Upgrade LocalAI from v2.25.0 to v3.9.0 with new features:
- **Agent Jobs Panel**: Schedule and manage background agentic tasks
- **Memory Reclaimer**: LRU eviction for loaded models, automatic VRAM cleanup
- **VibeVoice backend**: New voice synthesis support
Update README with:
- v3.9 feature highlights
- Complete CLI command reference
- Model presets table (tinyllama, phi2, mistral, gte-small)
- API endpoints documentation
- SecuBox Couche 2 integration notes
This is part of the v0.18 AI Gateway roadmap.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Replace Docker/LXC-based approach with direct binary download
- Download LocalAI v2.25.0 binary from GitHub releases
- Add localaictl CLI for install, model management, and service control
- Change default port to 8081 (avoid CrowdSec conflict on 8080)
- Remove secubox-app-localai-wb (merged into secubox-app-localai)
- Add model presets: tinyllama, phi2, mistral
Usage:
localaictl install
localaictl model-install tinyllama
/etc/init.d/localai enable && /etc/init.d/localai start
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Download Docker image layers directly via Registry API
- No dockerd or podman daemon required anymore
- Same approach as mitmproxy and magicmirror2 packages
- All backends included (llama-cpp, whisper, etc.)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Updated default version from v2.25.0 to v3.10.0
- Fixed binary URL format: local-ai-v3.10.0-linux-arm64
- Updated Docker image tag to v3.10.0-ffmpeg
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Add runtime_is_working() to verify daemon connectivity
- Falls back to standalone binary with helpful message if daemon not running
- Provides hint: /etc/init.d/dockerd start
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
When using `localaictl install --lxc`:
1. If podman/docker available: extracts rootfs from Docker image
- Includes ALL backends (llama-cpp, whisper, etc.)
- Creates LXC container with full LocalAI capabilities
2. If no docker/podman: falls back to standalone binary
- Limited backend support
This gives the best of both worlds:
- LXC lightweight container management
- Full Docker image backends
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
LocalAI changes:
- Rewrite localaictl to use Docker/Podman instead of standalone binary
- Use localai/localai:v2.25.0-ffmpeg image with all backends included
- Fix llama-cpp backend not found issue
- Auto-detect podman or docker runtime
- Update UCI config with Docker settings
New Ollama package:
- Add secubox-app-ollama as lighter alternative to LocalAI
- Native ARM64 support with backends included
- Simple CLI: ollamactl pull/run/list
- Docker image ~1GB vs 2-4GB for LocalAI
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Add secubox-app-localai package with LXC container support for LocalAI service
- Add luci-app-localai with dashboard, chat, models and settings views
- Implement RPCD backend for LocalAI API integration via /v1/models and /v1/chat/completions
- Use direct RPC declarations in LuCI views for reliable frontend communication
- Add LocalAI and Glances to secubox-portal services page
- Move Glances from services to monitoring section
Packages:
- secubox-app-localai: 0.1.0-r1
- luci-app-localai: 0.1.0-r8
- luci-app-secubox-portal: 0.6.0-r5
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
