Crate forge_engine 

forge-engine

Operational verification and evaluation engine for the local-first AI stack.

forge-engine owns the execution-heavy lane: compile mindstate, validate and apply structured patches, run checks, score candidate outcomes, persist operational evidence, and update causal edit-attribution state.

Authority boundary

• semantic-memory-forge owns raw verification/export wire truth.
• forge-memory-bridge owns transformation only.
• semantic-memory owns durable projected/queryable truth.
• forge-engine owns operational verification work on top of those crates.

The danger-sm-write feature remains a compatibility-only escape hatch and is not the canonical export/import path.

Re-exports

pub use config::ForgeConfig;

pub use error::ForgeError;

pub use error::ForgeResult;

pub use runtime::patch::apply_patch;

pub use runtime::patch::render_diff;

pub use runtime::patch::render_diff;

pub use runtime::patch::validate_patch;

pub use runtime::stabilizer::Stabilizer;

pub use runtime::compile_mindstate;

pub use runtime::extract_strategy_tags;

pub use runtime::DeltaPolicy;

pub use exec::is_env_allowed;

pub use exec::select_backend;

pub use exec::ContainerBackend;

pub use exec::ExecutionBackend;

pub use exec::HostBackend;

pub use adapters::CargoAdapter;

pub use adapters::ProjectAdapter;

pub use store::ForgeStore;

pub use config::ForgeLimits;

pub use lab::evaluate::compute_correctness;

pub use lab::evaluate::compute_scores;

pub use lab::evaluate::persist_eval_run;

pub use lab::archive_insert;

pub use lab::compute_cell_key;

pub use lab::compute_score_summary;

pub use lab::promote;

pub use lab::promote;

pub use lab::AlgebraSpec;

pub use lab::ArchiveUpdate;

pub use lab::BasisVersion;

pub use lab::CausalFingerprint;

pub use lab::CausalHypothesis;

pub use lab::EvalRunResult;

pub use lab::EvalSuite;

pub use lab::EvalTask;

pub use lab::ExperimentEvidenceBundle;

pub use lab::ExperimentRunner;

pub use lab::HypothesisStatus;

pub use lab::LocalExperimentRunner;

pub use lab::ParameterBounds;

pub use lab::ScoreVector;

pub use lab::VerificationPlan;

pub use lab::VerificationType;

pub use cea::attribute_effects;

pub use cea::build_edit_op_signature;

pub use cea::compute_run_hash;

pub use cea::load_graph;

pub use cea::load_graph_with_tx;

pub use cea::predict;

pub use cea::update_graph;

pub use cea::update_graph_with_tx;

pub use baseline::capture_baseline_provenance;

pub use baseline::BaselineDescriptor;

pub use baseline::BaselineSourceKind;

pub use baseline::ComparabilityPolicy;

pub use baseline::WorkspacePolicy;

pub use experiment::AnalysisRecord;

pub use experiment::CacheMode;

pub use experiment::EffectKind;

pub use experiment::EvidenceRecord;

pub use experiment::ExecutionRecord;

pub use experiment::ExperimentConfig;

pub use experiment::ExperimentDiff;

pub use experiment::ExperimentExportRecord;

pub use experiment::ExperimentMode;

pub use experiment::ExperimentResult;

pub use experiment::PairedExperimentRunner;

pub use experiment::RunIdentity;

pub use experiment::StatisticsPolicy;

pub use experiment::TrialRecord;

pub use experiment::TrialSide;

pub use experiment::TypedLocatedEffect;

pub use scoring::ComparabilityClass;

pub use scoring::ObjectiveKind;

pub use scoring::ObjectivePolicy;

pub use scoring::PatchExecutionPlan;

pub use scoring::PlannedCheck;

pub use export::compute_export_key;

pub use export::export_bundle;

pub use export::export_bundle_with_memory_write_through_compat;

pub use export::EpisodeExport;

pub use export::RENDERING_VERSION;

pub use tool_receipts::ForgeToolReceiptSink;

pub use failure::FailureClass;

pub use failure::FailureRecord;

pub use lab::evidence::build_hypothesis_edges;

pub use lab::evidence::compute_assessment;

pub use lab::evidence::generate_verification_plan;

pub use lab::evidence::local_hypothesis_support_confidence;

pub use lab::evidence::update_hypotheses_from_diff;

pub use lab::evidence::AssessmentCategory;

pub use lab::evidence::BundleScope;

pub use lab::evidence::ClaimStrength;

pub use lab::evidence::ContradictionState;

pub use lab::evidence::Covariates;

pub use lab::evidence::DroppedStep;

pub use lab::evidence::EvidenceAssessment;

pub use lab::evidence::HypothesisEdge;

pub use lab::evidence::HypothesisEdgeKind;

pub use lab::evidence::PairComparability;

pub use lab::evidence::PlanBudget;

pub use lab::evidence::ReceiptKind;

pub use lab::evidence::ReceiptRef;

pub use lab::evidence::ReceiptStorage;

pub use lab::evidence::SampleSupport;

pub use lab::evidence::StepRequirement;

pub use lab::evidence::Treatment;

pub use lab::evidence::VerificationPolicy;

pub use lab::evidence::VerificationState;

pub use lab::evidence::VerificationStep;

Modules

adapters

baseline

Baseline provenance and workspace policy for experiment comparability.

cea

config

error

exec

experiment

Experiment execution: baseline/patched paired trials with typed diffs.

export

Bundle export seam for the canonical Forge -> bridge -> memory lane.

failure

Failure classification, recording, and retry semantics.

invariants

lab

runtime

scoring

Objective-driven scoring policy.

store

tool_receipts

Structs

AttributedRunResult

AttributionTriple

CausalEdge

CausalGraph

CausalPrediction

CheckCommand

CheckResult

Claim

A claim extracted from source material.

CommandOutput

CommandTimings

CoverageSummary

EditOpSignature

EffectSignature

EvidenceBundle

An evidence bundle attached to a claim.

FileEdit

LedgerEntry

An entry in the claim ledger.

LineAttributionMap

LineRange

LocatedEffect

LogBundle

MindState

ParsedCheckOutput

PatchedWorkspace

RiskFlag

StructuredPatch

ValidationResult

Violation

Workspace

Enums

Anchor

AttemptPhase

CausalNode

CheckKind

ContainerRuntime

EditOp

ExecutionBackendKind

FileMode

LedgerEvent

Events that can be appended to the ledger.

UpdateResult

ViolationKind