api-reference.md

API Reference

Core Layer

core.graph.WorkflowGraph

The central DAG data structure.

Method	Description
add_node(node)	Add a node to the graph
remove_node(node_id)	Remove a node and its edges
get_node(node_id)	Get a node by ID
add_edge(src, tgt)	Connect two nodes (raises CycleError if cycle)
remove_edge(src, tgt)	Disconnect two nodes
topological_order()	Return nodes in execution order
mark_dirty(node_id)	Mark node and all descendants dirty
get_upstream(node_id)	Get all upstream node IDs
get_downstream(node_id)	Get all downstream node IDs

core.engine.Engine

The DAG execution engine.

Method	Description
execute(graph, node_id)	Execute up to node_id, return DataFrame
execute_all(graph)	Execute all nodes in topological order
get_cached(node_id)	Get cached output for a node

core.node_registry.NODE_REGISTRY

Dictionary of NodeTypeSpec objects defining all available node types.

core.profiler.DataProfiler

Compute column statistics for a DataFrame.

core.formatter.ConditionalFormatEngine

View-layer conditional formatting engine.

Agent Layer

agent.schemas.PipelineMutationBatch

Pydantic model for graph mutations.

agent.schemas.AppCommandBatch

Pydantic model for app-wide command batches generated by the AI.

agent.interpreter.AgentInterpreter

Apply PipelineMutationBatch to a live graph.

agent.chat_session.ChatSession

Multi-turn AI chat session with context injection.

agent.ai_backend.AIBackendRouter

Routes to the configured Gemini backend.

State Layer

state.app_state.AppState

Central reactive state object with Qt signals.

state.workspace.Workspace

Multi-tab workspace management.

state.history.HistoryStack

Undo/redo with action-level granularity.

IPC Layer

ipc.protocol.IPCCommand / IPCResult

Typed IPC message classes for worker communication.

ipc.worker.ComputeWorker

Headless worker process for data computation.