You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
This commit was created on GitHub.com and signed with GitHub’s verified signature.
New Features
Reference image inputs — ai image and ai text now accept repeatable --image references from local paths, file:// URLs, http(s):// URLs and data URLs
Vision stdin detection — ai text can distinguish piped image bytes from piped text, enabling image prompts from stdin without treating binary data as prompt text
Published CLI binary — packages ai as a Node-targeted dist/index.js bundle and points npm's bin/files metadata at dist, fixing global installs that could not execute the TypeScript source (#52)
Improvements
Node engine requirement — declares and documents the Node.js 20+ runtime requirement for the CLI package (#52)
Dynamic model discovery — replaced hardcoded fallback model lists with live fetches from the AI Gateway API, so new models are available immediately without a CLI update (#48)
Model capabilities & metadata — ai models now shows creator, capabilities, and pricing info for each model (#48)
Language-image model routing — language models tagged with image-generation (e.g. Gemini) are automatically detected and routed through generateText instead of generateImage (#48)
Stdin image support for language-image models — piping an image via stdin to a language-based image model now correctly passes the image data via the messages API (#48)
Improvements
--provider renamed to --creator — aligns with the AI Gateway's creator/model-name convention and reserves --provider for future multi-provider support (#48)
Deduplicated --json output — ai models --json now returns a flat array where each model appears once with all its capabilities (#48)
Gateway fetch resilience — failed gateway fetches are no longer permanently cached, so retries work correctly; requests time out after 5 seconds (#48)
Documentation overhaul — filled gaps across README, web docs, and added a SKILL.md for agent integration (#42)
Favicon & OG images — added favicon and Open Graph image generation to the web app (#40)
OG card refresh — updated OG card to match the portless design and removed outdated tagline (#41)
Bug Fixes
Mobile responsive overflow — fixed layout overflow on the landing page for mobile viewports (#47)
Breaking Changes
ai completions removed — the shell completions command has been removed (#48)
--provider flag renamed to --creator — update any scripts using --provider to use --creator instead (#48)