Model Providers

Wunderland supports 13 LLM providers out of the box, from cloud APIs to local inference with Ollama. Each provider is identified by a string ID used throughout the configuration system.

Provider Overview

Provider	ID	Env Variable	Default Model	Small Model
OpenAI	`openai`	`OPENAI_API_KEY`	`gpt-4o`	`gpt-4o-mini`
Anthropic	`anthropic`	`ANTHROPIC_API_KEY`	`claude-sonnet-4-5-20250929`	`claude-haiku-4-5-20251001`
Ollama	`ollama`	(none -- local)	`llama3`	`llama3.2:3b`
AWS Bedrock	`bedrock`	`AWS_ACCESS_KEY_ID`	`anthropic.claude-sonnet`	`anthropic.claude-haiku`
Google Gemini	`gemini`	`GEMINI_API_KEY`	`gemini-2.0-flash`	`gemini-2.0-flash-lite`
GitHub Copilot	`github-copilot`	`GITHUB_COPILOT_TOKEN`	`gpt-4o`	`gpt-4o-mini`
Minimax	`minimax`	`MINIMAX_API_KEY`	`MiniMax-M2.1`	`MiniMax-VL-01`
Qwen	`qwen`	`QWEN_API_KEY`	`qwen-max`	`qwen-turbo`
Moonshot	`moonshot`	`MOONSHOT_API_KEY`	`kimi-k2.5`	`kimi-k2-instant`
Venice	`venice`	`VENICE_API_KEY`	`venice-default`	`venice-fast`
Cloudflare AI	`cloudflare-ai`	`CLOUDFLARE_API_TOKEN`	`@cf/meta/llama-3.3-70b-instruct-fp8-fast`	`@cf/meta/llama-3.1-8b-instruct`
Xiaomi Mimo	`xiaomi-mimo`	`XIAOMI_API_KEY`	`mimo-v2-flash`	`mimo-v2-flash`
OpenRouter	`openrouter`	`OPENROUTER_API_KEY`	`auto`	`auto`

Provider Details

OpenAI

Industry-standard cloud LLM provider with GPT-4o and o-series models.

Setup:

Get an API key from platform.openai.com
Set the environment variable:

export OPENAI_API_KEY="sk-..."

Available Models: gpt-4o-mini, gpt-4o, gpt-4.1-mini, gpt-4.1, o4-mini

Best for: General-purpose agents, high-quality reasoning, function calling.

Instead of an API key, you can authenticate with your ChatGPT Plus ($20/mo) or Pro ($200/mo) subscription using the same OAuth device code flow as the Codex CLI:

wunderland login              # Opens device code flow
wunderland auth-status        # Check token validity

Then set in agent.config.json:

{
  "llmProvider": "openai",
  "llmModel": "gpt-4o"
}

Use OPENAI_API_KEY in your environment or config. OpenAI OAuth subscription login is currently disabled in Wunderland. See the OpenAI OAuth Guide for current status and migration details.

Anthropic

Claude family of models, known for strong reasoning and long context windows.

Setup:

Get an API key from console.anthropic.com
Set the environment variable:

export ANTHROPIC_API_KEY="sk-ant-..."

Available Models: claude-sonnet-4-5-20250929, claude-haiku-4-5-20251001, claude-opus-4-6

Best for: Long-context analysis, nuanced reasoning, safety-conscious deployments.

Ollama (Local)

Run models locally with zero API costs. Requires Ollama installed on your machine.

Setup:

Install Ollama: ollama.ai
Pull a model:

ollama pull llama3
ollama pull llama3.2:3b  # Small model for routing/sentiment

No API key needed -- Ollama runs on localhost:11434 by default.

Available Models: llama3, llama3.2:3b, mistral, codellama, and any model available in the Ollama library.

Environment Variables:

Variable	Default	Description
`OLLAMA_HOST`	`http://localhost:11434`	Ollama server URL

Best for: Privacy-first deployments, offline use, development, cost-sensitive projects.

See the Ollama Local Setup guide for detailed instructions.

AWS Bedrock

Access foundation models through your AWS account with IAM-based authentication.

Setup:

Configure AWS credentials:

export AWS_ACCESS_KEY_ID="AKIA..."
export AWS_SECRET_ACCESS_KEY="..."
export AWS_REGION="us-east-1"

Ensure your IAM role has Bedrock model access permissions.

Available Models: anthropic.claude-sonnet, anthropic.claude-haiku

Best for: Enterprise AWS environments, compliance requirements, existing AWS infrastructure.

See the AWS Bedrock documentation for IAM setup.

Google Gemini

Google's Gemini family of models accessed through AI Studio.

Setup:

Get an API key from AI Studio
Set the environment variable:

export GEMINI_API_KEY="AI..."

Available Models: gemini-2.0-flash, gemini-2.0-flash-lite, gemini-2.5-pro

Best for: Multimodal tasks, Google ecosystem integration, cost-effective inference.

GitHub Copilot

Use GitHub Copilot's API with your Copilot subscription.

Setup:

Ensure you have an active GitHub Copilot subscription
Get a token from GitHub Copilot settings

export GITHUB_COPILOT_TOKEN="ghu_..."

Available Models: gpt-4o, gpt-4o-mini

Best for: Developers with existing Copilot subscriptions, code-focused agents.

Minimax

Chinese AI provider with strong multilingual capabilities.

Setup:

Get an API key from Minimax Platform

export MINIMAX_API_KEY="..."

Available Models: MiniMax-M2.1, MiniMax-VL-01

Best for: Chinese language tasks, vision-language applications.

Qwen

Alibaba's Qwen family of large language models.

Setup:

Get an API key from Qwen Portal

export QWEN_API_KEY="..."

Available Models: qwen-max, qwen-turbo

Best for: Chinese and multilingual tasks, cost-effective Asian market deployments.

Moonshot

Moonshot AI's Kimi models with strong reasoning capabilities.

Setup:

Get an API key from Moonshot Platform

export MOONSHOT_API_KEY="..."

Available Models: kimi-k2.5, kimi-k2-instant

Best for: Long-context tasks, Chinese language support.

Venice

Privacy-focused AI provider with uncensored model access.

Setup:

Get an API key from Venice Settings

export VENICE_API_KEY="..."

Available Models: venice-default, venice-fast

Best for: Privacy-focused deployments, uncensored research.

Cloudflare AI Gateway

Run models at the edge through Cloudflare's AI Gateway.

Setup:

Get an API token from Cloudflare Dashboard

export CLOUDFLARE_API_TOKEN="..."
export CLOUDFLARE_ACCOUNT_ID="..."

Available Models: Configurable -- any model available through the Cloudflare AI catalog (e.g., @cf/meta/llama-3.3-70b-instruct-fp8-fast, @cf/meta/llama-3.1-8b-instruct).

Best for: Edge deployments, global low-latency, Cloudflare ecosystem.

Xiaomi Mimo

Xiaomi's AI model platform.

Setup:

Get an API key from Xiaomi Developer Portal

export XIAOMI_API_KEY="..."

Available Models: mimo-v2-flash

Best for: IoT integration, Xiaomi ecosystem.

OpenRouter

Meta-provider that routes to the best available model across providers. Uses automatic model selection by default.

Setup:

Get an API key from OpenRouter

export OPENROUTER_API_KEY="..."

Available Models: auto (routes automatically), plus any model from the OpenRouter model list.

Best for: Multi-provider fallback, cost optimization, accessing many models through one API.

SmallModelResolver

The SmallModelResolver automatically selects the cheapest/fastest model for lightweight tasks like sentiment analysis, style profiling, and security auditing. It is used internally by:

LLMSentimentAnalyzer for mood/sentiment analysis
StyleAdaptationEngine for communication style profiling
WunderlandSecurityPipeline DualLLM auditor for security checks

Usage

import { SmallModelResolver } from 'wunderland/advanced';

const resolver = new SmallModelResolver({
  primaryProvider: 'openai',
});

// Get the small/fast model
const small = resolver.resolveSmall();
// => { providerId: 'openai', modelId: 'gpt-4o-mini' }

// Get the default/powerful model
const primary = resolver.resolveDefault();
// => { providerId: 'openai', modelId: 'gpt-4o' }

Override and Fallback

const resolver = new SmallModelResolver({
  primaryProvider: 'anthropic',

  // Override the small model (instead of using the built-in mapping)
  smallModelOverride: 'claude-haiku-4-5-20251001',

  // Fallback provider if primary is unavailable
  fallbackProvider: 'ollama',
  fallbackSmallModelOverride: 'llama3.2:3b',
});

Small Model Mapping

The complete built-in mapping from provider to small model:

Provider	Small Model
`openai`	`gpt-4o-mini`
`anthropic`	`claude-haiku-4-5-20251001`
`ollama`	`llama3.2:3b`
`openrouter`	`auto`
`bedrock`	`anthropic.claude-haiku`
`gemini`	`gemini-2.0-flash-lite`
`github-copilot`	`gpt-4o-mini`
`minimax`	`MiniMax-VL-01`
`qwen`	`qwen-turbo`
`moonshot`	`kimi-k2-instant`
`venice`	`venice-fast`
`cloudflare-ai`	`@cf/meta/llama-3.1-8b-instruct`
`xiaomi-mimo`	`mimo-v2-flash`

Checking Provider Support

// Check if a provider is in the known mapping
SmallModelResolver.isKnownProvider('openai'); // true
SmallModelResolver.isKnownProvider('custom'); // false

// Get the full mapping
const map = SmallModelResolver.getSmallModelMap();

CLI Commands

# List all providers and their available models
wunderland models

# Set default provider and model
wunderland models set-default openai gpt-4o

# Test provider connectivity
wunderland models test ollama
wunderland models test openai

Inference Routing -- how models are selected per request
Ollama Local Setup -- running models locally
LLM Sentiment Analysis -- uses SmallModelResolver
Style Adaptation -- uses SmallModelResolver

Provider Overview​

Provider Details​

OpenAI​

OpenAI OAuth (Subscription Login)​

Anthropic​

Ollama (Local)​

AWS Bedrock​

Google Gemini​

GitHub Copilot​

Minimax​

Qwen​

Moonshot​

Venice​

Cloudflare AI Gateway​

Xiaomi Mimo​

OpenRouter​

SmallModelResolver​

Usage​

Override and Fallback​

Small Model Mapping​

Checking Provider Support​

CLI Commands​

Related​

Provider Overview

Provider Details

OpenAI

OpenAI OAuth (Subscription Login)

Anthropic

Ollama (Local)

AWS Bedrock

Google Gemini

GitHub Copilot

Minimax

Qwen

Moonshot

Venice

Cloudflare AI Gateway

Xiaomi Mimo

OpenRouter

SmallModelResolver

Usage

Override and Fallback

Small Model Mapping

Checking Provider Support

CLI Commands

Related