`agentic_fabric.config.llm`¶

LLM configuration shared by all runners.

Uses Anthropic Claude directly via ANTHROPIC_API_KEY. Falls back to OpenRouter if OPENROUTER_API_KEY is set. Supports local Ollama via OLLAMA_BASE_URL / AGENTIC_FABRIC_LLM_PROVIDER=ollama.

Module Contents¶

Classes¶

`LLMProvider`	Available LLM providers.
`LLMConfig`	Configuration for specific LLM use cases.

Functions¶

`get_llm`	Get configured LLM instance for agents.
`get_llm_or_raise`	Get configured LLM instance, raising if API key not set.
`get_llm_for_task`	Get LLM configured for a specific task type.
`get_reasoning_llm`	Get LLM optimized for complex reasoning tasks.
`get_creative_llm`	Get LLM optimized for creative tasks.
`get_code_llm`	Get LLM optimized for code generation.

Data¶

`DEFAULT_MODEL`
`MODELS`
`LLM_CONFIGS`

API¶

class agentic_fabric.config.llm.LLMProvider(*args, **kwds)¶

Bases: enum.Enum

Available LLM providers.

Initialization

ANTHROPIC = 'anthropic'¶

OPENROUTER = 'openrouter'¶

OLLAMA = 'ollama'¶

name()¶

value()¶

class agentic_fabric.config.llm.LLMConfig¶

Configuration for specific LLM use cases.

model: str = None¶

temperature: float = None¶

description: str = None¶

agentic_fabric.config.llm.DEFAULT_MODEL = None¶

agentic_fabric.config.llm.MODELS = None¶

agentic_fabric.config.llm.LLM_CONFIGS = None¶

agentic_fabric.config.llm.get_llm(model: str = DEFAULT_MODEL, temperature: float = 0.7, provider: agentic_fabric.config.llm.LLMProvider | None = None) → crewai.LLM | None¶

Get configured LLM instance for agents.

Args: model: Model identifier. Defaults to claude-haiku-4-5-20251001 (DEFAULT_MODEL) temperature: Sampling temperature (0.0-1.0). Lower = more focused, higher = more creative. provider: Force specific provider (ANTHROPIC, OPENROUTER, or OLLAMA). If None, auto-detects based on available API keys and env vars.

Available models: - claude-haiku-4-5-20251001 (default - fast, cost-effective) - claude-sonnet-4-5-20250929 (best for code and creative) - claude-sonnet-4-20250514 (capable general purpose) - claude-opus-4-20250514 (most capable) - openrouter/auto (fallback via OpenRouter) - qwen2.5:0.5b (local Ollama, no API key needed)

Returns: Configured LLM instance, or None if no API key set

Note: Tries ANTHROPIC_API_KEY first, falls back to OPENROUTER_API_KEY. When AGENTIC_FABRIC_LLM_PROVIDER=ollama or OLLAMA_BASE_URL is set, routes to a local Ollama server (no API key needed). Returns None if no provider is configured.

Example: >>> llm = get_llm() # Uses Claude Haiku 4.5 >>> llm = get_llm(“claude-opus-4-20250514”, temperature=0.3) >>> llm = get_llm(provider=LLMProvider.OPENROUTER) >>> llm = get_llm(provider=LLMProvider.OLLAMA) # Local Ollama

agentic_fabric.config.llm.get_llm_or_raise(model: str = DEFAULT_MODEL, temperature: float = 0.7, provider: agentic_fabric.config.llm.LLMProvider | None = None) → crewai.LLM¶

Get configured LLM instance, raising if API key not set.

Use this when you need to ensure an LLM is available.

Args: model: Model identifier (see get_llm for options) temperature: Sampling temperature (0.0-1.0) provider: Force specific provider (optional)

Returns: Configured LLM instance

Raises: ValueError: If neither ANTHROPIC_API_KEY nor OPENROUTER_API_KEY is set, and Ollama mode is not enabled.

agentic_fabric.config.llm.get_llm_for_task(task: str) → crewai.LLM | None¶

Get LLM configured for a specific task type.

Args: task: Task type - one of: ‘reasoning’, ‘creative’, ‘code’, ‘default’