Model Provider Examples

OpenAI

GPT-4o with function calling, streaming, and per-request options.

import { Agent, openai, defineTool } from "@agentium/core";
import { z } from "zod";

const agent = new Agent({
  name: "openai-agent",
  model: openai("gpt-4o", {
    apiKey: process.env.OPENAI_API_KEY,
  }),
  instructions: "You are a helpful assistant with tool access.",
  tools: [
    defineTool({
      name: "get_weather",
      description: "Get current weather for a city",
      parameters: z.object({ city: z.string() }),
      execute: async ({ city }) => `${city}: 22°C, Sunny`,
    }),
  ],
});

const result = await agent.run("What's the weather in Tokyo?");
console.log(result.text);

for await (const chunk of agent.stream("Tell me a joke")) {
  if (chunk.type === "text") process.stdout.write(chunk.text);
}

Anthropic

Claude Sonnet with extended thinking for step-by-step reasoning.

import { Agent, anthropic } from "@agentium/core";

const agent = new Agent({
  name: "claude-agent",
  model: anthropic("claude-sonnet-4-20250514", {
    apiKey: process.env.ANTHROPIC_API_KEY,
  }),
  instructions: "Think step by step before answering complex questions.",
  reasoning: {
    enabled: true,
    budgetTokens: 8_000,
  },
});

const result = await agent.run(
  "A bat and a ball cost $1.10 in total. The bat costs $1 more than the ball. How much does the ball cost?",
);

console.log("Thinking:", result.thinking);
console.log("Answer:", result.text);

Google Gemini

Gemini with multimodal input — images, audio, PDFs, and spreadsheets.

import { Agent, google, type ContentPart } from "@agentium/core";
import { readFileSync } from "node:fs";

const agent = new Agent({
  name: "gemini-agent",
  model: google("gemini-2.5-flash", {
    apiKey: process.env.GOOGLE_API_KEY,
  }),
  instructions: "You analyze documents and images with precision.",
});

const imageResult = await agent.run([
  { type: "text", text: "What's in this image?" },
  {
    type: "image",
    data: readFileSync("photo.jpg").toString("base64"),
    mimeType: "image/jpeg",
  },
] as ContentPart[]);

console.log(imageResult.text);

const pdfResult = await agent.run([
  { type: "text", text: "Summarize the key findings." },
  {
    type: "file",
    data: readFileSync("report.pdf").toString("base64"),
    mimeType: "application/pdf",
    filename: "report.pdf",
  },
] as ContentPart[]);

console.log(pdfResult.text);

Vertex AI

Gemini via Google Cloud with project-scoped authentication for enterprise deployments.

import { Agent, vertex } from "@agentium/core";

const agent = new Agent({
  name: "vertex-agent",
  model: vertex("gemini-2.5-flash", {
    project: process.env.GCP_PROJECT_ID,
    location: "us-central1",
  }),
  instructions: "You are an enterprise assistant running on Google Cloud.",
});

const result = await agent.run("Summarize the benefits of serverless architecture.");
console.log(result.text);

Ollama

Local model with no API key — fully offline, privacy-first inference.

import { Agent, ollama, defineTool } from "@agentium/core";
import { z } from "zod";

const agent = new Agent({
  name: "local-agent",
  model: ollama("llama3.1", {
    host: "http://localhost:11434",
  }),
  instructions: "You are a helpful assistant running locally. No data leaves this machine.",
  tools: [
    defineTool({
      name: "calculator",
      description: "Evaluate a math expression",
      parameters: z.object({ expression: z.string() }),
      execute: async ({ expression }) => {
        const sanitized = expression.replace(/[^0-9+\-*/().%\s]/g, "");
        return String(new Function(`return (${sanitized})`)());
      },
    }),
  ],
});

const result = await agent.run("What is 2^16 + 3^10?");
console.log(result.text);

DeepSeek

Reasoning model with chain-of-thought for complex math and logic problems.

import { Agent, deepseek } from "@agentium/core";

const agent = new Agent({
  name: "deepseek-reasoner",
  model: deepseek("deepseek-reasoner", {
    apiKey: process.env.DEEPSEEK_API_KEY,
  }),
  instructions: "Solve problems step by step with rigorous reasoning.",
});

const result = await agent.run(
  "A train leaves Station A at 60 mph. Another leaves Station B (300 miles away) at 80 mph toward A. When do they meet?",
);
console.log(result.text);

Mistral

Code generation with Codestral for programming tasks.

import { Agent, mistral } from "@agentium/core";

const agent = new Agent({
  name: "code-agent",
  model: mistral("codestral-latest", {
    apiKey: process.env.MISTRAL_API_KEY,
  }),
  instructions: "You are an expert programmer. Write clean, well-documented TypeScript.",
});

const result = await agent.run(
  "Write a TypeScript function that implements binary search on a sorted array.",
);
console.log(result.text);

xAI Grok

Grok with web search tool for grounded, real-time answers.

import { Agent, xai, defineTool } from "@agentium/core";
import { z } from "zod";

const agent = new Agent({
  name: "grok-agent",
  model: xai("grok-3", {
    apiKey: process.env.XAI_API_KEY,
  }),
  instructions: "You are a witty and knowledgeable assistant with live web search.",
  tools: [
    defineTool({
      name: "web_search",
      description: "Search the web for current information",
      parameters: z.object({ query: z.string() }),
      execute: async ({ query }) => `Top results for "${query}": ...`,
    }),
  ],
});

const result = await agent.run("What are the latest breakthroughs in battery technology?");
console.log(result.text);

Perplexity

Search-grounded answers with citations and domain filtering.

import { Agent, perplexity } from "@agentium/core";

const agent = new Agent({
  name: "perplexity-researcher",
  model: perplexity("sonar-pro", {
    apiKey: process.env.PERPLEXITY_API_KEY,
    search: {
      searchMode: "academic",
      searchRecencyFilter: "month",
      searchDomainFilter: ["nature.com", "arxiv.org", "sciencedirect.com"],
      returnRelatedQuestions: true,
      searchContextSize: "high",
    },
  }),
  instructions: "You are a research assistant. Cite your sources.",
});

const result = await agent.run(
  "What are the latest FDA-approved treatments for Alzheimer's disease?",
);
console.log(result.text);

const raw = result.raw as any;
if (raw?.citations) console.log("Sources:", raw.citations);
if (raw?.related_questions) console.log("Related:", raw.related_questions);

Cohere

Command R+ optimized for RAG with document search.

import { Agent, cohere, defineTool } from "@agentium/core";
import { z } from "zod";

const agent = new Agent({
  name: "cohere-rag-agent",
  model: cohere("command-r-plus", {
    apiKey: process.env.CO_API_KEY,
  }),
  instructions:
    "You are a knowledge-grounded assistant. Always use the search tool before answering. Cite your sources.",
  tools: [
    defineTool({
      name: "search_docs",
      description: "Search internal documentation",
      parameters: z.object({ query: z.string() }),
      execute: async ({ query }) =>
        JSON.stringify([
          { title: "Deployment Guide", content: `Guide for: ${query}...`, source: "docs/deploy.md" },
          { title: "API Reference", content: `API info for: ${query}...`, source: "docs/api.md" },
        ]),
    }),
  ],
  maxToolRoundtrips: 3,
});

const result = await agent.run("How do I deploy our service to production?");
console.log(result.text);

Meta Llama

Llama 4 Scout via the official Meta API with vision support.

import { Agent, meta, type ContentPart } from "@agentium/core";
import { readFileSync } from "node:fs";

const agent = new Agent({
  name: "llama-agent",
  model: meta("Llama-4-Scout-17B-16E-Instruct", {
    apiKey: process.env.LLAMA_API_KEY,
  }),
  instructions: "You are a helpful, harmless, and honest assistant.",
});

const textResult = await agent.run("Explain the difference between TCP and UDP.");
console.log(textResult.text);

const imageData = readFileSync("diagram.png").toString("base64");
const visionResult = await agent.run([
  { type: "text", text: "Describe this architecture diagram." },
  { type: "image", data: imageData, mimeType: "image/png" },
] as ContentPart[]);
console.log(visionResult.text);

AWS Bedrock

Foundation models (Mistral, Nova, Llama) via AWS with IAM authentication.

import { Agent, awsBedrock, defineTool } from "@agentium/core";
import { z } from "zod";

const agent = new Agent({
  name: "bedrock-agent",
  model: awsBedrock("mistral.mistral-large-2402-v1:0", {
    region: "us-east-1",
  }),
  instructions: "You are a helpful assistant running on AWS Bedrock.",
  tools: [
    defineTool({
      name: "lookup_order",
      description: "Look up an order by ID",
      parameters: z.object({ orderId: z.string() }),
      execute: async ({ orderId }) =>
        JSON.stringify({ id: orderId, status: "shipped", eta: "2026-03-05" }),
    }),
  ],
  maxToolRoundtrips: 3,
});

const result = await agent.run("Where is my order ORD-12345?");
console.log(result.text);

AWS Claude

Claude via Bedrock with extended thinking and AWS billing.

import { Agent, awsClaude } from "@agentium/core";

const agent = new Agent({
  name: "aws-claude-agent",
  model: awsClaude("us.anthropic.claude-sonnet-4-20250514-v1:0", {
    awsRegion: "us-east-1",
  }),
  instructions: "You are an enterprise assistant. Think step by step.",
  reasoning: {
    enabled: true,
    budgetTokens: 4_000,
  },
});

const result = await agent.run(
  "Analyze the trade-offs between microservices and monolithic architectures for a 10-person startup.",
);
console.log("Thinking:", result.thinking);
console.log("Answer:", result.text);

Azure OpenAI

GPT-4o on Azure with enterprise compliance and private endpoints.

import { Agent, azureOpenai, defineTool } from "@agentium/core";
import { z } from "zod";

const agent = new Agent({
  name: "azure-agent",
  model: azureOpenai("gpt-4o", {
    apiKey: process.env.AZURE_OPENAI_API_KEY,
    endpoint: process.env.AZURE_OPENAI_ENDPOINT,
    deployment: "my-gpt4o-deployment",
    apiVersion: "2024-10-21",
  }),
  instructions: "You are an enterprise assistant for internal use.",
  tools: [
    defineTool({
      name: "search_docs",
      description: "Search internal documentation",
      parameters: z.object({ query: z.string() }),
      execute: async ({ query }) => `Found 3 results for "${query}"...`,
    }),
  ],
});

const result = await agent.run("Find the onboarding docs for new engineers.");
console.log(result.text);

Azure Foundry

Open-source models (Phi, Llama, Mistral) hosted on Azure AI Foundry.

import { Agent, azureFoundry } from "@agentium/core";

const agent = new Agent({
  name: "foundry-agent",
  model: azureFoundry("Phi-4", {
    apiKey: process.env.AZURE_API_KEY,
    endpoint: process.env.AZURE_ENDPOINT,
  }),
  instructions: "You are a fast, efficient assistant powered by Phi-4 on Azure.",
});

const result = await agent.run("Explain recursion with a simple example.");
console.log(result.text);

OpenAI-Compatible

Connect to Together, Groq, Fireworks, or OpenRouter with a custom baseURL.

import { Agent, openai } from "@agentium/core";

// Together AI — 100+ open-source models
const togetherAgent = new Agent({
  name: "together-agent",
  model: openai("meta-llama/Meta-Llama-3.1-70B-Instruct-Turbo", {
    apiKey: process.env.TOGETHER_API_KEY,
    baseURL: "https://api.together.xyz/v1",
  }),
  instructions: "You are a helpful assistant.",
});

// Groq — ultra-fast inference
const groqAgent = new Agent({
  name: "groq-agent",
  model: openai("llama-3.3-70b-versatile", {
    apiKey: process.env.GROQ_API_KEY,
    baseURL: "https://api.groq.com/openai/v1",
  }),
  instructions: "You are a fast, helpful assistant.",
});

// Fireworks — optimized open-source inference
const fireworksAgent = new Agent({
  name: "fireworks-agent",
  model: openai("accounts/fireworks/models/llama-v3p1-70b-instruct", {
    apiKey: process.env.FIREWORKS_API_KEY,
    baseURL: "https://api.fireworks.ai/inference/v1",
  }),
  instructions: "You are a helpful assistant.",
});

// OpenRouter — access models from multiple providers
const openrouterAgent = new Agent({
  name: "openrouter-agent",
  model: openai("anthropic/claude-sonnet-4-20250514", {
    apiKey: process.env.OPENROUTER_API_KEY,
    baseURL: "https://openrouter.ai/api/v1",
  }),
  instructions: "You are a helpful assistant.",
});

const result = await togetherAgent.run("Explain how transformers work in 3 sentences.");
console.log(result.text);

Custom Provider

Implement the ModelProvider interface for any LLM API.

import {
  Agent, registry,
  type ModelProvider, type ChatMessage,
  type ModelResponse, type StreamChunk,
} from "@agentium/core";

class MyProvider implements ModelProvider {
  readonly providerId = "my-provider";
  readonly modelId: string;
  private apiKey: string;

  constructor(modelId: string, config: { apiKey: string }) {
    this.modelId = modelId;
    this.apiKey = config.apiKey;
  }

  async generate(messages: ChatMessage[]): Promise<ModelResponse> {
    const lastMessage = messages[messages.length - 1]?.content;
    const res = await fetch("https://api.my-llm.com/v1/chat", {
      method: "POST",
      headers: {
        Authorization: `Bearer ${this.apiKey}`,
        "Content-Type": "application/json",
      },
      body: JSON.stringify({ model: this.modelId, prompt: lastMessage }),
    });
    const data = await res.json();
    return {
      message: { role: "assistant", content: data.text },
      usage: { promptTokens: data.usage.input, completionTokens: data.usage.output, totalTokens: data.usage.total },
      finishReason: "stop",
      raw: data,
    };
  }

  async *stream(): AsyncGenerator<StreamChunk> {
    yield { type: "text", text: "Streaming not implemented" };
    yield { type: "finish", finishReason: "stop", usage: undefined };
  }
}

registry.register("my-provider", (modelId, config) =>
  new MyProvider(modelId, config as { apiKey: string }),
);

const agent = new Agent({
  name: "custom-agent",
  model: registry.resolve("my-provider", "my-model-v1", {
    apiKey: process.env.MY_API_KEY!,
  }),
  instructions: "You are a helpful assistant.",
});

const result = await agent.run("Hello!");
console.log(result.text);

Model Switching

Change models at runtime without rebuilding agents — useful for fallback strategies or A/B testing.

import { Agent, openai, anthropic, ollama } from "@agentium/core";

const models = {
  fast: openai("gpt-4o-mini"),
  smart: openai("gpt-4o"),
  reasoning: anthropic("claude-sonnet-4-20250514"),
  local: ollama("llama3.1"),
};

function createAgent(tier: keyof typeof models) {
  return new Agent({
    name: `assistant-${tier}`,
    model: models[tier],
    instructions: "You are a helpful assistant.",
  });
}

const fastAgent = createAgent("fast");
const smartAgent = createAgent("smart");

let result = await fastAgent.run("What is 2 + 2?");
console.log("Fast:", result.text);

result = await smartAgent.run("Analyze the geopolitical implications of AI regulation.");
console.log("Smart:", result.text);

// Runtime fallback: retry with a different model on failure
async function runWithFallback(input: string) {
  const tiers: (keyof typeof models)[] = ["fast", "smart", "reasoning"];
  for (const tier of tiers) {
    try {
      const agent = createAgent(tier);
      return await agent.run(input);
    } catch {
      console.log(`${tier} failed, trying next...`);
    }
  }
  throw new Error("All models failed");
}

const fallbackResult = await runWithFallback("Explain quantum entanglement.");
console.log(fallbackResult.text);

Core Examples

Multi-Agent & Orchestration

Knowledge & RAG

Deployment & Transport

Monitoring & Cost

Advanced & Integrations

End-to-End Applications

Model Provider Examples

OpenAI

Anthropic

Google Gemini

Vertex AI

Ollama

DeepSeek

Mistral

xAI Grok

Perplexity

Cohere

Meta Llama

AWS Bedrock

AWS Claude

Azure OpenAI

Azure Foundry

OpenAI-Compatible

Custom Provider

Model Switching

​OpenAI

​Anthropic

​Google Gemini

​Vertex AI

​Ollama

​DeepSeek

​Mistral

​xAI Grok

​Perplexity

​Cohere

​Meta Llama

​AWS Bedrock

​AWS Claude

​Azure OpenAI

​Azure Foundry

​OpenAI-Compatible

​Custom Provider

​Model Switching

OpenAI

Anthropic

Google Gemini

Vertex AI

Ollama

DeepSeek

Mistral

xAI Grok

Perplexity

Cohere

Meta Llama

AWS Bedrock

AWS Claude

Azure OpenAI

Azure Foundry

OpenAI-Compatible

Custom Provider

Model Switching