DSL Executor SDK

The DSL Executor SDK is a Python library for loading and executing declarative, graph-based DSL workflows. It enables both local and remote execution of modular Python tasks defined as nodes within a Directed Acyclic Graph (DAG), while also supporting advanced routing and orchestration.

This SDK is useful for systems that need to define, version, and execute workflows composed of independently packaged Python modules.

---

Installation

Prerequisites

• Python 3.7 or above
• pip installed

From source (local setup)

Clone your project and install the SDK from the root directory:

cd systems/dsl/dsl_executor
pip install .

For development use (with test dependencies):

pip install -e .[dev]

---

Usage Example

Assume you have already created and uploaded a DSL workflow using the DSL Creator Server, and it is now registered in your workflow database.

Step 1: Load and Execute the Workflow

from dsl_executor import new_dsl_workflow_executor, parse_dsl_output

# Set required environment variable or pass the full base URI
WORKFLOWS_API_URL = "http://your-workflow-registry-service"

# Initialize executor from workflow ID
executor = new_dsl_workflow_executor(
    workflow_id="sample_workflow",
    workflows_base_uri=WORKFLOWS_API_URL,
    is_remote=False,          # Set to True for remote execution
    addons={"incr": 2}        # Optional override of globalSettings
)

# Define input to the DSL system
input_data = {
    "user_input": {"x": 10}
}

# Execute the workflow
output = executor.execute(input_data)

# Parse final output or specific module output
print("Final Output:", parse_dsl_output(output))
print("Module 2 Output:", parse_dsl_output(output, module_name="module2"))

---

Module Breakdown

DSLWorkflowExecutor (Abstract Base Class)

Defines the base interface for all executors:

• execute(input_data): Executes the workflow given the input.
• estimate(): Placeholder for estimating execution cost/resources.
• load_modules(): Optional hook for module setup.

---

DSLWorkflowExecutorLocal

Concrete implementation of DSLWorkflowExecutor for local execution.

• Performs topological sorting on the workflow graph

• Uses LocalCodeExecutor to:

• Download or copy module code from codePath

• Install Python dependencies (from requirements.txt)
• Load and invoke eval() on each node

---

DSLWorkflowExecutorRemote

Remote execution wrapper. Delegates execution to a remote DSL runtime service using RemoteDSLExecutor.

• Requires POLICIES_SYSTEM_URL environment variable
• Used for offloading execution to another system or cluster

---

new_dsl_workflow_executor(...)

Factory function to create either a DSLWorkflowExecutorLocal or DSLWorkflowExecutorRemote instance based on the is_remote flag.

Arguments:

Parameter	Description
workflow_id	ID of the registered workflow
workflows_base_uri	URL of the workflow registry
is_remote	If True, creates a remote executor
addons	Dict merged into globalSettings
executor_id	Optional unique ID for the remote executor

---

parse_dsl_output(output, module_name="")

Helper to extract either the final output or a specific module's output from the full DSL result.

• If module_name is omitted, returns the final sink node output.
• If provided, returns the output for that specific module.

---