the platform
Self-Hosted AI Agent Orchestration: A Full Tour
Clone the repo, bring the stack up, install a model, then build, run, and trace an agent — the full tour.
Walkthrough · 5:54
Voice In, Voice Out: A No-Code AI Agent Pipeline
What This Video Covers
A voice agent built entirely on the canvas and served by local engines. Speech goes in, Whisper transcribes it, a small local model (Llama 3.2) answers in two or three plain sentences, and Kokoro reads the answer back. The video installs the voice sample from Settings → Samples, after installing the two speech models from the registry.
It walks through each node — the audio input, the transcribe node and its error branch, the LLM's system prompt, the speak node — then records a question in the test panel, plays the spoken answer, and reads the run trace to see where the time goes: text-to-speech is the slowest step. It ends by showing that any of these models can be swapped, for another local model or a hosted API, without touching the agent.
Try It Yourself
More Walkthroughs
retrieval
Give Your LLM a Memory: RAG with File Upload and Web Crawl
Upload a PDF, crawl a docs site, and wire both into an agent as retrieval tools — no code.
hybrid
Run Local and Remote AI Models in One Agent
A local model triages and handles the easy work; the hard question goes to a hosted API — one graph, each branch calling its own model.
deploy
How to Deploy TheYgent to Kubernetes
Build the images, load them into a cluster, apply the manifests, and sign in — the same stack, on Kubernetes.