Skip to main content

LLM Module API Specification

Version: v1.1.9 | Status: Stable | Last Updated: February 2026

1. Overview​

The llm module manages Large Language Model integrations for Codomyrmex. It currently supports Ollama as a local inference backend and integrates with the Fabric framework for orchestration.

2. Core Components​

2.1 Ollama Integration​

  • OllamaManager: Manages the local Ollama instance process.
  • ModelRunner: Executes inference requests against Ollama models.
  • OutputManager: Handles streaming responses and formatting.
  • ConfigManager: Manages Ollama-specific configuration.

2.2 Fabric Integration​

  • FabricManager: Interface to the Fabric framework.
  • FabricOrchestrator: Coordinates multi-model workflows.
  • FabricConfigManager: Fabric configuration logic.

2.3 Configuration​

  • LLMConfig: Data structure for LLM settings.
  • get_config() -> LLMConfig: Retrieve current configuration.
  • set_config(config: LLMConfig): Update configuration.
  • reset_config(): Revert to defaults.

3. Usage Example​

from codomyrmex.llm import OllamaManager, ModelRunner

# Start Ollama
manager = OllamaManager()
manager.start()

# Run inference
runner = ModelRunner(model="llama3")
response = runner.generate("Explain quantum computing")
print(response)