release: merge develop into main for v0.30.2

Author: DavidsonGomes

.github/workflows/dashboard.yml

@@ -2,24 +2,29 @@ name: Build & Publish Docker Images (Swarm)
 
 # Builds evo-nexus-runtime and evo-nexus-dashboard images using the
 # Swarm-specific Dockerfiles (Dockerfile.swarm and Dockerfile.swarm.dashboard)
-# and pushes them to GitHub Container Registry (ghcr.io). The upstream
-# Dockerfile and Dockerfile.dashboard are intentionally NOT used here:
-# they are for local `docker compose up`, while these variants carry the
-# Swarm-specific bootstrap (entrypoint.sh, start-dashboard.sh, terminal-
+# and pushes them to Docker Hub under the `evoapicloud` namespace. The
+# upstream Dockerfile and Dockerfile.dashboard are intentionally NOT used
+# here: they are for local `docker compose up`, while these variants carry
+# the Swarm-specific bootstrap (entrypoint.sh, start-dashboard.sh, terminal-
 # server native build, both CLIs on PATH).
 #
 # Triggers:
 #   * push to main            -> builds + publishes :latest
 #   * push of a version tag   -> builds + publishes :vX.Y.Z and :latest
 #   * manual dispatch         -> builds without pushing (dry-run)
+#
+# Required repository secrets:
+#   * DOCKERHUB_USERNAME  — Docker Hub account with push rights to evoapicloud
+#   * DOCKERHUB_TOKEN     — access token (Settings → Security → New Access Token)
 on:
   push:
     branches: [main]
     tags: ['v*']
   workflow_dispatch:
 
 env:
-  REGISTRY: ghcr.io
+  REGISTRY: docker.io
+  NAMESPACE: evoapicloud
   FORCE_JAVASCRIPT_ACTIONS_TO_NODE24: 'true'
 
 jobs:
@@ -28,7 +33,6 @@ jobs:
     runs-on: ubuntu-latest
     permissions:
       contents: read
-      packages: write
     strategy:
       fail-fast: false
       matrix:
@@ -42,26 +46,22 @@ jobs:
       - name: Checkout
         uses: actions/checkout@v4
 
-      - name: Lowercase owner for ghcr.io
-        id: owner
-        run: echo "owner_lc=${GITHUB_REPOSITORY_OWNER,,}" >> "$GITHUB_OUTPUT"
-
       - name: Set up Docker Buildx
         uses: docker/setup-buildx-action@v3
 
-      - name: Login to GitHub Container Registry
+      - name: Login to Docker Hub
         if: github.event_name != 'workflow_dispatch'
         uses: docker/login-action@v3
         with:
           registry: ${{ env.REGISTRY }}
-          username: ${{ github.actor }}
-          password: ${{ secrets.GITHUB_TOKEN }}
+          username: ${{ secrets.DOCKERHUB_USERNAME }}
+          password: ${{ secrets.DOCKERHUB_TOKEN }}
 
       - name: Extract metadata (tags, labels)
         id: meta
         uses: docker/metadata-action@v5
         with:
-          images: ${{ env.REGISTRY }}/${{ steps.owner.outputs.owner_lc }}/${{ matrix.image }}
+          images: ${{ env.REGISTRY }}/${{ env.NAMESPACE }}/${{ matrix.image }}
           # Boa prática: publicar :latest junto com a tag versionada sempre
           # que um release (tag v*) for empurrado. Push para a branch main
           # sem tag também atualiza :latest.

.github/workflows/docker-publish.yml

@@ -82,6 +82,7 @@ config/workspace.yaml
 config/routines.yaml
 config/triggers.yaml
 config/heartbeats.yaml
+config/smart-router.json
 
 # ── Claude Code session data ─────────────
 .claude/projects/

.gitignore

@@ -5,6 +5,18 @@ All notable changes to this project will be documented in this file.
 The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.1.0/),
 and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
+## [0.30.2] - 2026-04-23
+
+Patch release focused on CI/distribution: Docker images now ship from the official `evoapicloud` namespace on Docker Hub (public, no auth required on Swarm managers), and the legacy dashboard-only workflow was removed to unblock the build pipeline.
+
+### Changed
+
+- **Docker images published to Docker Hub under `evoapicloud/`** — the Swarm workflow (`.github/workflows/docker-publish.yml`) now pushes `evo-nexus-runtime` and `evo-nexus-dashboard` to `docker.io/evoapicloud/*` instead of `ghcr.io`. Requires `DOCKERHUB_USERNAME` + `DOCKERHUB_TOKEN` secrets on the repo. The `evonexus.stack.yml` template and `README.swarm.md` were updated to reference the public images — no `OWNER` placeholder to fill in, no `docker login` needed on Swarm managers.
+
+### Removed
+
+- **Redundant `dashboard.yml` workflow** — deleted `.github/workflows/dashboard.yml`, which built a Python+React-only dashboard image (`Dockerfile.dashboard`) that nothing consumes. The Swarm workflow already produces a strict superset (with terminal-server and both CLIs). This also fixes the build failure on `main` caused by a TypeScript peer-dep conflict in the legacy Dockerfile.
+
 ## [0.30.1] - 2026-04-23
 
 Patch release focused on thread UX polish: session now swaps cleanly when switching threads via the sidebar, the agent is briefed explicitly about running inside a persistent thread (not a fresh one-shot session), the assignee dropdown stops hiding agents, and fresh installs no longer inherit Evolution-specific goal seed data.

CHANGELOG.md

@@ -75,10 +75,12 @@ long-running service. External REST APIs do not.
   resolver called `letsencryptresolver` (rename in the stack labels if
   your resolver has a different name).
 - A DNS A record pointing `evonexus.<yourdomain>` to the Swarm ingress.
-- Docker images published to a registry (the included GitHub Actions
-  workflow publishes to `ghcr.io/<owner>/evo-nexus-{dashboard,runtime}`).
-- Either make the ghcr.io packages public, or run `docker login ghcr.io`
-  on every Swarm manager before deploying.
+- Docker images are pulled from Docker Hub at
+  `evoapicloud/evo-nexus-{dashboard,runtime}` (published by the included
+  GitHub Actions workflow on every version tag and push to `main`).
+- The images are public, so no `docker login` is needed on the Swarm
+  managers. If you fork the project and publish to your own namespace,
+  update the `image:` lines in `evonexus.stack.yml`.
 
 ### Steps
 
@@ -90,9 +92,10 @@ long-running service. External REST APIs do not.
 2. **Open Portainer → Stacks → Add stack.**
    - Name: `evonexus`
    - Paste the contents of `evonexus.stack.yml`
-   - Replace the three placeholders: `OWNER` (your ghcr.io namespace),
-     `evonexus.example.com` (your hostname, two places), and
-     `traefik-public` (your Traefik overlay network name, several places)
+   - Replace the placeholders: `evonexus.example.com` (your hostname,
+     two places) and `traefik-public` (your Traefik overlay network
+     name, several places). The images already point at the official
+     `evoapicloud/evo-nexus-*` namespace on Docker Hub.
    - Click **Deploy**
 
 3. **Watch the containers come up.** `evonexus_dashboard` serves the SPA

README.swarm.md

@@ -1,6 +1,6 @@
 {
   "name": "@evoapi/evo-nexus",
-  "version": "0.30.1",
+  "version": "0.30.2",
   "description": "Unofficial open source toolkit for Claude Code — AI-powered business operating system",
   "keywords": [
     "claude-code",

cli/package.json

@@ -0,0 +1,25 @@
+{
+  "_comment": "OpenRouter Smart Router config. Copy to config/smart-router.json to customize. IDs validated against https://openrouter.ai/api/v1/models on 2026-04-23 — verify before editing. All listed models support tool/function calling.",
+  "enabled": true,
+  "port": 4891,
+  "host": "127.0.0.1",
+  "cooldown_ms": 60000,
+  "upstream_timeout_ms": 60000,
+  "breaker": {
+    "enabled": true,
+    "failure_threshold": 20,
+    "window_ms": 10000,
+    "reset_ms": 30000
+  },
+  "free_models": [
+    "openai/gpt-oss-120b:free",
+    "qwen/qwen3-coder:free",
+    "meta-llama/llama-3.3-70b-instruct:free",
+    "z-ai/glm-4.5-air:free"
+  ],
+  "paid_models": [
+    "google/gemini-2.5-flash",
+    "moonshotai/kimi-k2",
+    "anthropic/claude-sonnet-4.5"
+  ]
+}

config/smart-router.example.json

@@ -122,10 +122,20 @@ def _schedule_hc(get_app_fn) -> None:
 
 
 def start_health_check_thread(get_app_fn) -> None:
-    """Start the background health check scheduler (idempotent)."""
-    global _hc_started
+    """Start the background health check scheduler (idempotent).
+
+    Runs the first pass on a short-delay timer (not inline) so app startup
+    isn't blocked by slow/unreachable Postgres connections. Without this
+    initial pass, the classify_worker would spam the log for up to
+    _INTERVAL_S (5 min) on boot if a connection is offline, because stale
+    status='ready' from the last session won't be reconciled until the
+    first scheduled tick.
+    """
+    global _hc_started, _hc_timer
     with _hc_lock:
         if _hc_started:
             return
         _hc_started = True
-    _schedule_hc(get_app_fn)
+    _hc_timer = threading.Timer(5.0, _hc_loop, args=(get_app_fn,))
+    _hc_timer.daemon = True
+    _hc_timer.start()