pllmLocal experiments
Serve declarative experiments, Python objects, and Python factories locally.
pllm gateway --local resolves one typed Experiment, starts its required roles as child processes, and exposes the trusted loopback Responses and Chat Completions API. The gateway does not replace the experiment's pipeline or profile.
Target forms
Use file.py:object for a Python file, module:object for an importable module, or .json/.yaml for declarative configuration. Python targets are imported code and therefore require --trust-python in unattended commands. Add --factory only when the selected object is a zero-argument callable returning an Experiment. Raw python file.py execution is intentionally unsupported.
The repository's composition example exposes both experiment and build_experiment; examples/pllm.yaml provides the equivalent declarative form.
Examples
Serve a Python experiment locally
Import the checked-in Experiment object and start both provider roles behind the trusted loopback gateway.
pllm gateway --local --experiment examples/composition.py:experiment --trust-pythonBuild an experiment with a Python factory
Use --factory when the selected Python object is a zero-argument callable rather than an existing Experiment.
pllm gateway --local --experiment examples/composition.py:build_experiment --factory --trust-pythonServe a declarative experiment
JSON and YAML targets are data, so they do not require --trust-python.
pllm gateway --local --experiment examples/pllm.yamlApplication endpoint
Applications connect to http://127.0.0.1:8080/v1 by default. Inference and preparation children remain provider-role services, not application-facing OpenAI-compatible endpoints. Stop the gateway to shut down its local role topology.