ccapi_skill

Overview

This skill provides an OpenAI-compatible gateway to 60+ multimodal AI models across text, image, video, and audio. It lets you call multiple providers (OpenAI, Anthropic, Google, Bytedance, Suno, ElevenLabs, etc.) through a single API endpoint by only changing the base_url in existing OpenAI SDK usage. Use it to centralize model selection, failover, and billing while retaining familiar SDK calls.

How this skill works

The skill exposes the CCAPI endpoint (https://api.ccapi.ai/v1) and accepts provider/model-id strings (e.g., openai/gpt-5.2). It routes requests to the chosen provider, supports chat completions, streaming, function calling, image/video/audio generation, async task polling, and returns OpenAI-compatible responses and error formats. Use the provided endpoints for chat, images, video tasks, and audio; GET /v1/models lists available models and pricing.

When to use it

• When migrating or building apps that already use the OpenAI SDK but need multiple providers
• When you want a single endpoint to access text, image, video, and audio models
• When you need provider failover, transparent USD billing, and centralized model discovery
• When generating async video or long-running media tasks and polling for results
• When you want streaming chat or function-calling support across alternative LLMs

Best practices

• Pass models in provider/model-id format (e.g., openai/gpt-5.2) to ensure correct routing
• Set base_url to https://api.ccapi.ai/v1 and send Authorization: Bearer <key> with SDK calls
• Poll video generation task endpoints (GET /v1/video/generations/{task_id}) for completion
• Use GET /v1/models to filter by provider, context window, and pricing before choosing a model
• Enable streaming for chat where low-latency token delivery matters; handle standard OpenAI-style errors

Example use cases

• Replace base_url in an existing OpenAI Python or Node.js client to test alternative LLMs without code changes
• Generate images with google/nano-banana and produce music with suno/chirp-v5 from a single backend
• Submit a text-to-video job with bytedance/seedance-2.0, poll the task, and serve the resulting video URL
• Use streaming on deepseek/deepseek-chat for a live assistant with real-time token output
• Aggregate model selection in a UI that lets users toggle providers and pricing per request