A deep technical breakdown of Vibra-Ingenn's dual-mode receiver/caller pipeline, local REST daemon, and compiled Go graph execution flow.
Traditional AI agents are unidirectional: they either act strictly as an API client calling out to external cloud intelligence (like a web chat window), or they serve as a local model host listening for instructions (like Ollama or LM Studio).
Vibra-Ingenn is a bidirectional execution junction. It sits in the middle of your stack, acting simultaneously as a high-speed command receiver (for IDEs and local scripts) and an outbound compiler (driving local and cloud model inference to coordinate system workflows).
Vibra-Ingenn does not force you into a specific workflow tool. You can drive the local engine from three distinct entry interfaces, making it highly versatile:
Instantly integrates with Claude Desktop, Cursor, Cline, Windsurf, or VS Code. AI models call tools and run recipes on your machine in real-time.
The app runs a small service that listens on your own machine only. The built-in dashboard uses it to run and inspect recipes. Nothing outside your computer can reach it.
Run workflows directly inside local build scripts, cron jobs, task schedulers, or terminal consoles via vibra run <path_to_recipe>.
During a recipe execution, Vibra-Ingenn takes over. It manages variable state and credentials securely in-memory. For example, you can query a local model to parse codebase headers, format the output, and pipe it straight to a cloud model for a deeper code review.
Because Vibra-Ingenn handles all variable sanitization, HTTP connections (configured with a generous 120s timeout for large token generation tasks), and file operations natively in compiled Go code, the LLM is completely isolated from the execution bugs that plague traditional Python frameworks.
{
"workflow_id": "hybrid_local_cloud_code_review",
"tasks": [
{
"id": "extract_local_code",
"action_type": "http_request",
"target": "http://localhost:11434/v1/chat/completions",
"payload": {
"body": {
"model": "llama3.2:3b",
"messages": [{"role": "user", "content": "Extract function signatures..."}]
}
},
"output": { "store_as": "local_extracted_headers" }
},
{
"id": "cloud_code_review",
"trigger": "after:extract_local_code",
"action_type": "http_request",
"target": "https://openrouter.ai/api/v1/chat/completions",
"payload": {
"headers": { "Authorization": "Bearer {{env.OPENROUTER_API_KEY}}" },
"body": {
"model": "anthropic/claude-opus-5",
"messages": [{"role": "user", "content": "Review these signatures: {{local_extracted_headers.choices[0].message.content}}"}]
}
}
}
]
}
A saved recipe is a file, and the engine can run it with no AI connected at all. That makes Vibra-Ingenn a lightweight local automation runner as well as a tool for AI assistants.
Once you've used an AI to compile a JSON recipe once, you can save it and run it as often as you like from the dashboard, the command line or a scheduled task—with no LLM connection and no API token costs — provided the recipe contains no ai_call step.
You don't need to pay API token costs to run the same tasks every day. The LLM is only the compiler (writing the JSON recipe once). Vibra-Ingenn is the runtime engine, executing saved recipes natively at machine speeds.