Model Creator — Workspace Models for Open WebUI

An Open WebUI Tool that interviews you, drafts a Workspace Model (agent preset on top of a base LLM), and persists it via POST /api/v1/models/create — private by default.

Pair with Skill Creator to build Skills, then bind them onto Models.

License: MIT · Author: IANUSTEC · Requires Open WebUI >= 0.10.0

Features

• Saves for real — does not only dump a draft in chat; calls the Models API with the user’s session token.
• Native function calling by default (params.function_calling = "native").
• Auto-bind from context: knowledge, tools, skills, actions, filters, capabilities, default features, and builtin tools.
• Ask when vague: if the draft has little context, lists everything visible to the user and asks what to enable (list_model_options / bindings_confirmed).
• Private by default — empty access_grants (owner-only).
• MODEL.md or JSON input with YAML frontmatter.
• Single-file — paste into Workspace → Tools.

Requirements

• Open WebUI >= 0.10.0 (Workspace Models + Skills APIs)
• Permission: workspace.models (or admin)
• Python deps (auto-installed from frontmatter): httpx, pydantic, PyYAML

Installation

Option A — Open WebUI community

1. Open the tool page on the Open WebUI community site.
2. Click Get / Import.

Option B — manual

1. Workspace → Tools → +
2. Paste model_creator.py
3. Save and enable the tool on a model/chat

Usage

The model should:

1. Interview: purpose, base LLM, system prompt, domain signals (RAG, web, code…).
2. Draft MODEL.md (or JSON).
3. Call create_model(content, context=...).
4. If the tool asks for bindings, choose ids (or call again with bindings_confirmed=true).

Example MODEL.md

See examples/python-tutor.md.

---
id: python-tutor
name: Python Tutor
base_model_id: gpt-4o
description: Patient Python tutor for beginners. Use for learning Python,
  debugging student code, or explaining concepts.
temperature: 0.4
function_calling: native
system: |
  You are a patient Python tutor for {{ USER_NAME }}.
  Prefer short examples and check understanding.
---

Tools exposed

Function	Purpose
create_model	Parse, auto-bind / ask, persist
update_model	Update an existing model by id
validate_model	Validate without saving
list_model_options	List knowledge/tools/skills/actions visible to you

Valves

Valve	Default	Meaning
default_private	true	Empty access_grants
default_active	true	is_active=true
default_base_model_id	""	Fallback base LLM id
auto_bind_from_context	true	Infer bindings from text
ask_when_vague	true	Ask before creating vague agents
owui_api_base	""	Override API base (else request.base_url)

Related

• Skill Creator
• Generate Documents
• Generate Slides
• Generate Spreadsheets

License

MIT — see LICENSE.