RAG / Knowledge Agent

The backend agent that manages embedding and semantic retrieval over your company's institutional knowledge, giving every other agent in the orchestration access to it as shared memory.

What Is the RAG / Knowledge Agent?

The RAG / Knowledge Agent is one of TotalApp's Core Agents. It manages embedding and semantic retrieval over your organization's internal documents, historical records, and logs — giving every other agent in the orchestration access to that institutional knowledge as one shared memory layer.

Internal documents and past records are the institutional knowledge every team relies on, but without a shared retrieval layer, every agent that needs it would have to implement its own embedding and search logic — a recipe for drift and duplicated pipelines. The RAG / Knowledge Agent centralizes this: index once, query from anywhere.

In one sentence

Documents, records, and logs are embedded into one searchable collection — any other agent can query it by meaning, not just keywords, and every result comes with a source citation.

How It Works — Index Once, Query Everywhere

1. Company-Wide Indexing → 2. Semantic Retrieval → 3. Cited Answer

Phase	What happens
1. Company-Wide Indexing	Documents, records, and logs are embedded into one searchable collection. New documents are embedded and made searchable automatically — no manual re-indexing step required.
2. Semantic Retrieval	Retrieval is by semantic similarity, not exact keyword matching, so a query can find relevant material even when it doesn't share exact wording with the source document.
3. Cited Answer	Every retrieval comes with a source reference, so answers can be traced back to the document, record, or log entry they came from and independently verified.

One shared layer, not one per agent

Any other agent — Core or Industry-Specific — can query the same knowledge collection instead of implementing its own retrieval logic. This is what makes it a genuinely shared memory layer rather than a feature bolted onto one specific screen.

Key Capabilities

Company-Wide Indexing

Documents, records, and logs embedded into one searchable collection, kept current as new material arrives.

One Shared Layer

Every other agent queries the same knowledge base — no duplicated retrieval logic to maintain or let drift out of sync.

Citations Included

Every retrieval comes with a source reference, so answers can be traced and verified rather than taken on faith.

Tenant & Role Scoped

Access to the knowledge base respects the same RBAC and tenant boundaries as everything else in TotalApp — nothing bypasses entitlements just because it's a semantic search.

Relationship to the 2-Tier Architecture

The RAG / Knowledge Agent is the semantic-retrieval half of TotalApp's 2-Tier Architecture: deterministic, exact-match data lookups are handled by other engines, while fuzzy, meaning-based retrieval over unstructured knowledge goes through this agent. Workflows that need both — an exact record lookup plus contextual background from past documents — use the two together rather than picking one over the other.

Frequently Asked Questions

Do I need to manually re-index when new documents are added?

No — new documents are embedded and made searchable automatically as they're added, with no manual re-indexing step required.

Can a Core Agent and an Industry-Specific Agent both query the same knowledge base?

Yes — any agent, Core or Industry-Specific, can query the same shared knowledge collection. There's no separate index per agent type.

Is semantic search here the same as TotalApp's "Finding Similar Records" feature?

They rely on the same underlying Knowledge Engine approach to embedding and similarity search, but this agent specifically serves other agents in the orchestration as a shared knowledge layer, distinct from a single screen's own record-matching feature. See Semantic Search for the screen-level feature.

Does retrieval respect tenant boundaries?

Yes — access to the knowledge base is scoped by tenant and role exactly the same as every other data source in TotalApp, so one tenant's indexed documents are never retrievable by another tenant.