What problem does it solve? Accessing modern AI models like FLUX, Veo, Claude, and Tavily normally requires separate accounts, SDKs, and GPU infrastructure for each provider. This Skill unifies 250+ serverless AI apps behind a single CLI so you can run image generation, video creation, LLM calls, web search, 3D generation, and Twitter automation with one command. ## Core Features & Use Cases - Unified App Execution: Run any app with belt app run <namespace/app> --input <json>, with automatic local file uploads for images, audio, and video inputs. - App Discovery & Samples: Search, filter, and inspect apps with belt app list, belt app search, and generate ready-to-edit input templates via belt app sample. - Async Task Management: Submit long-running jobs with --no-wait and poll results later using belt task get <task-id>. - Use Case: Generate a product video by running belt app run google/veo-3-1-fast --input '{"prompt": "drone shot over a forest at sunset"}', then post it to Twitter with x/post-create in the same session. ## Quick Start Install the CLI with the inference.sh install script, run belt login, then ask the agent to run an app such as generating an image with falai/flux-dev-lora using a text prompt.
