Run cloud-based AI apps via the inference.sh CLI with JSON outputs.
npx skills add https://github.com/MarbleSodas/Mavis --skill inference-sh-cli-marblesodas
Please help me install this Agent Skill. Skill: inference-sh-cli Source: https://github.com/MarbleSodas/Mavis/tree/main/optional-skills/devops/cli Command: npx skills add https://github.com/MarbleSodas/Mavis --skill inference-sh-cli-marblesodas
It solves the problem of juggling many AI provider APIs by giving you a single, consistent CLI to discover and run image, video, LLM, and search apps in the cloud.
Example: A user wants both an image and a short video for a product launch—first search for the best matching image app ID, run it with a prompt, then search and run a video app using the returned style/theme guidance.
Use the inference-sh CLI to generate media by running an app after searching for its app ID, for example: "User asks the assistant to use the inference.sh CLI (infsh) to generate a FLUX image for the prompt 'a sunset over mountains' and return the resulting MEDIA URLs."
High-intent search queries and answers about installing and using this skill.
The inference.sh CLI solves juggling multiple provider APIs by offering a single, consistent terminal interface to discover and run cloud-based image, video, LLM, and search apps without needing provider-specific integration.
To authenticate and run inference.sh commands, you need authenticated access to the infsh tool in your terminal environment, which allows you to execute cloud AI apps and parse structured JSON results.
You find the correct app ID before generating media by using the CLI's app discovery-first workflow, which utilizes search and list commands to locate the exact app ID dynamically instead of hard-coding potentially brittle identifiers.
Yes, you can get JSON outputs with media URLs from AI image generation by executing infsh commands with the --json flag, which parses the results to surface direct links for your generated images and videos.
The inference.sh CLI supports workflows beyond image generation, covering video generation, avatar and lipsync creation, AI search and research, as well as 3D, TTS, audio, and automation categories.
The best way to avoid brittle hard-coding when running cloud AI apps is to use a discovery-first workflow that performs app lookups by name via terminal commands before execution, ensuring deterministic and accurate app selection.
HarriesRun 150+ AI apps via the infsh CLI with JSON output.
FatOtterExtract JPEG thumbnails and metadata from media files using ffprobe and Pillow.
eddieluengDiscover and run inference.sh apps from a terminal CLI with JSON output.
skkeoriwAutomate image, video, and audio generation via ComfyUI REST and WebSocket APIs.
