Document Collections with Relational Queries

Fauna stores data as documents but lets a query join across collections the way a relational database would. A Beam agent reads a record from the collection a workflow is watching, using the query a team has approved, and writes an update or a new document back once its rule applies, whether that means marking a record processed or creating a related entry. This keeps data in one place without a separate export step to a reporting database. Queries that return more matches than expected, or that hit a schema conflict, are set aside for a person to check before any write goes through.

ACID Transactions Across Documents

Fauna commits changes across multiple documents as a single transaction, so a set of related writes either all succeed or none do. A Beam agent bundles the writes a workflow needs, such as updating a record and creating a linked entry at once, into one transaction and submits it under the rule a team has approved. If the transaction commits, the agent notifies whoever owns the affected records; if it fails, nothing partial is left behind to clean up. Transactions that fail repeatedly, or that touch a collection outside the agent's configured scope, are handed to a person to investigate.

Schema and Index Management

A Fauna database defines indexes that determine which queries run efficiently, and those indexes need to match the fields a workflow actually searches on. A Beam agent reads query patterns from an approved configuration and confirms the needed index exists before running a lookup, rather than issuing a query that scans an entire collection. Where an index is missing for a query a team wants to automate, the agent notes the gap instead of running a slow, unindexed query silently. Any change to the schema, such as adding a field type or altering a collection's structure, is left to a person to plan and apply.