Skip to content

Model Providers

DevDuck supports 14 model providers with smart auto-detection.

Auto-Detection Priority

DevDuck checks for credentials in this order and uses the first available:

# Provider Detection
1 Amazon Bedrock AWS_BEARER_TOKEN_BEDROCK or AWS STS credentials
2 Anthropic ANTHROPIC_API_KEY
3 OpenAI OPENAI_API_KEY
4 GitHub Models GITHUB_TOKEN or PAT_TOKEN
5 Google Gemini GOOGLE_API_KEY or GEMINI_API_KEY
6 Cohere COHERE_API_KEY
7 Writer WRITER_API_KEY
8 Mistral MISTRAL_API_KEY
9 LiteLLM LITELLM_API_KEY
10 LlamaAPI LLAMAAPI_API_KEY
11 SageMaker SAGEMAKER_ENDPOINT_NAME
12 LlamaCpp LLAMACPP_MODEL_PATH
13 MLX Apple Silicon + strands_mlx installed
14 Ollama Fallback (always available)

Manual Selection

# Force specific provider
export MODEL_PROVIDER=anthropic
devduck

# With specific model
export MODEL_PROVIDER=bedrock
export STRANDS_MODEL_ID=us.anthropic.claude-sonnet-4-20250514-v1:0
devduck

# Common parameters
export STRANDS_MAX_TOKENS=60000
export STRANDS_TEMPERATURE=1.0

Provider Configuration

# Option 1: AWS credentials (default profile)
aws configure

# Option 2: Bearer token
export AWS_BEARER_TOKEN_BEDROCK=...

# Option 3: Explicit
export MODEL_PROVIDER=bedrock
export STRANDS_MODEL_ID=us.anthropic.claude-sonnet-4-20250514-v1:0

export ANTHROPIC_API_KEY=sk-ant-...
# Optional:
export STRANDS_MODEL_ID=claude-sonnet-4-20250514

export OPENAI_API_KEY=sk-...
# Optional:
export STRANDS_MODEL_ID=gpt-4o

export GOOGLE_API_KEY=...
# Optional:
export STRANDS_MODEL_ID=gemini-2.5-flash

# Just have Ollama running
ollama serve

# Optional:
export OLLAMA_HOST=http://localhost:11434
export STRANDS_MODEL_ID=qwen3:1.7b  # macOS default

pip install strands-mlx
export MODEL_PROVIDER=mlx
export STRANDS_MODEL_ID=mlx-community/Qwen3-1.7B-4bit

Multi-Model with use_agent

Use different models for different tasks within the same session:

# Use Anthropic for creative writing
use_agent(
    prompt="Write a haiku about artificial intelligence",
    system_prompt="You are a minimalist poet.",
    model_provider="anthropic"
)

# Use local Ollama for sensitive data
use_agent(
    prompt="Summarize this confidential document",
    system_prompt="You summarize documents concisely.",
    model_provider="ollama",
    model_settings={"model_id": "qwen3:8b"}
)

# Use environment configuration
use_agent(
    prompt="Analyze this data",
    system_prompt="You are a data analyst.",
    model_provider="env"
)

→ See Multi-Agent for more patterns.

Ollama Smart Defaults

DevDuck picks sensible Ollama models per platform:

OS Default Model
macOS qwen3:1.7b
Linux qwen3:30b
Windows qwen3:8b

Override with STRANDS_MODEL_ID.