Release notes¶

Changes in each published nlp-shap version. Newest first.

Unreleased¶

ExactEstimator for lazy coalition enumeration with estimand-plugin aggregation.
iter_mask_ints() and iter_masks() for streaming masks without materializing 2^n coalitions.
estimate_attributions() wires coalition payoffs to Shapley or Banzhaf aggregators.
Vectorized exact Shapley aggregation for complete characteristic tables.
run_iter() and run_stream for bounded pending-task scheduling.
exact estimator entry point under nlp_shap.estimators.

build_coalition_key uses compact binary hashing instead of JSON serialization.
Marginal estimand aggregation uses integer bitmask keys and cached factorial weights.
sample_masks yields an iterator instead of a materialized tuple.

RunArchive with SQLite index, blob storage, and lazy history_lazy() iteration.
CoalitionRecord rows with mask bytes, utility, timing, and cache-hit metadata.
CoalitionDedupRegistry and build_coalition_key() SHA256 coalition keys.
HotResultStore LRU cache for recent generations.
InferenceScheduler with bounded max_inflight concurrency and SchedulerMetrics.

Masking codec: PackedMask, MaskCodec with legacy-compatible packed hashing.
MaskSpace for projecting feature splits to full masks.
View-based MaskedSnapshot and MaskBuilder with structural sharing.
Absence policies: DeletePolicy, PadPolicy, NeutralPolicy.
TokenPartitioner and built-in plugin registration for tokens, delete, pad, and neutral.
numpy core dependency for mask packing.

Domain types: ConversationSnapshot, Turn, Message, PlayerSet, CoalitionMask, CooperativeGame, and conversation enums.
Protocols for backends, estimators, value functions, partitions, absence policies, and embedding providers.
PluginRegistry with packaging entry-point discovery.
ExplainConfig (Pydantic v2) with YAML load/save helpers.
Configuration guide and expanded API reference pages.

Sphinx theory pages for cooperative games and estimands (Shapley vs Banzhaf).
Usage guide for estimand aggregators, results, and manifests.
Expanded API reference for public modules.
Example notebook examples/estimands_toy_game.ipynb.
Furo theme, getting-started page, embedded notebook rendering, and API module layout.

ShapleyAggregator and BanzhafAggregator for coalition-weighted attribution from masks and payoffs.
Estimand enum and wire types for honest estimand labelling.
ExplainResult with required estimand field.
RunManifest, RunManifestPayload, and parse_manifest() for run-archive metadata.
EstimandAggregator protocol.