Unified Vector Space
Even separately built knowledge DBs are aligned into the same vector space. A single query can cross multiple DBs.
MAGELLAN Knowledge DB | Knowledge Platform
Organization-wide
knowledge,
in a single vector space.
MAGELLAN Knowledge DB is a platform that supports organization-wide knowledge DB (RAG) construction. It consolidates knowledge DBs that tend to proliferate across departments and projects into a form usable cross-cuttingly within a unified vector space. Add the ComPath feature and external public information — such as primary sources from multiple countries — can be used alongside internal knowledge.
The knowledge is inside
the company.
But you can't
use it cross-cuttingly.
01
RAGs built by each department and project use different vector spaces. You cannot query across knowledge DBs.
02
Which knowledge can be shown to whom — proper access control at the DB level is hard, and integration stalls.
03
Simple vector-store RAG has limits when it comes to reliable keyword search and grasping conceptual connections.
Unify, govern,
and read by structure.
Per-DB Access Control
Fine-grained control over who can query which knowledge — at the DB level. Cross-cutting use and governance, reconciled.
Hybrid Structure
Beyond simple vector-store RAG — keywords and concept structures such as an LLM Wiki are also incorporated. Reliable retrieval, by meaning or by word.
Scattered knowledge
becomes
the power to decide.
Organization-wide knowledge that is unified, cross-cuttingly accessible, and properly governed becomes the foundation not only for human judgment, but also for the AI agents that run on the front lines (Ravel and others) and for the FDE co-build alongside them. Knowledge accumulated across departments is reliably retrieved when needed — once-fragmented knowledge becomes a single foundation that fuels decision-making.
With ComPath,
use external primary sources
alongside your knowledge.
Adding the ComPath feature to MAGELLAN Knowledge DB lets you use, alongside internal knowledge, external public information — such as primary sources from multiple countries — within the same vector space. ComPath detects "what is happening" from primary sources such as permitted government sites and major media in each country, presents them as titles and short summaries you can grasp at a glance, and links to the original sources. The full original text is not displayed, and sources are always cited — a copyright-respectful way to place the external "now" on the same foundation as your internal knowledge.
ComPath is an add-on. Stack as many external primary sources onto your internal knowledge DBs as you need.