Expand description
Public Rust facade for DAG-ML.
Modules§
- aggregation
- bundle
- campaign
- controller
- controller_
registry - Loader for declarative
*.controller.yamlregistries. - data
- dsl
- error
- fold
- generation
- graph
- ids
- metrics
- observability
- ADR-12 observability hooks.
- oof
- phase
- plan
- policy
- provenance
- relation
- rng
- runtime
- selection
Structs§
- Aggregate
Representation - Aggregated
Prediction Block - Aggregation
Controller Result - Aggregation
Controller Spec - Aggregation
Controller Task - Aggregation
Policy - Artifact
Handle Record - Artifact
Id - Artifact
Input Spec - Artifact
Materialization Request - Artifact
Payload Materialization Record - Artifact
Ref - Augmentation
Policy - Branch
Id - Branch
View Plan - Bundle
Data Requirement - Bundle
Id - Bundle
Prediction Block Cache Record - Bundle
Prediction Cache Payload - Bundle
Prediction Cache Payload Set - Bundle
Prediction Cache Record - Bundle
Prediction Requirement - Bundle
Replay Execution - Campaign
Fingerprint Spec - Campaign
Spec - Candidate
Score - Cartesian
Product Representation - Columnar
Prediction Cache Block - Columnar
Prediction Cache Manifest - Columnar
Prediction Cache Store - Combination
Plan - Compiled
Pipeline Dsl - Controller
Id - Controller
Manifest - Controller
Registry - DagMl
Error Descriptor - A stable ADR-11 error payload that can be serialized across bindings.
- Data
Binding - Data
Handle Record - Data
Materialization Request - Data
Model Shape Plan - Data
Output Provenance - Data
Plan - Data
Plan Step - Data
Provider View Spec - Data
View Handle Record - Data
View Policy - Data
View Request - Data
View Selector - Edge
Contract - Edge
Spec - Evaluation
Result - Execution
Bundle - Execution
Plan - Execution
Scope Plan - Explanation
Block - An EXPLAIN-phase output block (ADR-12 explain contract). Explanations are a
node output returned in the
NodeResult— like predictions, they cross as data, not as an opaque host handle. Thepayloadshape is controller-defined (e.g. per-feature importances); the core does not interpret it. Explanations are only valid in theEXPLAINphase. - External
Data Plan Envelope - Feature
Selection Policy - File
Artifact Manifest - Versioned, file-backed artifact manifest. This is a manifest/portability
layer only: it records portable
ArtifactRefmetadata for a bundle’s refit artifacts. It does not deserialize ML objects or materialize artifact payloads; payload stores remain future work. - File
Artifact Manifest Entry - One persisted artifact entry. Mirrors the bundle
RefitArtifactRecordidentity (node, controller, artifact and params fingerprint) while requiring theArtifactRefto be portable so the manifest stays movable with its payloads. - File
Artifact Manifest Store - File-backed artifact manifest store rooted at a directory.
- File
Artifact Payload Store - File
Prediction Cache Entry - File
Prediction Cache Manifest - File
Prediction Cache Store - FitInfluence
Diagnostic - FitInfluence
Task - Fold
Assignment - FoldId
- FoldSet
- Generation
Choice - Generation
Dimension - Generation
Param Override - Generation
Spec - Graph
Interface - Graph
Plan - Graph
Spec - GroupId
- GroupK
Fold Spec - Handle
Ref - InMemory
Aggregated Prediction Store - InMemory
Artifact Store - InMemory
Data Provider - InMemory
Lineage Recorder - InMemory
Prediction Cache Store - InMemory
Prediction Store - KFold
Spec - Leakage
Unit Policy - Lineage
Id - Lineage
Record - Model
Input Fusion Policy - Model
Input Port Spec - Model
Input Spec - Monte
Carlo Cartesian Representation - NodeId
- Node
Plan - Node
Result - Node
Spec - Node
Task - Observation
Id - Observation
Prediction Block - OofCampaign
- OofMatrix
- Open
Lineage RunEvent Options - Operator
Selector - Parallel
Scheduler - Phase
Execution Schedule - Pipeline
DslBranch - Pipeline
DslBranch Step - Pipeline
DslConcat Branch - Pipeline
DslConcat Transform Step - Pipeline
DslData Port - Pipeline
DslGeneration Choice - Pipeline
DslGeneration Dimension - Pipeline
DslGeneration Param Override - Pipeline
DslGenerator Stage - Pipeline
DslGenerator Step - Pipeline
DslMerge Model Step - Pipeline
DslMerge Selector - Pipeline
DslMerge Step - Pipeline
DslOperator Step - Pipeline
DslPrediction Port - Pipeline
DslSequence Step - Pipeline
DslShape Plan - Pipeline
DslSpec - Pipeline
DslTuning Spec - Pipeline
DslVariant Choice - PortRef
- Port
Schema - Port
Spec - Prediction
Block - Prediction
Cache Materialization Record - Prediction
Cache Materialization Request - Prediction
Input Spec - Prediction
Join Policy - Ranked
Candidate - Reduction
Plan - Refit
Artifact Record - Refit
Slot Plan - Regression
Metric Report - Regression
Target Block - Relation
Contract - Replay
Phase Request - Representation
Combo Selection Record - Representation
Compatibility Report - Representation
Replay Manifest - Representation
Sample Observation Mapping - Research
Provenance Export - Research
Provenance Package - Research
Provenance Package File - Research
Provenance Package Validation - RunContext
- RunId
- Runtime
Controller Registry - Sample
Id - Sample
Relation - Sample
Relation Set - Schema
Migration Policy - Seed
Context - Selection
Decision - Selection
Metric - Selection
Policy - Sequential
Scheduler - Shape
Delta - Split
Invocation - Stack
Fixed Representation - Stack
Padded Masked Representation - Stacking
FitContract - StratifiedK
Fold Spec - Stratified K-fold: each sample is validated exactly once (OOF-safe like
plain K-fold), but folds are balanced by a per-sample class label so every
fold mirrors the overall class distribution.
stratamaps each sample id to its class label (identity-keyed metadata — never feature values). - Target
Id - Variant
Execution Spec - Variant
Id - Variant
Plan
Enums§
- Aggregation
Controller Input - Aggregation
Controller Output - Aggregation
Method - Aggregation
Weights - Artifact
Backend - Artifact
Policy - Augmentation
Scope - Branch
View Mode - Combination
Mode - Controller
Capability - Controller
FitScope - DagMl
Error - Data
Plan Step Kind - Data
Request Partition - Entity
Unit Level - Evaluation
Scope - Feature
Selection Scope - FitBoundary
- FitInfluence
Mechanism - FitInfluence
Policy - Fold
Partition - Generation
Strategy - Granularity
- Handle
Kind - Inference
Features - Meta
RowDomain - Meta
Training Features - Metric
Objective - Missingness
Policy - Model
Input Fusion Mode - Nested
CvSpec - Inner (nested) cross-validation policy.
- Node
Kind - Phase
- Pipeline
DslBranch Mode - Pipeline
DslGenerator Mode - Pipeline
DslGenerator Value - Pipeline
DslMerge Output - Pipeline
DslParam Generator - Pipeline
DslSelection Spec - Pipeline
DslStep - Port
Cardinality - Port
Kind - Prediction
Join Key - Prediction
Level - Prediction
Partition - Prediction
Unit Id - Reduction
Axis - Reduction
Method - Reduction
Role - Reduction
Task Compatibility - Refit
Strategy - Regression
Metric Kind - Representation
Cardinality - Representation
Compatibility Outcome - Representation
Compatibility Severity - Representation
Missing Source Policy - Representation
Plan - RngPolicy
- Selection
Protocol - Shape
Delta Kind - Split
Unit
Constants§
- AGGREGATION_
CONTROLLER_ RESULT_ SCHEMA_ ID - AGGREGATION_
CONTROLLER_ RESULT_ SCHEMA_ VERSION - AGGREGATION_
CONTROLLER_ TASK_ SCHEMA_ ID - AGGREGATION_
CONTROLLER_ TASK_ SCHEMA_ VERSION - BUNDLE_
PREDICTION_ CACHE_ FORMAT - CAMPAIGN_
SPEC_ SCHEMA_ ID - CAMPAIGN_
SPEC_ SCHEMA_ VERSION - CONTROLLER_
MANIFEST_ SCHEMA_ ID - CONTROLLER_
MANIFEST_ SCHEMA_ VERSION - DAGML_
OPENLINEAGE_ FACET_ SCHEMA_ URL - DATA_
OUTPUT_ PROVENANCE_ KEY - DATA_
OUTPUT_ PROVENANCE_ SCHEMA_ ID - DATA_
OUTPUT_ PROVENANCE_ SCHEMA_ VERSION - DATA_
PLAN_ SCHEMA_ ID - DATA_
PLAN_ SCHEMA_ VERSION - EXECUTION_
BUNDLE_ FILE - EXECUTION_
BUNDLE_ SCHEMA_ VERSION - EXECUTION_
PLAN_ FILE - EXECUTION_
PLAN_ SCHEMA_ ID - EXECUTION_
PLAN_ SCHEMA_ VERSION - EXTERNAL_
DATA_ PLAN_ ENVELOPE_ SCHEMA_ VERSION - FILE_
ARTIFACT_ MANIFEST_ FILE - FILE_
ARTIFACT_ MANIFEST_ SCHEMA_ VERSION - FILE_
PREDICTION_ CACHE_ MANIFEST_ FILE - FILE_
PREDICTION_ CACHE_ STORE_ SCHEMA_ VERSION - GRAPH_
SPEC_ SCHEMA_ ID - GRAPH_
SPEC_ SCHEMA_ VERSION - LINEAGE_
RECORDS_ FILE - MIN_
READABLE_ EXECUTION_ BUNDLE_ SCHEMA_ VERSION - MIN_
READABLE_ PREDICTION_ CACHE_ PAYLOAD_ SCHEMA_ VERSION - MIN_
WRITABLE_ EXECUTION_ BUNDLE_ SCHEMA_ VERSION - MIN_
WRITABLE_ PREDICTION_ CACHE_ PAYLOAD_ SCHEMA_ VERSION - MODEL_
INPUT_ SPEC_ SCHEMA_ ID - MODEL_
INPUT_ SPEC_ SCHEMA_ VERSION - NODE_
RESULT_ SCHEMA_ ID - NODE_
RESULT_ SCHEMA_ VERSION - NODE_
TASK_ SCHEMA_ ID - NODE_
TASK_ SCHEMA_ VERSION - OBSERVABILITY_
FIELD_ ALLOWLIST - Frozen ADR-12 telemetry field allowlist. Every field emitted by this module must appear here, and each entry is an identifier or a count — never data. Adding a field requires an ADR-12 update and a review per the privacy rule.
- OPENLINEAGE_
RUN_ EVENT_ SCHEMA_ URL - PIPELINE_
DSL_ SCHEMA_ ID - PIPELINE_
DSL_ SCHEMA_ VERSION - PREDICTION_
CACHE_ PAYLOAD_ SCHEMA_ VERSION - PROV_
JSONLD_ FILE - RESEARCH_
PROVENANCE_ SCHEMA_ VERSION - RO_
CRATE_ METADATA_ FILE - SELECTION_
DECISION_ SCHEMA_ ID - SELECTION_
DECISION_ SCHEMA_ VERSION - SELECTION_
POLICY_ SCHEMA_ ID - SELECTION_
POLICY_ SCHEMA_ VERSION
Traits§
Functions§
- aggregate_
observation_ predictions - aggregate_
sample_ predictions_ by_ unit - build_
aggregated_ prediction_ cache_ payload - build_
aggregated_ prediction_ cache_ record - build_
execution_ bundle - build_
execution_ bundle_ with_ prediction_ contracts - build_
execution_ bundle_ with_ prediction_ requirements - build_
execution_ plan - build_
openlineage_ run_ event - build_
openlineage_ run_ event_ from_ package_ files - build_
prediction_ cache_ payload - build_
prediction_ cache_ record - build_
research_ provenance_ export - build_
research_ provenance_ package - bundle_
prediction_ requirement_ key - campaign_
fingerprint - capabilities_
support_ fit_ influence - compile_
pipeline_ dsl - compile_
pipeline_ dsl_ with_ controller_ registry - compile_
pipeline_ dsl_ with_ generation - compile_
pipeline_ dsl_ with_ generation_ and_ controller_ registry - dispatch_
custom_ aggregation_ task - dispatch_
custom_ observation_ aggregation - dispatch_
custom_ sample_ aggregation - emit_
oof_ refusal - Emit the ADR-12 out-of-fold leakage refusal event with stable taxonomy fields
(
category/codemirrorcrate::DagMlError::OofLeakage) so log consumers can alert on refusals without parsing messages. - enumerate_
variants - execution_
bundle_ schema_ migration_ policy - fold_
set_ fingerprint - generation_
spec_ fingerprint - join_
oof_ campaign_ features - join_
oof_ features - load_
yaml_ manifest_ from_ path - Load and validate a single manifest from disk.
- load_
yaml_ manifests_ from_ dir - Walk a directory for
*.controller.yamlfiles and return a deterministically ordered, validated manifest list. Duplicatecontroller_ids across files are rejected so a directory cannot silently ship two definitions of the same controller. - lower_
nirs4all_ compat_ pipeline_ dsl - node_
span - Build the per-node span (ADR-12), nested under the current phase span so node telemetry is attributed to its run, plan and controller. Identifiers only.
- oof_
campaign_ fingerprint - parse_
pipeline_ dsl_ json - parse_
yaml_ manifest - Parse a single YAML manifest from raw text. The same
ControllerManifestdeserializer is reused so YAML and JSON cannot drift on field shape. - phase_
span - Build the per-phase-scope span (ADR-12).
run_id/plan_idcorrelate concurrent or overlapping runs; emptyvariant_id/fold_idmean the field is not applicable to the current phase. Fields are identifiers only. - prediction_
cache_ payload_ schema_ migration_ policy - refit_
artifact_ input_ key - regression_
report_ to_ candidate_ score - relation_
set_ fingerprint - resolve_
inner_ cv - Resolve the effective inner-CV policy for a node: a node-local policy
overrides the campaign-global default;
Nonemeans no nested CV. - resolve_
pipeline_ dsl_ minimal_ aliases - score_
regression_ aggregated_ block - score_
regression_ prediction_ block - select_
candidate - select_
candidate_ groups - validate_
data_ binding_ envelope - validate_
inner_ fold_ set_ within_ outer - Enforce the nested-CV invariant: every sample in
inner— both the top-level universe and every fold’s train/validation members — must be an outer-fold training sample (never an outer-validation sample). Holds by construction for dag-ml-built inner folds, and also validates inner folds supplied from elsewhere. Refuses with an OOF-validation error on any leaking sample. - validate_
oof_ campaign - validate_
prediction_ blocks_ against_ folds - validate_
prediction_ blocks_ are_ oof - validate_
prediction_ cache_ payload_ matches_ record - validate_
research_ provenance_ package_ files