🤖 Model Providers

📦 Supported Providers

🔍 Auto-Detection Priority

DevDuck checks for credentials in this order:

1. BedrockAWS_BEARER_TOKEN_BEDROCK or AWS STS credentials
2. AnthropicANTHROPIC_API_KEY
3. OpenAIOPENAI_API_KEY
4. GitHubGITHUB_TOKEN or PAT_TOKEN
5. GeminiGOOGLE_API_KEY or GEMINI_API_KEY
6. CohereCOHERE_API_KEY
7. WriterWRITER_API_KEY
8. MistralMISTRAL_API_KEY
9. LiteLLMLITELLM_API_KEY
10. LlamaAPILLAMAAPI_API_KEY
11. SageMakerSAGEMAKER_ENDPOINT_NAME
12. LlamaCppLLAMACPP_MODEL_PATH
13. MLXApple Silicon detected + strands_mlx installed
14. OllamaFallback (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

🔀 Multi-Model with use_agent

Use different models for different tasks within the same session:

# Use Bedrock for main agent, but OpenAI for a specific task
use_agent(
    prompt="Write a haiku about coding",
    system_prompt="You are a poet",
    model_provider="openai",
    model_settings={"model_id": "gpt-4o"}
)

# Use Ollama for local processing
use_agent(
    prompt="Summarize this text",
    system_prompt="You summarize text concisely",
    model_provider="ollama",
    model_settings={"model_id": "qwen3:8b"}
)

# Use environment config
use_agent(
    prompt="Analyze data",
    system_prompt="You are a data analyst",
    model_provider="env"  # Uses STRANDS_* env vars
)

🦙 Ollama Smart Defaults

Ollama model auto-selected based on OS: