The Omniscient Quick Start Guide provides a concise overview of the Omniscient framework, which is designed for decision-routing and execution. The framework emphasizes a requirements-driven approach, allowing users to specify what they need without dictating how those needs should be met. The core system intelligently selects the most suitable model, agent, or tool for execution, ensuring that all decisions made within the framework are transparent and testable.
To get started with Omniscient, users can install the package in development mode using pip install -e . and verify the installation with a simple Python command. The guide also outlines how to run tests, including architecture tests and a comprehensive suite of tests to ensure the framework’s integrity.
Basic usage examples illustrate how to utilize models, agents, and tools within the framework. For instance, the run_with_routing function automatically selects a model based on specified requirements, while the run_task function allows for automatic agent selection based on the task type. Additionally, users can execute registered tools with the execute_tool function.
The guide provides instructions for creating new components, such as agents, tools, and models. Each section includes code snippets demonstrating how to define a new agent or tool, register it, and integrate it into the core system. This modular approach allows for easy expansion and customization of the framework.
Key architectural rules are emphasized, such as the importance of routing logic being centralized in the core, execution logic residing in the app or agents, and the necessity of using registries for component selection. Common patterns for extending functionality without breaking existing code are also highlighted, along with debugging tips to inspect registered components and routing decisions.
The project structure is clearly laid out, showing the organization of core decision-making, execution, orchestration, capabilities, and external interfaces. Finally, the guide encourages users to explore example agents, run tests, and read additional documentation for further insights into expanding the framework. The architecture is described as locked, indicating that any deviations from the established structure should be carefully considered and refined.
Author: Jeremy Engram
I turn coffee into code...
