Per-run dispatch metadata threaded through the engine and executor. More...

#include <run_context.h>

Collaboration diagram for neograph::graph::RunContext:

Public Attributes
std::shared_ptr< CancelToken >	budget_cancel_token
	Shared sibling-cancellation scope for budget-aware nodes.

std::uint64_t	cache_execution_id = 0
	Engine-assigned identity for this run or resume call.

std::shared_ptr< CancelToken >	cancel_token
	Operation-scoped cooperative cancellation handle.

std::optional< std::chrono::steady_clock::time_point >	deadline
	Absolute monotonic-clock deadline supplied through RunMetadata, when set.

asio::any_io_executor	dispatch_executor
	RunMetadata::dispatch_executor: where a node attempt that needs its own worker runs.

std::shared_ptr< OwnedManagedBudgetLease >	managed_budget_lease
	Host-only standalone currency; never recovered from caller JSON.

std::uint64_t	model_token_budget = 0
	Program-supplied token ceiling and terminal signal. Zero/null leave generic Core runs unchanged.

std::string	resume_interrupt_id
	#406: the interrupt resume_value answers, when it answers this invocation's own NodeInterrupt: its PendingInterrupt::id, the task id that RunResult::interrupt_id and RunResult::interrupts report, whether the resume named it (RunMetadata::resume_interrupt_id) or answered the first one.

std::optional< json >	resume_value
	Value supplied to GraphEngine::resume(), for the one invocation the resume answers, in the step it re-enters (#388, #393): the node or Send whose NodeInterrupt it answers, or the node an interrupt_before gate held.

std::string	run_id
	Durable Program run identity used by host-brokered journal effects.

bool	shared_scheduler = false
	RunMetadata::shared_scheduler: blocking node bodies should not hold the calling executor, which other runs and host timers share.

int	step = 0
	Current super-step index.

std::shared_ptr< Store >	store
	Optional cross-thread shared memory.

StreamMode	stream_mode = StreamMode::ALL
	Mirrors RunConfig::stream_mode.

std::string	thread_id
	Mirrors RunConfig::thread_id.

std::shared_ptr< ToolExecutionController >	tool_execution_controller
	Shared host admission boundary for every tool call in this run.

ToolExecutionIdentity	tool_execution_identity
	Stable host/Program identity copied to each dispatched tool call.

ToolGate	tool_gate
	Effective tool policy for this invocation.

std::string	trace_id
	Per-run trace correlator supplied through RunMetadata.

std::shared_ptr< UsageAccumulator >	usage
	Shared token accounting sink for this run and its subgraphs.

Detailed Description

Per-run dispatch metadata threaded through the engine and executor.

GraphNode::run(NodeInput) consumes this through NodeInput::ctx. Workers that need an isolated copy take it by value; the common path takes it by const reference.

Definition at line 71 of file run_context.h.

Member Data Documentation

◆ cache_execution_id

std::uint64_t neograph::graph::RunContext::cache_execution_id = 0

Engine-assigned identity for this run or resume call.

Used only for execution-local cache isolation; it is not derived from user data.

Definition at line 112 of file run_context.h.

◆ dispatch_executor

asio::any_io_executor neograph::graph::RunContext::dispatch_executor

RunMetadata::dispatch_executor: where a node attempt that needs its own worker runs.

Empty: the process-wide attempt pool.

Definition at line 102 of file run_context.h.

◆ resume_interrupt_id

std::string neograph::graph::RunContext::resume_interrupt_id

#406: the interrupt resume_value answers, when it answers this invocation's own NodeInterrupt: its PendingInterrupt::id, the task id that RunResult::interrupt_id and RunResult::interrupts report, whether the resume named it (RunMetadata::resume_interrupt_id) or answered the first one.

Empty when resume_value is the approval of the interrupt_before gate that held this node (#393, #396), and whenever resume_value is empty. A node behind a gate that also asks its own question tells the gate's approval from the answer to its question by this field.

Definition at line 137 of file run_context.h.

◆ resume_value

std::optional<json> neograph::graph::RunContext::resume_value

Value supplied to GraphEngine::resume(), for the one invocation the resume answers, in the step it re-enters (#388, #393): the node or Send whose NodeInterrupt it answers, or the node an interrupt_before gate held.

Every other invocation sees it empty, as does every invocation of a later step and of a fresh run, so a node that asks pauses the run again instead of taking an answer to another question.

Definition at line 126 of file run_context.h.

◆ tool_gate

ToolGate neograph::graph::RunContext::tool_gate

Effective tool policy for this invocation.

It includes any inherited parent policy and this engine's own policy, so a subgraph cannot weaken a parent gate by using a different GraphEngine instance.

Definition at line 145 of file run_context.h.

The documentation for this struct was generated from the following file:

include/neograph/graph/run_context.h