A modular, configuration-driven framework for building and executing LangGraph-based agent workflows. Define agents, tools, and workflows entirely via YAML — no code changes required.
- Agentic Graph Execution using LangGraph
- Dynamic loading of agents, tools, and workflows from YAML configs
- CLI interface for running agents, invoking workflows, and visualizing graphs
- Extensible architecture: plug in new agents/tools without modifying core logic
| Component | Description |
|---|---|
| Tools | Stateless, atomic operations (e.g., fetch prices, run inference). Configured in configs/tools/, implemented in src/agentic/tools/. |
| Agents | Reasoning nodes with optional LLMs and tool access. Configured in configs/agents/, implemented in src/agentic/agents/. |
| Workflows | Directed graphs of agents defining execution flow. Configured in configs/graphs/, built via graph_builder.py. |
configs/ # YAML configs for agents, tools, workflows
src/
agentic/ # Core runtime: agents, tools, graph builder
cli/ # CLI commands (run, list, visualize)
tests/ # Unit tests
nelishka-cliRun unit and integration tests:
pytest tests/