DAGent - Directed Acyclic Graphs (DAGs) as AI Agents
DAGent is a Python library designed to streamline the creation of AI agents by structuring them into workflows using directed acyclic graphs (DAGs).
Key Features:
- Tool Abstraction: Functions are treated as tools that LLMs can utilize, with docstrings and annotations aiding in LLM inference.
- Function Nodes: Executes Python functions, which can be attached to Decision Nodes.
- Decision Nodes: LLMs select functions to run from available options. Tool descriptions are autogenerated and stored in
Tool_JSONfolder, which can be manually edited for better agent reliability. - Model Flexibility: Supports different LLM models for inference and tool description generation, including Groq and Ollama.
- Data Passing: Facilitates data transfer between functions using the
prev_outputparameter.
Use Cases:
- Automated task execution using AI agents.
- Building complex workflows with LLM-driven decision-making.
- Integrating existing Python code into AI agent frameworks.
