v12.0.0 Conductor - Multi-Agent Orchestration
Sentinel Desktop v12.0.0 — Conductor
Multi-agent orchestration, persistent memory, SSH control, and fleet management.
What's New
v8.0 Webhand — Embedded Browser Automation
- Playwright browser control (Chromium/Firefox/WebKit)
- 11 web actions: web_open, web_click, web_type, web_read, web_extract, web_wait_for, web_screenshot, web_eval_js, web_download, web_upload, web_tabs
- Dual-mode detection (auto web vs native routing)
- Self-signed cert whitelist for IT appliances
- IT appliance login detection (10 vendors: SonicWall, FortiGate, UniFi, Meraki, pfSense, OPNsense, MikroTik, NinjaOne, ConnectWise, IT Glue)
- Session vault — encrypted cookie persistence and restore
- Web recorder — capture browser actions as replayable JSON scripts
v9.0 Netops — SSH/Network Device Control
- SSH client via paramiko with connect/disconnect/run/context-manager
- Device-aware command runner (Cisco IOS/NX-OS, Juniper JunOS, FortiGate, MikroTik, pfSense, Linux)
- Output parser for interfaces, ARP, routing, ping, version
- 5 executor actions: ssh_connect, ssh_disconnect, ssh_run, ssh_show, ssh_ping
v10.0 Sentinel Server — Fleet/Daemon Mode
- Daemon service manager (start/stop/heartbeat/job tracking)
- Fleet manager (register/unregister/heartbeat multi-node tracking)
- Persistent job queue (submit/claim/complete/fail/cancel with priority)
- 14 new API endpoints (/daemon/, /fleet/, /jobs/*)
v11.0 Memory — Persistent Memory and RAG
- Episodic memory — timestamped JSONL with search, 30-day auto-compression
- Semantic memory — SQLite key-value facts with categories, tags, access tracking
- Working memory — in-memory session scratchpad with buckets
v12.0 Conductor — Multi-Agent Orchestration
- Task planner — rule-based goal decomposition with dependency detection
- Parallel executor — concurrent subtask execution respecting dependency graph
- Result synthesizer — merge multi-agent results with status aggregation
- Conductor coordinator — end-to-end plan, execute, synthesize pipeline
Stats
- 5,943 tests passing (606 new since v7.0)
- 50 files changed, 8,394 insertions
- 56 tool schemas registered for LLM tool-calling
- Zero lint errors
Upgrade
git pull
pip install -r requirements.txt
pip install playwright && playwright installFull Changelog: v3.1.0...v12.0.0
Full Changelog: v6.0.0...v12.0.0