Skip to content

Releases: jherforth/Monastery

Monastery v0.1.2-alpha

Pre-release

Choose a tag to compare

@jherforth jherforth released this 26 Jun 22:30

Release Date: 2026-06-26

The second alpha. If v0.1.0 gave Monastery a brain — LLMs, chat, projects, Git — then v0.1.2 gives
it hands: it can now deploy what you build to your own infrastructure, dispatch agentic work to
Hermes
, and read its own deployment failures and try to fix them. Still very much alpha: the
deployment paths depend heavily on your own homelab setup and you'll occasionally coach the LLM
through a build — but the core loop of prompt → build → ship to your own boxes now runs end to end.

🚀 Self-Hosting & Deployment (new)

The headline of this release: a Self-Host Wizard that takes a Monastery project, commits it to
your Git forge, and builds + deploys it on your own platform.

Coolify

  • One-click deploy via the Coolify API (Dockerfile build pack)
  • Clone-at-build Dockerfiles — the image clones your repo at build time, so deploys work with
    IP / .local / self-signed hosts that Coolify's normal Git validation rejects
  • Smarter server selection — de-prioritizes the built-in localhost server so apps land on your real
    VPS, with a graceful fallback when a project has no deployment server
  • Deployment persistence (tracks the app UUID) for redeploy-in-place and 404 self-heal
  • Base64 Dockerfile payloads and SSH/Git URL parsing fixes

Dokploy

  • Full git-based deploy over Dokploy's tRPC API (create → save git provider → build type → deploy)
  • Container port auto-detected from the Dockerfile EXPOSE and applied before deploy
  • Fixes for auth, endpoint URLs, the dashboard link, the saveGitProvider schema, and the
    committed-Dockerfile flow (resolves "Github Provider not found" / "select-a-container")

Cloudflare Tunnel (both platforms)

  • Optional auto-wired cloudflared sidecar — paste a tunnel token and Monastery deploys the
    connector for you
  • Apps are published on a stable high host port for LAN + tunnel reachability; point your Cloudflare
    Public Hostname at the returned http://127.0.0.1:<port> service URL

Pocketbase backend (initial)

  • Configure one shared Pocketbase instance in Settings; the wizard injects POCKETBASE_URL
    (build-time and runtime) into deployed apps
  • A toggle-gated instruction set teaches the connected LLM how to build against Pocketbase
    (JS SDK usage, auth, collections) only when you're building a database-backed app

🔧 Deployment Monitoring & AI Auto-Fix (new)

  • Pull a failed deployment's build log straight into chat with one click
  • The connected LLM diagnoses the failure and applies fixes to your project files — then redeploy
  • Graceful fallback: when a platform can't serve the remote build log (a known upstream Dokploy
    limitation, issue #4687), the LLM proactively reviews your Dockerfile/config instead of dead-ending

🤖 AI Agents — Hermes Integration (new)

  • Connect an external Hermes agent in Settings → Hermes Agent (proxied via POST /api/hermes/run)
  • Agent mode toggle routes your chat through the agent instead of the raw LLM
  • Agent roles — Architect, Coder, Reviewer, Tester, Documenter, Deployer — as a selectable lens
    over your messages, multi-select capped at 2
  • Hardened tool-call handling that surfaces Hermes's tool steps inline
  • Shared-workspace docs so Hermes's own file writes show up in your Monastery project

⚡ Smarter Streaming

  • Auto-continue — responses cut off by the model's output-token cap resume automatically
    (capped, abort-aware, and toggleable), so long generations finish without manual clicking
  • Token usage shown per response when the endpoint reports it, accumulated across continuations
  • Provider-aware response handling and token caps (local Ollama vs cloud) to stop premature cut-offs
  • Fixes for runaway continuation loops and SSE stream parsing

🗂️ Workspace & Editor

  • Create and delete files and directories directly from the sidebar
  • File/folder uploads and drag-and-drop in the file tree
  • SVG rendering in the preview pane

🎨 UI & Polish

  • Consolidated chat toolbar — Auto-continue / Pocketbase / Agent-mode toggles collapsed onto a
    single compact row, with agent roles revealed inline when Agent mode is on
  • Refreshed branding and LLM "thinking" animation
  • Numerous compile and stability fixes across the Rust backend and React UI

⚠️ Alpha caveats: deployment behavior depends heavily on your Coolify/Dokploy setup (remote vs.
local build servers, tunnels, registries), and you should expect to occasionally guide the LLM
through a build fix. Back up anything you care about before pointing Monastery at it.


Full Changelog: v0.1.0-alpha...v0.1.2-alpha

v0.1.0-alpha

v0.1.0-alpha Pre-release
Pre-release

Choose a tag to compare

@jherforth jherforth released this 17 Jun 14:38

Monastery v0.1.0-alpha

Release Date: 2026-06-17

The inaugural alpha release of Monastery — a self-hosted sanctuary for AI-assisted coding. Connect your LLMs, manage projects with Git, deploy to your own infrastructure, and dispatch specialized AI agents - expect little from this initial offering.


Core Platform

Multi-LLM Support

  • Connect any OpenAI-compatible endpoint (Ollama, vLLM, DeepSeek, OpenAI, Groq, etc.)
  • Add, validate, and manage multiple endpoints in Settings
  • Auto-select active endpoint; mark favorites
  • Local endpoint detection with "local" badge

Chat Interface

  • Real-time SSE streaming with markdown rendering
  • Code block syntax highlighting with copy buttons
  • File attachment support (images, code files)
  • Auto-apply AI-generated code blocks to project files
  • Shell command detection and execution
  • Reasoning window — collapsible panel showing LLM thinking/reasoning tokens, capped at ~12 rows, scrollable
  • Stop/interrupt generation mid-stream
  • Welcome screen with suggested prompts

Project Management

  • Multi-project support with SQLite persistence
  • File tree explorer in sidebar
  • Multi-tab code editor (Monaco-based, custom Monastery dark theme)
  • File read/write via backend API
  • Preview pane for live output

Session Persistence

  • Chat sessions tied to projects
  • Create, switch, and delete sessions
  • Message history stored in SQLite via backend

Git Integration

Four Git forges supported with full CRUD + test workflow:

Forge Type Setup
GitHub Cloud API key (repo scope)
GitLab Cloud / Self-hosted API key (api, read_user)
Forgejo Self-hosted URL + API key
Gitea Self-hosted URL + API key

Features

  • Wizard-style connection setup with token guide links
  • Browse repositories with branch selection
  • Clone repos as new Monastery projects
  • Push projects to new repos on any forge
  • Git status display (branch, ahead/behind, changed files)
  • Auto-commit with snapshot creation before push

Agent System

Six built-in specialized AI agents, each with a tuned system prompt:

Agent Icon Role
Architect 🏗️ Plans project structure, chooses patterns
Coder 💻 Writes and edits code files
Reviewer 🔍 Code review — bugs, security, anti-patterns
Tester 🧪 Writes unit/integration tests
Documenter 📝 README, API docs, inline comments
Deployer 🚀 Dockerfile generation, deployment config

Invocation Methods

  • Chat quick-actions — collapsible button row above chat input (toggle with Agents header)
  • Editor toolbar — Explain / Refactor / Add Tests buttons dispatch Reviewer, Coder, and Tester with current file context
  • Natural language — ask the main chat to use an agent

How It Works

  1. Agent receives its system prompt + full project context (capped at 200KB)
  2. LLM streams response via SSE with reasoning support
  3. Response appears as live-streamed assistant message in chat

Self-Host Deployment

Hosting Services (Settings)

Connect to your self-hosted infrastructure:

Service Type Use
Dokploy PaaS Deploy apps, databases, Docker containers
Coolify PaaS Vercel/Netlify alternative for self-hosted
Pocketbase Backend Database, auth, file storage

Each service card shows connection status:

  • Not Verified — saved but not validated
  • Connected — validated successfully
  • Failed — error with message

Self-Host Wizard (Ctrl+Shift+D)

One-click deployment modal:

  • Auto-detects connected platforms and database backends
  • Framework detection from package.json (Next.js, Vite, React, Vue, Express, Fastify, Node)
  • Auto-generates optimized multi-stage Dockerfile if none exists
  • Creates application on Dokploy or Coolify
  • Triggers deployment and returns dashboard URL
  • Configurable app name and port

Snapshot Version Control

  • Create named snapshots of entire project state
  • Auto-snapshots before git commits
  • List, view, diff, and restore snapshots
  • Revert to any snapshot from chat markers

Sidebar

Four-tab collapsible left panel:

Tab Content
Files Project file tree with sync status indicators
Sessions Chat history with create/delete
Agents 6 built-in agents with descriptions, tool badges, and idle status
Integrations Live status of LLM endpoints, Git forges, and Hosting Services with count badges

Settings Modal

Three-tab configuration center:

Tab Content
LLM Endpoints Add/validate/delete OpenAI-compatible endpoints
Git Forges Connect GitHub, GitLab, Forgejo, Gitea
Hosting Services Connect Dokploy, Coolify, Pocketbase

Architecture

┌──────────────────────────────────────────────────┐
│                   Browser (React + Vite)          │
│  ┌─────────┐ ┌──────────┐ ┌───────────────────┐  │
│  │ Chat    │ │ Editor   │ │ Preview           │  │
│  │ +Agents │ │ +Toolbar │ │                   │  │
│  └─────────┘ └──────────┘ └───────────────────┘  │
│  ┌────────────────────────────────────────────┐  │
│  │ Sidebar (Files/Sessions/Agents/Integrations)│  │
│  └────────────────────────────────────────────┘  │
└────────────────────┬─────────────────────────────┘
                     │ REST + SSE
┌────────────────────▼─────────────────────────────┐
│              Rust Backend (Axum)                  │
│  ┌──────────┐ ┌────────┐ ┌───────────────────┐   │
│  │ Chat     │ │ Agents │ │ Deploy (Coolify/   │   │
│  │ Stream   │ │ Run    │ │ Dokploy)           │   │
│  └──────────┘ └────────┘ └───────────────────┘   │
│  ┌──────────┐ ┌────────┐ ┌───────────────────┐   │
│  │ Git      │ │Hosting │ │ Snapshots         │   │
│  │ Forges   │ │Services│ │                   │   │
│  └──────────┘ └────────┘ └───────────────────┘   │
│  ┌──────────────────────────────────────────┐    │
│  │ SQLite (endpoints, projects, sessions,    │    │
│  │  git_connections, hosting_connections)    │    │
│  └──────────────────────────────────────────┘    │
└──────────────────────────────────────────────────┘

Tech Stack

  • Frontend: React 18, TypeScript, Vite, Tailwind CSS, Zustand, SWR, Monaco Editor
  • Backend: Rust, Axum, SQLx, async-openai, reqwest
  • Database: SQLite
  • Streaming: Server-Sent Events (SSE) with reasoning support
  • Container: Docker Compose

What's Next (v0.2.0)

  • External agent support (Hermes, Open Claw)
  • Deployment progress streaming in wizard
  • Environment variable management for deployments
  • Pocketbase schema auto-generation
  • Agent pipeline orchestration (chain agents together)
  • Multi-model agent support (use different LLMs per agent)

Documentation

  • README.md — Project overview and quick start
  • ARCHITECTURE.md — High-level architecture
  • PROJECT_VISION.md — Design philosophy
  • docs/AGENTS.md — Agent system documentation
  • docs/SELF_HOST_WIZARD_PLAN.md — Deployment wizard plan
  • packages/web-ui/README.md — Frontend documentation