Skip to content

Explanation

Knowledge Base

Centralize documents and knowledge, connect external sources, and retrieve useful context through semantic search and the document graph.

3 min

At a glance

The Knowledge Base centralizes your documents and files in a single space. Import your content and find information in seconds through intelligent search and the relationship graph.

Documents are presented in a list by default, with a grid available when browsing cards is more useful. Each document keeps its source, indexing state, and last activity visible. File formats use an opaque, high-contrast tile.

In the application, open the Knowledge Base at /base-de-connaissances and the Graph at /base-de-connaissances/graphe.

What you'll do

  • Import documents manually (PDF, DOCX, Markdown, TXT, HTML) or from external sources
  • Index content supplied through your Live Data connectors
  • Search intelligently using natural language and find relevant passages across all your documents
  • Visualize the links between your documents in an interactive graph that reveals hidden connections
  • Automatically link detected entities (suppliers, products, locations, etc.) to your business model

Document library

How it works

Every imported document goes through an automatic five-step process:

  1. Extraction -- raw text is extracted from the file (PDF, Word, etc.).
  2. Chunking -- the text is split into short segments so search results are precise rather than returning an entire document.
  3. Indexing -- each segment is analyzed to understand its meaning, enabling query-based search instead of exact keyword matching.
  4. Entity detection -- names of people, organizations, locations, products, etc. are identified and linked to your ontology.
  5. Graph construction -- links between documents are computed (shared entities, common tags, content similarity).

This processing runs in the background. You can keep working while your documents are being analyzed.

Concrete example: you import a supplier contract as a PDF. The system extracts the text, detects the supplier name "Dupont & Fils" and the product "Stainless steel", then connects this contract to other documents mentioning the same entities in the graph.

  • Live Data: create file connectors (Google Drive, OneDrive, SharePoint, Dropbox) to automatically index your documents in the Knowledge Base.
  • Ontology: entities detected in your documents are linked to your business model entities, enriching both modules.
  • Workflows: use semantic search blocks to query your document base from a workflow.
  • AI Agent: the agent can rely on your documents to answer contextual questions.

Expected outcome

At the end of this section, you will have a centralized document base, searchable in natural language, with a relationship graph that reveals connections between your documents and your business model.

Need help?

Write to us: Support and contact.

Uploading a file and making it searchable are separate stages. Check processing status, then search for a term from a known passage. Open the result and compare it with the original document.

For a Client–Commande contract, use a dated version and explicit name. Finding a general clause does not prove that it applies to a particular order. Check the customer, version, period and supporting passage.

Document indexes and structured records may have different freshness. Importing a document does not automatically create its mentioned instances and relationships. See Answer using records and documents.

Did this page help you achieve your goal?