| Dhenara Agent DSL (DAD), an open-source framework for building, running, and managing AI agents! GitHub: https://github.com/dhenara/dhenara-agent Documentation: https://docs.dhenara.com/dhenara-agent/introduction Key Features: - Agent Lifecycle Management: Create, run, resume, and manage agents - Run Controls: Resume execution from any point to save on API costs - Built-in Observability: OpenTelemetry tracing with Jaeger and Zipkin support - Event System: Fine-grained flow control - Open Development: Build using standard tools (Git, VS Code) - No Platform Lock-in: Build powerful command-line agents similar to Claude Code using just Git and VS Code Simple Architecture: DAD uses three main components: 1. Nodes: Basic execution units (API calls, file operations)
2. Flows: Collections of nodes with execution logic
3. Agents: High-level coordinators Sample: outer_loop_flow = FlowDefinition()
# Outer loop
outer_loop_flow.for_each(
id="category_processor",
statement=ObjectTemplate(expression="$expr{$hier{categories}.outcome.structured.categories}"),
item_var="category",
index_var="category_index",
max_iterations=10,
body=FlowDefinition().for_each(
# Inner loop
id="item_processor",
statement=ObjectTemplate(expression="$expr{category.items}"),
item_var="item",
index_var="item_index",
max_iterations=20,
body=item_processing_flow
)
)
While still in alpha, we've designed it to handle complex agent behaviors with robust control flow. We're Working On:- Budget control with pause/resume capabilities - Additional node types and integrations - Parallel execution for complex workflows - Enhanced visualization tools Try it out: `pip install dhenara-agent` We welcome your feedback on what would make this more useful for your AI agent projects! Thanks. |
How?