metaspn-entities 0.1.7

Canonical entity resolution, aliasing, and merges for MetaSPN systems

metaspn-entities

Identity layer for MetaSPN systems.

Features

• Canonical entity IDs
• Deterministic identifier normalization + alias resolution
• Merge history and reversible soft undo
• SQLite backend using stdlib sqlite3
• Event emission payloads for EntityResolved, EntityMerged, EntityAliasAdded
• Optional filesystem snapshot export

Quick usage

from metaspn_entities import EntityResolver

resolver = EntityResolver()
resolution = resolver.resolve("twitter_handle", "@some_handle")
events = resolver.drain_events()
print(resolution.entity_id, resolution.confidence)

API notes

• resolve(identifier_type, value, context=None) -> EntityResolution
• add_alias(entity_id, identifier_type, value, ...)
• merge_entities(from_entity_id, to_entity_id, reason, ...)
• undo_merge(from_entity_id, to_entity_id, ...) (implemented as reverse merge with redirect correction)
• drain_events() -> list[EmittedEvent]
• export_snapshot(output_path) to inspect SQLite state as JSON

Event Contract Guarantees

drain_events() returns EmittedEvent objects whose event_type and payload are schema-compatible with metaspn-schemas entity events.

• EntityResolved payload keys:
  • entity_id, resolver, resolved_at, confidence, schema_version
• EntityMerged payload keys:
  • entity_id, merged_from, merged_at, reason, schema_version
• EntityAliasAdded payload keys:
  • entity_id, alias, alias_type, added_at, schema_version

Datetime fields are emitted as UTC ISO-8601 strings for deterministic serialization.

M0 Ingestion Adapter

For worker/runtime integration, use resolve_normalized_social_signal(...) with a normalized signal envelope.

from metaspn_entities import EntityResolver, resolve_normalized_social_signal

resolver = EntityResolver()
signal = {
    "source": "social.ingest",
    "payload_type": "SocialPostSeen",
    "payload": {
        "platform": "twitter",
        "author_handle": "@some_handle",
        "profile_url": "https://example.com/profiles/some-handle",
    },
}

result = resolve_normalized_social_signal(resolver, signal)
print(result.entity_id, result.confidence)
for event in result.emitted_events:
    print(event.event_type, event.payload)

Adapter behavior:

• Extracts deterministic identifier candidates from normalized payloads.
• Resolves a primary identifier, then adds remaining identifiers as aliases.
• Returns only events produced during the adapter call.

M1 Context API

Profiler/router workers can read consolidated context using:

• resolver.entity_context(entity_id, recent_limit=10)
• resolver.confidence_summary(entity_id)

Both APIs resolve canonical redirects first, so merged IDs return coherent context.

M2 Recommendation Context API

Recommendation and drafter workers can consume:

• resolver.recommendation_context(entity_id)

The recommendation context includes:

• identity confidence
• activity recency (days)
• interaction history summary (evidence count + source distribution)
• preferred channel hint
• relationship stage hint (cold / warm / engaged)
• merge-safe continuity fields keyed to canonical entity IDs

M3 Outcome Attribution API

Outcome evaluators can map attempt/outcome references back to canonical entity lineage:

• resolver.attribute_outcome(references)

Supported references include entity_id, email, canonical_url, handles, domains, and names.

Attribution guarantees:

• canonical merge redirects are resolved before returning entity_id
• output includes explicit confidence for downstream learning logic
• deterministic tie-breaks are applied by score, then hit count, then entity ID