Collect claims, not mystery objects
Values arrive with typed evidence, source URLs, confidence, and provider identity.
Open-source entity enrichment ยท v0.6.0
Enrichmentyou can audit.
Resolve people and companies without losing the source behind each field.
Provider neutralBring approved search, public data, browser, or internal sources.
Evidence intactKeep URL, observation time, provider, and confidence with every claim.
Review explicitSurface conflicts and inferred claims before automated action.
Provenance first
Every field keeps its proof.
Enrichfold turns provider output into deterministic decisions while preserving what was observed, where, and when.
Reconcile deterministically
Stable rules select a suggested value and preserve every competing claim.
Gate uncertain results
Conflicts and inferred claims are marked for review instead of silently passing.
Control research cost
Reserve provider units before execution and expose partial coverage when work is skipped or fails.
Search connected
Find sources across 11 search engines.
Enrichfold searches Exa, Parallel, You.com, Tavily, Linkup, Seltz, TinyFish, Nimble, Browserbase, Serper, and DuckDuckGo. It merges duplicate URLs with explainable scores. Scrapefold calls the same search API and handles page retrieval by URL.
Search results remain sources. Use WebSearchProvider to map them into claims with evidence, then let ResearchEngine apply its review gates.
search.py
from enrichfold.search import SearchOptions, search
results = await search(
"example.com company",
SearchOptions(engines=("parallel", "tavily")),
)
for result in results:
print(result.url, result.score, result.engines)Keenable connected
Search broadly. Decide carefully.
The optional Keenable adapter discovers sources. Your mapper decides which results become claims, so search output never masquerades as verified fact.
provider.py
from enrichfold import (
Claim, Entity, Evidence, KeenableProvider,
ProviderSpec, ResearchEngine,
)
def map_results(entity, results):
for item in results:
yield Claim(
field="summary",
value=item["description"],
kind="inferred",
evidence=Evidence(
source_url=item["url"],
observed_at=item["acquired_at"],
confidence=0.8,
provider="keenable",
),
)
provider = KeenableProvider(
lambda entity: f'{entity.identifiers["domain"]} company',
map_results,
)
result = ResearchEngine([
ProviderSpec("keenable", provider, reserved_units=1),
]).run(Entity.company(domain="example.com"))Get started
Add provenance to your enrichment pipeline.
Install the Python package, connect the providers you trust, and keep review decisions in your application.