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.