Documentation · Mode 3 · Production today

Retrieval.

The indexed-corpus integration surface. Powered by Aeira. Connectors ingest your sources, Aeira indexes them, ACLs gate the data at retrieval. Fast at query time, cheap per query.

See all three surfaces Start with Apps today →
Production today · docs landing

The Retrieval engine is in production with a deep data-ingestor catalog, ACL-aware hybrid search, and customer-tunable ranking. This page covers what ships today. The full public API reference and authoring docs land here as the surface stabilizes. See the architecture page for Aeira's ACL-aware index in more depth, or the integrations catalog for vendor coverage.

Capabilities

What Retrieval ships today.

Concrete capabilities running in customer deployments. The public authoring API and detailed docs are still being stabilized; the engine underneath is not.

103 in-house data ingestors

Ingestors cover knowledge sources, ticketing, CRM, code, files, observability, and data warehouses. New ingestors land weekly. Data ingestors are distinct from MCP servers, which power Apps and Agent tool calls. Many vendors offer both.

Airbyte bridge for the long tail

Beyond the in-house ingestor catalog, the Airbyte bridge extends Retrieval to hundreds more SaaS and database sources. Anything in the Airbyte ecosystem can become a Retrieval source.

Hybrid search

Keyword, vector, and entity-graph traversal in one query. The engine picks the right mix per query. Customers do not have to choose a search mode; they get the best of all three.

Customer-extensible ranking, trust, authority

Relevance, recency, source quality, trust signals: all shapable through policy code. The engine ships sensible defaults; customers extend with their own rules. Authorization stays platform-enforced and is never customer-overridable.

Architectural commitments

The shape of Retrieval.

Four things hold across every release of the surface. Everything else is implementation that will evolve.

ACLs gate at the index

Authorization is an architectural property, not a postprocess filter. Documents the user cannot read never enter the candidate set. Ranking and authority are kept separate from authorization.

Authority is customer-shapable. Authorization is not.

Ranking, relevance, recency, and trust signals are yours to shape. Access control is enforced by the platform and is not customer-overridable. Two contracts, one floor.

Aeira is the engine

Aeira is Sparcle's retrieval and policy engine. It owns the index, the authorization graph, and the audit chain. The same code paths run in the free single-user install and in production org deployments.

No data leaves your perimeter

The index lives where you deploy Bolt. Embeddings are generated by a model you choose. Sparcle does not host a hosted-corpus tier. The data stays where you put it.

Decision rule

When to reach for Retrieval.

Use Retrieval when the answer lives across a large corpus and the right document isn't predictable from the question. If the answer is in one of three known APIs, it's an App. If the answer is one of ten thousand documents you've written internally over five years, it's Retrieval.

Retrieval pairs naturally with the other two surfaces. An Agent can use Retrieval as a tool to ground its reasoning. An App can link to a Retrieval result as a typed row. The three surfaces are composable, not exclusive.

In the meantime

Apps is live. Retrieval shares its primitives.

The same identity, ACL, and audit-chain machinery you wire up for Apps today is what gates Retrieval tomorrow. Get the platform wired in now, light up Retrieval when the surface ships.

Try Apps today Read the architecture