Interconnector

Multi-Machine Coordination

While Guaardvark is powerful as a standalone application, it truly excels when deployed as a distributed network. The Interconnector is the foundational protocol that links these nodes together. It enables a family of local Guaardvark instances to work as a single, coordinated system, sharing resources and knowledge across your local network without ever leaking data to the public internet.

Synchronization is handled through a secure Master-Client architecture. You can designate one machine (perhaps a powerful workstation with a high-end GPU) as the master node, and connect other devices (like laptops, Raspberry Pis, or edge servers) as clients. The Interconnector then manages the secure flow of state, ensuring that your project files, custom rules, and model configurations are consistent across your entire fleet.

Seamless File and Document Sync

The Interconnector provides a robust Local Sync Engine that monitors your project directories for changes. When you upload a document on your primary workstation, it can be automatically pushed to your other connected devices. This ensures that your RAG Search indexes are always up to date, regardless of which machine you are currently using to chat with your AI. The sync process is peer-to-peer over your local network, maximizing speed and maintaining complete privacy.

Broadcast of Autonomous Learnings

The most advanced feature of the Interconnector is the Learning Broadcast. When a Guaardvark instance running the Self-Improvement Engine identifies and fixes a bug in its codebase, it validates the patch and then broadcasts that "learning" to every other node in the network. This allows your entire fleet of devices to benefit from the autonomous improvements discovered by any single machine, creating a self-healing and self-optimizing infrastructure.

Key Networking Capabilities

• Model Sharing: Download a model once on your master node and share it across the network to save bandwidth and storage space.
• Node Management: A dedicated UI for managing connected devices, approving new nodes, and monitoring the health of your local AI fleet.
• Conflict Resolution: Intelligent versioning ensures that file changes are merged correctly between machines, with manual override options if needed.
• Air-Gapped Sync: Designed to work in environments with no external internet access; the Interconnector only requires local network connectivity to function.

Configure It in Settings

The Interconnector is configured entirely from the running app — there is no YAML file to hand-edit. Open Settings → Interconnector on each machine you want to link. The modal handles enable/disable, node mode, naming, API keys, sync entities, and auto-sync interval.

On the master node (the machine other devices will sync to)

1. Open Settings from the left rail, then click the Interconnector card to open the configuration modal.
2. Toggle Enable Network Interconnector.
3. Set Node Mode to Master (Server). The Node Name defaults to your System Name from Branding — override it if you want a more descriptive name (e.g. workstation-rtx4090).
4. Under Master Server Settings, leave Require API Key Authentication on (recommended unless every machine is on a fully trusted LAN).
5. Click Generate API Key. The key is shown once in a yellow alert — copy it immediately. After you navigate away only the truncated hash remains; if you lose the key, regenerate a new one.
6. Optional: click Test File Scanning to confirm the master can index its project tree before clients try to pull from it.
7. Save. The master's address is its LAN IP plus port 5002 (e.g. http://192.168.1.50:5002).

On each client node

1. Open Settings → Interconnector on the client.
2. Toggle Enable Network Interconnector.
3. Set Node Mode to Client.
4. Give it a Node Name (e.g. laptop-dean or pi-cluster-3).
5. Under Master Server Connection, paste the Master Server Address and the Master API Key you copied earlier.
6. Click Test Connection to confirm the client can reach the master.
7. Under Sync Settings, optionally turn on Enable Auto-Sync and set a Sync Interval (seconds) — the default is 300 (5 minutes). Pick which entity types to sync: clients, projects, and websites are enabled by default.
8. Save. The client registers with the master automatically and starts sending heartbeats every 30 seconds.

Once both sides are configured the master's Connected Nodes table populates within a minute. From that table the master can approve pending registrations, broadcast code or data updates to selected clients, and watch sync status in real time. Manual one-shot syncs are available alongside the auto-sync schedule via the Sync Now button.