Model Management
ModelFlow supports 29 model providers. Manage cloud APIs and local Ollama models in one place, each with roles, tools, and context parameters.
Supported Providers
Text models: OpenAI, Anthropic, DeepSeek, Kimi, Qwen, Gemini, Grok, Doubao, GLM, Baichuan, Hunyuan, Yi, StepFun, MiniMax, Mistral, Ollama, and custom endpoints. Image: DALL·E, Stable Diffusion, Wanxiang. Video: Wanxiang, Vidu, Kling, Runway, Pika. TTS: EdgeTTS, Browser TTS.
| Category | Examples |
|---|---|
| Text / Chat | OpenAI, Anthropic, DeepSeek, Kimi, Qwen, Gemini, Grok, Doubao, GLM, Baichuan, Hunyuan, Yi, StepFun, MiniMax, Mistral, Ollama |
| Image | DALL·E, Stable Diffusion, Wanxiang |
| Video | Wanxiang, Vidu, Kling, Runway, Pika |
| Audio / TTS | EdgeTTS, Browser TTS |
Adding a Model
In Model Management, click Add Model and fill in name, provider, Base URL, API Key, model ID, max context, temperature, timeout, and retries. You can also bind a role, enable tools, and set capability tags and skill level.
Connection Test
- OK Success: API reachable and model responds
- Auth Auth OK but test rejected: API key valid, model name may be wrong
- Fail Failed: unreachable or authentication failed
Ollama Local Models
Choose Ollama as the provider and ModelFlow auto-discovers installed local models from http://localhost:11434/api/tags. Make sure Ollama is running.
Enable / Disable
Enable or disable models at any time. Disabled models won't appear in @mentions or workflow model pickers. Free tier allows up to 3 enabled models.
Default Role
Each model can be bound to a role. The role defines the model's identity—system prompt, capability boundaries, default toolset. The model is just a replaceable backend; the role gives AI persistent identity.