API Reference synaptic v#0.2.7

Synaptic provides a declarative workflow engine with a DSL for orchestrating LLM-backed steps, human-in-the-loop pauses, and resumable executions.

OTP application entry point that starts the registry, runtime supervisor, PubSub, and Finch pool required by Synaptic.

Demo workflow that collects email/phone details with LLM-based routing.

A small workflow available only in the dev environment so you can try the Synaptic engine end-to-end from iex -S mix.

Internal orchestrator that glues workflow definitions to runtime runners.

Behaviour for integrating with 3rd party eval services like Braintrust, LangSmith, etc.

Experimental API for running workflows against multiple inputs and scorers.

Registry used to name runner processes by run id and enumerate active runs.

GenServer that executes a workflow definition, tracking context, waiting states, retries, history, PubSub events, and suspend/resume logic.

DynamicSupervisor responsible for spinning up Synaptic.Runner processes.

Behaviour and helper types for step-level scoring in Synaptic workflows.

Immutable snapshot passed to scorers when evaluating a step.

Standardized result returned by scorers.

Metadata structure for compiled workflow steps.

Main test runner module for executing Synaptic workflows from YAML test definitions.

Validates test expectations against workflow execution results. Supports regex matching for context field values.

Formats test results for console and JSON output.

Parses and validates YAML test definition files for Synaptic workflows.

Helper utilities for invoking LLM providers from workflow steps.

Behaviour for LLM adapter implementations.

Minimal OpenAI chat client built on Finch.

Struct describing an LLM-callable tool.

DSL entry point for defining Synaptic workflows.