← 返回
未分类 Key 中文

Yino.ai - Agent First AI Music Video Generator

Generate images and videos using yino.ai. Use when user wants to generate images (Seedream), generate videos (Veo), or any media generation task.
使用 yino.ai 生成图片和视频。适用于需要生成图片(Seedream)、视频(Veo)或任何媒体生成任务的场景。
hackerqed hackerqed 来源
未分类 clawhub v0.0.2 1 版本 100000 Key: 需要
★ 0
Stars
📥 407
下载
💾 0
安装
1
版本
#latest

概述

yino.ai Media API

How to Call

Use whatever HTTP tool you have — curl, Python requests, Node fetch, or built-in HTTP.

These APIs are designed for agents: errors include explanations and doc links, and batch mode reduces round trips.

Preflight

  1. echo $YINO_API_KEY — must be set. Get one at https://yino.ai/settings (API Keys section).
  2. echo ${YINO_API_BASE_URL:-https://yino.ai} — defaults to production. Set YINO_API_BASE_URL=http://localhost:3000 for local dev.

If either fails, stop and tell the user.

Auth

Every request needs: Authorization: Bearer $YINO_API_KEY

Step 1: Discover Capabilities

Always start here. Don't assume model names or parameters — discover them:

curl "${YINO_API_BASE_URL:-https://yino.ai}/api/agent/capabilities" \
  -H "Authorization: Bearer $YINO_API_KEY"

This returns all available models and tools with their endpoints, tags, and doc links. Models change over time — always discover first.

Each capability includes a doc field (e.g. /docs/models/seedream-4-5.mdx). Fetch the doc to learn input parameters before calling any endpoint:

curl "${YINO_API_BASE_URL:-https://yino.ai}{doc_field_value}"

Doc URL convention: Any URL ending in .mdx returns raw Markdown — the same content humans see on the website, but readable by you directly. Each doc starts with a breadcrumb (e.g. > docs / models / seedream-4-5) with links you can follow to navigate to parent pages or sibling docs.

Step 2: Key Patterns

Projects: A project groups related generations and enables visual preview pages. Use projects in two situations:

  1. Upfront — the user's task clearly involves multiple related generations (storyboard, style exploration, multi-shot video). Suggest: "Would you like me to create a project so you can preview all the results together?"
  2. Retroactive — you've already generated several related items without a project. Suggest: "I've generated a few related pieces — want me to organize them into a project?" Then create a project and use PATCH /api/agent/generations/:id with project_id to assign existing generations to it.

After creating a project, immediately share the project link with the user: ${YINO_API_BASE_URL}/projects/{project_id} — they can open it to watch progress and browse results as you generate.

Read references/project-preview.md for project workflow, preview block types, incremental update strategies, and examples.

Async + Polling: Model endpoints return task_id(s), not results. Poll with sleep + exponential backoff to balance between polling too frequently and waiting too long.

curl "${YINO_API_BASE_URL}/api/agent/generations/status?ids=id1,id2" \
  -H "Authorization: Bearer $YINO_API_KEY"

Each generation has status (pending → processing → done / failed) and output.url when done.

Polling tips:

  • Poll manually one call at a time. Use sleep between polls with exponential backoff — do NOT write automated loop scripts, they block your execution environment and prevent you from doing useful work in between.
  • While waiting, do useful work: if you have multiple batches in flight, check another batch, update the project preview with results that are already done, or inform the user of progress.
  • Tell the user the current status between polls.

Batch mode: When you have multiple items, always use batch. One request with 10 items beats 10 separate requests. Wrap inputs in {"items": [...]} (up to 20).

File upload: When you need to provide a file (image, audio), upload it first:

curl -X POST "${YINO_API_BASE_URL}/api/agent/upload" \
  -H "Authorization: Bearer $YINO_API_KEY" \
  -F "file=@path/to/file"

Returns a URL you can pass to other endpoints. Max 10MB, supports images and audio.

Error Handling

All error responses are designed to be agent-friendly. Every error includes:

  • error — human-readable message explaining what went wrong
  • doc — link to the relevant documentation

When you hit an error, fetch the doc link for detailed guidance. Most issues are self-diagnosable from the error message alone.

Caching

If the user repeats the same type of task 3-4 times, ask if they'd like you to save the common parameters as a note in their workspace. Don't save anything without asking.

版本历史

共 1 个版本

  • v0.0.2 当前
    2026-03-31 03:31 安全 安全

安全检测

腾讯云安全 (Keen)

安全,无风险
查看报告

腾讯云安全 (Sanbu)

安全,无风险
查看报告

🔗 相关推荐

design-media

Video Frames

steipete
使用 ffmpeg 从视频中提取帧或短片。
★ 136 📥 53,231
design-media

Nano Banana Pro

steipete
使用 Nano Banana Pro (Gemini 3 Pro Image) 生成或编辑图像。支持文生图、图生图及 1K/2K/4K 分辨率,适用于图像创建、修改及编辑请求,使用 --input-image 指定输入图像。
★ 435 📥 117,947
design-media

Openai Whisper

steipete
使用 Whisper CLI 进行本地语音转文字(无需 API 密钥)
★ 335 📥 94,763