Architecture
One coding interface. An explicit compute destination.
The model host, developer workspace and network transport have separate responsibilities.
Your GPU. Your AI. Your Control. Anywhere. · Created by Paladugu Ganesh Naidu
Architectural flow
One developer interface. One chosen inference destination.
This is a conceptual diagram of a supported SSH-model configuration, not a claim of automatic remote-host provisioning. Inference occurs on the selected model host; ordinary workspace actions remain subject to the local extension’s permissions.
VS Code interface and tools
The extension is embedded in VS Code. It presents Ask, Plan and Agent modes; selected files and diagnostics can provide context. File proposals and ordinary command execution remain under the development host and its permissions.
Inference routing
Choose local Ollama, an Ollama server on a user-managed SSH host or an explicitly configured compatible endpoint. The choice determines where prompts and included context are processed.
Reviewable execution
Agent workflows involve requests, proposed edits, approvals, permitted actions and verification. A generated answer is not evidence that a file has been saved or that a command passed. Users should review outputs and failure states.
SSH and trust boundaries
SSH transports model requests to a reachable authenticated machine; it does not provide automatic GPU provisioning, air-gapped operation, model downloads or unrestricted remote terminal access.
Continue exploring TuxNest
Our tagline is “Tuxnest is India's first plug-and-play hybrid AI agent ecosystem”. Self-hosted inference requires appropriate host security, model capacity, power and networking; it does not establish automatic legal compliance.