Demo

Knowledge bases

Turn your documents into a reliable knowledge base

Import your files, index websites, transcribe audio and video, and screen every document before it feeds into answers.

Read the “Preparing your documents” guideDiscuss your project
DocumentsWebsitesContent validation Import
DocumentTypeStatus
User guide.pdfPDFApproved
Pricing 2026.xlsxExcelApproved
Offer overview.pptxPowerPointAnalyzing
Onboarding training.mp4VideoTranscribed
Forum export.htmlHTMLTo review

Documents marked “To review” stay quarantined: they are never used in answers.

Interface diagram: documents in a corpus and their validation status (sample content).

The problem

Why it matters

An organization’s useful knowledge is spread across PDFs, slide decks, spreadsheets, intranet pages and recordings. A language model has no access to it, and makes things up when asked.

These contents therefore need to be gathered, split and indexed, while making sure that no inappropriate or booby-trapped document pollutes the answers.

In practice

  1. Create a knowledge baseA knowledge base groups the documents of one domain; an assistant can query several of them, each with its own priority.
  2. Add contentBatch file upload, scheduled website indexing, or automatic synchronization through a workflow.
  3. ScreenOptionally, each document is analyzed before indexing; questionable cases are quarantined for a human decision.
  4. IndexText is extracted, split into chunks, enriched with its context and then embedded for search.

Capabilities

What you can do

Each capability is shown with its actual availability.

  • PDF, Word (.docx), PowerPoint (.pptx), Excel (.xlsx) documentsAvailable
  • Plain text, Markdown, CSV, HTML, JSONAvailable
  • Batch upload20 files of 20 MB per upload by default, adjustable by the administrator.Available
  • Website indexingStart URL, include/exclude patterns, depth, page count, linked PDFs, sitemap, scheduling.Pro license
  • Audio and video transcriptionMP3, WAV, M4A, MP4, MOV, MKV… via Whisper (OpenAI) or Mistral.Requires setup
  • OCR for images and scanned PDFsVia the Mistral OCR service (key required).Requires setup
  • Content validation before indexingFile safety, sensitive content, personal data, prompt injection; quarantine and human decision.Requires setup
  • Knowledge base versions with approvalProduced by workflows: comparison, approval, publication.Requires setup
  • One-off attachment in a conversationAnalyzed for the conversation without being added to the knowledge base.Available
  • SharePoint, Google Drive, OneDrive, Confluence, S3 connectorsNot available today; can be achieved by developing a connector.Under consideration
  • .doc, .xls, .ppt and OpenDocument formatsNot available

Benefits

What you gain

Reuse what you have

Your documents become usable as they are, with no prior rewriting.

Up-to-date sources

Scheduled website indexing and folder synchronization: the assistant keeps pace with your content.

A trusted gateway

With validation enabled, nothing is indexed until it is accepted: a rejected document never feeds an answer.

Requirements and limitations

Sourced answers (RAG)

Hybrid search, contextual enrichment and verified citations: every answer points to the excerpts that support it.

Document workflows

Synchronize folders, FTP/SFTP servers and websites, have changes approved and publish knowledge base versions on a schedule.

Security and governance

Roles, SSO including FranceConnect+, white-label organizations, GDPR retention, audit log, content validation and migration between environments.

Next step

Best practices for getting accurate answers from the very first tests.

Read the “Preparing your documents” guideContact us