Every Action Is a Decision
Decision-Model-Driven Execution for Structured Agent Workflows
DecisionAct is a Python agent framework for building task-focused agents with predefined intents, tools, and outcomes (with reply templates). Decision models interpret user requests and select actions; your Python functions execute them.
Using pip:
pip install decisionactUsing uv inside a project:
uv add decisionactCurrent source and documentation: v0.1.2. Earlier releases: v0.1.1 and v0.1.0.
See the changelog for detailed release notes and upgrade guidance.
v0.1.2 adds fresh inputs for repeated tool invocations, earlier-result input choices, structured decision state, one input-confidence recheck and renewable call allowances with a hard cap. See the calculator example and upgrade notes.
Read the documentation for the quickstart, agent-building guides, execution flow, and API reference.
MIT - AnilKumarBorige