FileExplorer, and useMeechi into isolated, maintainable components and hooks.
- •Refactored native-sync and engine logic into a clean adapter pattern.
- •Enforced strict architecture guards and licensing checks across packages.
- •Expanded test coverage and implemented comprehensive CI smoke gates.
- •Restored backdrop-filter glass blur and hardened fallback UI tokens.
Major infrastructure unblocking offline inference and robust PWA support.
- •Full implementation of bundled Ollama sidecar for offline local LLM inference.
- •New relay_client.rs with proxy auth middleware for authenticated sidecar routing.
- •Added next-pwa, web app manifest, and WebLLM caching for offline support.
- •Hardened release script with semver validation and change detection.
Focusing on reliable data persistence and UI refinement.
- •Implemented WebRTC LAN sync, QR pairing, and conflict copy.
- •Added 4-tier display (Seed/Lagom/Ginga/Jiyu) to the app header.
- •Fixed semantic link mark fragmentation and color persistence bugs.
- •Resolved comment counter flicker and HTML leakage.
Desktop deployment update.
- •Included v1.0.8 MSI and EXE installers in public assets for deployment.
Major milestone for local-first architectural stability.
- •Implemented Tier 1 Local Hub for true data ownership.
- •Resolved critical Vercel build and ESM module resolution errors.
- •Fixed AI identity persistence bugs.
Transforming the landing experience into a philosophical manifesto.
- •Launched the "Sanctuary" landing page design.
- •Added "Studio of Two" origin story on the About page.
- •Introduced the community support channel on GitHub.
Meechi is evolving every day in the Studio of Two.