The set of software boundaries, interaction surfaces, and protocol contracts through which humans, AI agents, and system components communicate. In the context of LLMs and spatial computing, interfaces include node-based visual editors, chat frontends, API gateways, and multimodal input layers that mediate access to underlying AI or infrastructure capabilities.
Semantic Classification
Content
Node based
- Rivet (ironcladapp.com) is a Human-Agent Interaction Surfaces for Infrastructure Hardware and Edge based Agents using Large Language Models
- Node-Graph Visual Programming Interface is a Human-Agent Interaction Surfaces for Large Language Models which is open source and locally hosted. also see Langflow
- Langflow Human-Agent Interaction Surfaces for Large Language Models builder, with slightly more features than Node-Graph Visual Programming Interface which is another one I need to try soon from their github
- ComfyUI is a Human-Agent Interaction Surfaces for Stable Diffusion Image Model. It allows very high levels of control over Diffusion Models by leveraging open source extensions and a vibrant developer and creator community. The tool’s flexibility and efficiency comes from its innovative design philosophy, that prioritizes user customization and interaction.
LLM and multimodal local
- Description: Web-based UI inspired by ChatGPT, designed for high extensibility.
- Features:
- Workspaces for personalised assistants (similar to GPT’s custom setups).
- OpenAI-compatible endpoints for streamlined backend integration.
- Optimised for responsiveness, especially on touchscreen devices.
- Use Cases: General-purpose use, roleplay (RP), and advanced configuration.
- Limitations: Lack of comprehensive documentation remains a significant barrier.
- Link: Open WebUI GitHub
- Description: Web-based interface focusing on customisation and versatility.
- Features:
- Presets for reusable configurations.
- YAML and
.envsupport for endpoint and key management.
- Limitations:
- Export/migration tools are cumbersome.
- Slower responsiveness compared to Open WebUI.
- Link: LibreChat GitHub
- Description: Universal adapter popular for roleplay.
- Features:
- Wide compatibility with backends like Koboldcpp.
- Growing focus on general use beyond RP.
- Limitations: Outdated UI design limits appeal.
- Link: SillyTavern GitHub
- Description: Standalone desktop application for local inference.
- Features:
- Easy to use with a modern UI.
- Suitable for new users and traditional “Windows-style” workflows.
- Limitations: Closed source; outbound connections for updates raise privacy concerns.
- Link: LM Studio
- Description: Native macOS application with extensive feature sets.
- Features:
- Integrates advanced tools like text-to-speech (TTS).
- Highly optimised for macOS environments.
- Link: Msty App
- Description: Lightweight Android app for local models.
- Features:
- Mobile-focused with offline support.
- Works as a “SillyTavern Lite” alternative.
-
Link: ChatterUI GitHub
- Strengths:
- Lightweight backend with robust feature support (e.g., roleplay, text-to-speech).
- Integrates well with Open WebUI and SillyTavern.
- Limitations: UI is functional but lacks visual polish.
- Link: Koboldcpp GitHub
- Strengths:
- Minimalist server UI with OpenAI-compatible API.
- Excellent for developers due to fast updates.
- Limitations: Limited feature set compared to Oobabooga and Open WebUI.
- Link: Llama.cpp GitHub
- Strengths:
- Broad feature set including image generation and voice capabilities.
- Stable for solo usage.
- Limitations: Slower performance compared to newer backends like TabbyAPI or vLLM.
-
Link: Oobabooga GitHub
- Open WebUI: Can integrate vision models, image generation, TTS, and more with third-party tools like Azure or Together.ai.
- Koboldcpp: Supports some multimodal backends but lacks native syntax highlighting.
- LM Studio: Quick model mounting for API integrations.
-
AnythingLLM: Flexible for agent development and experimentation.
- For Beginners: LM Studio or Msty provide ease of use with minimal setup.
- For Advanced Users: Open WebUI offers extensive customisation and backend compatibility.
- For Roleplay: SillyTavern excels in flexibility with multiple backends.
- For Multimodal Needs: Combine Open WebUI with specific vision or TTS backends.
-
For Developers: Use Llama.cpp for rapid updates or Koboldcpp for lightweight integration.
- Open WebUI GitHub
- LibreChat GitHub
- SillyTavern GitHub
- Koboldcpp GitHub
- Llama.cpp GitHub
- Oobabooga GitHub
- Msty App
- ChatterUI GitHub