@onioneko/kiki-llm-anthropic

Kiki · Middleware framework for LLM agents

npm install @onioneko/kiki-llm-anthropic

Anthropic Claude LLM provider as middleware. Calls the API and pushes the assistant response onto ctx.messages, then calls next() to allow response-processing middleware downstream.

Usage

import { compose } from "@onioneko/kiki-core";
import { createAnthropicMiddleware } from "@onioneko/kiki-llm-anthropic";

const chain = compose([
  createAnthropicMiddleware({ model: "claude-sonnet-4-20250514" }),
]);

const ctx = {
  messages: [{ role: "user", content: "What is 2 + 2?" }],
  system: "Be concise.",
};
await chain(ctx);
// ctx.messages[1] is the assistant response

Configuration

model is required. Other fields are optional with sensible defaults.

createAnthropicMiddleware({
  model: "claude-sonnet-4-20250514", // required
  apiKey: "...",       // or ANTHROPIC_API_KEY env var
  maxTokens: 16384,    // default: 16384
  client: anthropic,   // optional: injected Anthropic client (for testing)
});

Behavior

• Non-terminal — calls next() after pushing the response, allowing response-processing middleware to be composed after the LLM.
• Prompt caching — applies cache_control on the system prompt and the penultimate user message.
• Multi-modal — handles both string and ContentBlock[] message content.
• Config precedence — explicit config > defaults.

Exports

• createAnthropicMiddleware(config) — Factory function returning a Middleware
• AnthropicConfig — Configuration type
• Re-exports from @onioneko/kiki-types: Message, ConversationContext, ContentBlock, TextBlock, ImageBlock