What problem does it solve? Calling different LLM providers normally means juggling separate SDKs, credential formats, and prompt templates. This Skill documents write-language, a package that wraps the Vercel AI SDK behind a single writeLanguageResponse() call with a registry of providers, models, and named prompt templates, so you can switch providers or debug failures without relearning each API. ## Core Features & Use Cases - One-call multi-provider generation: Call writeLanguageResponse() with a provider, apiKey, and agent name to generate HTML or Markdown output across Groq, OpenAI, Anthropic, Google, Vertex, Bedrock, Cloudflare, and more. - Prompt template registry: Use AGENT_PROMPTS templates like question, summarize, answer-cite-sources, and knowledge-graph-nodes, with before/after hooks that parse structured output into result.extract. - Provider and model registry: Query LANGUAGE_MODELS, getModelsByProvider(), or getModelsByCapability() to discover valid providers and vision- or reasoning-capable models. - Use Case: You need to add a new LLM provider to your app. Follow the Skill to add a case to createLLMProvider and a registry entry, then validate it with the packed apiKey format rules for Cloudflare, Vertex, or Bedrock. ## Quick Start Ask the AI to call writeLanguageResponse with provider groq, your API key, the question agent, and a query, then read the error field before using the returned content.
