{
  "id": "page:docs-research-ai-cli",
  "_kind": "Page",
  "_file": "wiki/docs/research/ai-cli.md",
  "_cluster": "wiki",
  "attributes": {
    "nodeKind": "Page",
    "sourcePath": "docs/research/ai-cli.md",
    "sourceKind": "repo-docs",
    "title": "ai-cli — image + video gen via CLI",
    "displayName": "ai-cli — image + video gen via CLI",
    "slug": "docs/research/ai-cli",
    "articlePath": "wiki/docs/research/ai-cli.md",
    "article": "\n# ai-cli — image + video gen via CLI\n\n> **Status:** Assimilation research note (no code change beyond the atlas record).\n> **Date:** 2026-06-23.\n> **Tagline (requester):** \"Image + video gen via CLI.\"\n> **Method:** vendor homepage <https://ai-cli.dev> + npm/GitHub corroboration. Behavioral claims are vendor-stated and not independently exercised.\n> **Provenance (verified 2026-06-23):** open source, **Apache-2.0**, by **Vercel Labs** — `github.com/vercel-labs/ai-cli`, npm `ai-cli` (v0.3.1; npm self-describes as \"agent-native\" with \"predictable artifact outputs\").\n> **Atlas record:** `tool:ai-cli` — `packages/atlas/graph/domain/tools/ai-cli.yaml` (evidence `evidence:ai-cli-homepage-2026-06`, `evidence:ai-cli-npm-2026-06`).\n\n---\n\n## 0. Bottom line\n\nA terminal CLI for generating **text, images, and video** from prompts, routed through the **Vercel AI Gateway** to models from OpenAI, Anthropic, Google, Black Forest Labs, and ByteDance. Built for piping and automation: multi-model comparison, stdin/stdout chaining, inline terminal previews, JSON metadata.\n\n## 1. What it does\n\n`ai image/video/text \"prompt\"` → generated output, with live model discovery and the ability to run the same prompt across multiple providers for comparison.\n\n## 2. How it works\n\n- Routes to providers via the **Vercel AI Gateway**; short model names auto-resolve (e.g. `flux-2-pro`, `gpt-image-2`).\n- **Live model discovery** — no hardcoded model list (`ai models`).\n- **Multi-model comparison** — same prompt across providers simultaneously.\n- **Pipeable**: stdin/stdout chaining (e.g. `ai image | ai video`), inline terminal image preview via graphics protocols, JSON metadata output for automation.\n- Minimal setup; configured via environment variables.\n\n## 3. Install & usage\n\n```bash\nnpm install -g ai-cli\nai image \"prompt\"\nai video \"prompt\"\nai text  \"prompt\"\nai models            # list available models\n```\n\n## 4. Relevance to babysitter / agent workflows\n\n- A scriptable media-generation primitive an orchestrated agent could call for asset generation steps (e.g. producing diagrams/thumbnails/video in a pipeline), with JSON output suited to automation.\n- The multi-provider gateway model overlaps conceptually with the repo's own provider/transport-adapter layer — useful as a reference point, not a dependency.\n- Recorded as catalog awareness (`used_for: skill-area:video-processing`); not wired into babysitter.\n\n## 5. Comparison (atlas edges)\n\nNo direct `alternative_to` recorded — no equivalently-scoped multi-provider media-gen CLI currently in the catalog. Conceptually adjacent to the LLM-gateway tools (`tool:openrouter`, `tool:litellm`) but those are text/inference proxies, not media-gen CLIs.\n\n## 6. Caveats / open questions\n\n- License/OSS/repo **verified** (Apache-2.0, Vercel Labs) via npm + GitHub. Early version (0.3.x).\n- Provider/model list, pricing, auth model (env vars → which keys / Vercel AI Gateway billing), and video-gen latency/quality are vendor claims, not independently exercised here.\n",
    "documents": []
  },
  "outgoingEdges": [],
  "incomingEdges": [
    {
      "from": "page:docs",
      "to": "page:docs-research-ai-cli",
      "kind": "contains_page"
    }
  ]
}