dag-ml-data Table Of Contents

Use this as a validation map before development starts.

Area

File

Purpose

Validate

Entry point

README.md

Project scope, layout and quick start

The repo can be understood in under five minutes

Agent handoff

AGENTS.md

Rules for autonomous implementation work

A new agent knows boundaries and green gate

Architecture

docs/ARCHITECTURE.md

Data-layer modules and DAG-ML frontier

No execution graph responsibility leaks in

ABI

docs/ABI.md

Data provider ABI and ownership

Host buffers stay host-owned

Shared contract schemas

docs/contracts/coordinator_data_plan_envelope.schema.json, docs/contracts/feature_fusion_selector.schema.json

JSON Schemas for the coordinator data-plan envelope and feature fusion selector produced for dag-ml

Fixtures and bindings declare the same v1 wire shapes

Rationale

docs/RATIONALE.md

Split from DAG-ML and design tradeoffs

Data scope is independently defensible

MVP acceptance

docs/MVP_ACCEPTANCE.md

First data-contract target for UC6/UC11

Data plans support stacking without owning OOF logic

Capability matrix

docs/CAPABILITY_MATRIX.md

Full nirs4all replacement data surface

Data contracts expose identity without owning OOF

OOF fixtures

docs/OOF_FIXTURES.md

Shared tiny campaign data contracts

DAG-ML can consume fixture identities without guessing

Roadmap

docs/ROADMAP.md

Sequenced delivery phases

Every phase has an observable definition of done

Status

docs/STATUS.md

Current scaffold state and next actions

No hidden implementation claims

Test plan

docs/TEST_PLAN.md

Schema, planner and ABI tests

Fingerprints and path solving are covered

Supported surface

docs/SUPPORTED.md

Release support matrix and public-signature policy

Production, conformance, experimental and backlog surfaces are separated

Aggregation interop

docs/AGGREGATION_INTEROP.md

Mapping between data-side reducers and coordinator aggregation policies

Cross-repo reducer drift is explicit

Performance probes

docs/PERFORMANCE.md

Private 0.2.0 performance sanity probes

Critical data-path regressions are measurable

Source design

docs/design/source/ml_data_specification_v1.md

Full ML_DATA contract

Used as implementation source of truth

Core crate

crates/dag-ml-data-core

Schema, representation and data-plan primitives

cargo test -p dag-ml-data-core passes

C ABI crate

crates/dag-ml-data-capi

FFI-safe helpers and DataVTable shape

Header mirrors Rust ABI structs

CLI crate

crates/dag-ml-data-cli

Local validation and fingerprinting

Example schema fingerprints

Python provider package

crates/dag-ml-data-capi/bindings/python

Installable stdlib-only ctypes provider shim (dag_ml_data_provider)

Python can materialize, view, read identity/features/targets/tensors and release handles

Validation Checklist

Check

Command

Rust formatting

cargo fmt --all --check

Rust tests

cargo test --workspace

Lints

cargo clippy --workspace --all-targets -- -D warnings

Contract schema syntax

python3 -m json.tool docs/contracts/coordinator_data_plan_envelope.schema.json >/dev/null && python3 -m json.tool docs/contracts/feature_fusion_selector.schema.json >/dev/null

Shared contract drift

DAG_ML_REPO=../dag-ml python3 scripts/validate_contracts.py

Example schema

cargo run -p dag-ml-data-cli -- fingerprint-schema examples/minimal_schema.json

Python ABI smoke

cargo build -p dag-ml-data-capi --lib && PYTHONPATH=crates/dag-ml-data-capi/bindings/python python3 examples/python/provider_smoke.py --envelope examples/fixtures/oof_campaign/coordinator_data_plan_envelope_nir.json --request examples/fixtures/oof_campaign/materialization_request_model_base_x.json