Bridge SDK is a Python SDK for defining workflow pipelines with automatic dependency
management. Pipelines are defined using @pipeline.step decorators and step_result
annotations, then executed by the Bridge orchestrator on the Poolside platform.
make venv # Create virtual environment
make sync # Install dependencies
make proto # Generate protocol buffers
make test # Run testsPre-commit hooks are configured — run uv run pre-commit install after cloning.
bridge_sdk/pipeline.py— Pipeline class and PIPELINE_REGISTRYbridge_sdk/step.py—@stepdecorator (deprecated) andget_dsl_output()bridge_sdk/step_function.py— StepFunction wrapper and STEP_REGISTRYbridge_sdk/step_data.py— StepData Pydantic modelbridge_sdk/annotations.py—step_result()annotation helperbridge_sdk/function_schema.py— JSON schema generation from function signaturesbridge_sdk/cli.py— CLI commands:check,config get-dsl,runbridge_sdk/bridge_sidecar_client.py— gRPC client for agent integrationbridge_sdk/proto/bridge_sidecar.proto— gRPC service definitionexamples/— Example pipelines (agent_example.py, example.py)tests/— Test suite
This repo ships a bridge-pipelines skill at skills/bridge-pipelines/ that
teaches agents how to build Bridge pipelines using this SDK. It is distributed
via pkit as part of the bridge registry.
When making changes that affect how the SDK is used — such as changes to the
public API, decorators, annotations, CLI commands, agent integration, configuration
format, or usage patterns — you must also update the bridge-pipelines skill
in skills/bridge-pipelines/ so it stays in sync with the SDK.
The skill files:
skills/bridge-pipelines/SKILL.md— Core usage guide (quick start, concepts, examples)skills/bridge-pipelines/references/agents.md— Agent integration via BridgeSidecarClientskills/bridge-pipelines/references/cli.md— CLI command referenceskills/bridge-pipelines/references/web-integration.md— Web UI and REST API integration