LLM providers

Provider components install Sample middleware so a document can talk to cloud or local models with no custom runtime wiring. Model routing is by the model prop.

Built-in providers

These ship in packages/core/components/:

AnthropicProvider.md — cloud models via the Anthropic API.
OllamaProvider.md — local models via a running Ollama server.
LlamafileProvider.md — local models via llamafile.
Sample.md — the sampling call itself.
Instruction.md — surfaces a system prompt as visible, composable content.

One document, two models

Providers nest. <Sample> routes each prompt to the matching model and combines the results — all in a single markdown file.

packages/core/examples/hello-world.md

---
title: Hello World
---

# {meta.title}

<AnthropicProvider model="claude-opus-4-5">
  <OllamaProvider model="llama3.2">
    <Instruction system="You are a creative comedian.">
      <Sample model="llama3.2">
        Smart: <Sample prompt="Say something smart" model="claude-opus-4-5" />
        Joke:  <Sample prompt="Tell me a joke" model="llama3.2" />
        Combine Smart and Joke into one smart joke
      </Sample>
    </Instruction>
  </OllamaProvider>
</AnthropicProvider>

Running provider documents

For this release, provider documents need the built-in components on the search path:

xmd run packages/core/examples/hello-world.md --component-dir packages/core/components

The example above also needs an ANTHROPIC_API_KEY in the environment and a local Ollama server running llama3.2. Embedding the built-in components into the binary (so --component-dir isn't required) is on the roadmap.

Next: Reference →