Studio assistant backend

Shared model service for the Hou-lm / Houdini and Ae-lm / After Effects clients.

Architecture

flowchart TB
    subgraph clients["Artist workstation"]
        H["Hou-lm · Houdini"]
        A["Ae-lm · After Effects"]
    end
    subgraph backend["studio-assistant-backend"]
        W["Open WebUI · accounts and API"]
        G["LiteLLM · model routing"]
        W <--> G
    end
    H <--> W
    A <--> W
    G <--> L["Local models · Ollama"]
    G <--> C["Cloud models"]

Loading

Clients send context, images and tool results through the service. Commands execute locally in Houdini or AE. The Compose example supplies Open WebUI, LiteLLM, PostgreSQL and GPU-backed Ollama; an existing compatible service can also serve the clients.

Setup

1. Follow server setup to deploy and configure models.
2. Give artists the Open WebUI base URL and their user API keys.
3. Install Hou-lm or Ae-lm on workstations.

Models need tool-call support; visual workflows also need image input. Backend files are not required on artist workstations.

Existing stack updates · Optional LiteLLM stream adapter · Development · Security · MIT