Document ingestion pipeline

Ragie takes in documents from connected sources and prepares them for retrieval by breaking them into searchable chunks. A Beam agent reads new and changed source files and applies the customer rule to ingest, tag, or hold each one before it enters the index. The ingestion result is written back with its status for the team. When a document fails to parse, arrives in an unsupported format, or carries a sensitivity marker the rule does not cover, the agent holds it and notifies the owner, so a person decides whether that content belongs in the knowledge base.

Retrieval query handling

Ragie answers retrieval requests by returning the passages most relevant to a query from the indexed content. A Beam agent reads the incoming query and applies the approved rule to run the retrieval, filter by permission, and pass the results to the calling application. The request and what was returned are logged for audit. When a query returns nothing useful, crosses a permission boundary, or asks for content the rule restricts, the agent withholds the result and routes the case to a person, keeping access decisions on sensitive material with the team.

Source freshness syncing

Ragie keeps its index aligned with the source documents as they change over time. A Beam agent reads update signals from connected sources and applies the customer rule to re-ingest a changed file, retire a removed one, or note a version shift. Each update is logged against the source for reference. When a source disappears without warning, a bulk change arrives, or a document's history looks inconsistent, the agent pauses syncing and alerts the owner, so a person confirms the change is intended before the knowledge base reflects it.