base-de-connaissances

---

title: Knowledge Base sidebar_label: Knowledge Base (16) sidebar_position: 3 doc_type: reference

Knowledge Base

Overview

Search your documents, ask questions, and leverage your knowledge base.

Prerequisites

• API key with the MCP preset and required scopes
• Configured MCP client (see Configuration)

Quick start examples

Search your documents

1. List your document sources :

{
  "name": "knowledge_list_sources",
  "arguments": {}
}

2. Search the documents :

{
  "name": "knowledge_search",
  "arguments": {
    "query": "politique de remboursement",
    "maxResults": 5
  }
}

Ask a question with citations

1. Get a natural language answer with citations :

{
  "name": "knowledge_ask",
  "arguments": {
    "question": "Quelle est la politique de retour ?"
  }
}

---

knowledge_analytics

[ANALYTICS] Get comprehensive analytics for the Knowledge Library.

USE WHEN:

• Need to assess knowledge base health and coverage
• Want to find gaps in documentation
• Monitor search usage and quality metrics RETURNS: Overview (counts, storage), usage (queries, top searches), coverage (by source/category), quality metrics REQUIRES: none (all optional); optional: espaceId, periodDays, staleDaysThreshold RELATED:
• knowledge_search: Search documents
• knowledge_list_sources: List sources EXAMPLE: "What is the coverage of our knowledge base?" -> this tool

Scope	mcp.read
Read-only	Yes
Destructive	No
Idempotent	Yes

Parameters

Parameter	Type	Required	Description
espaceId	string (uuid)	No	Filter analytics to a specific espace
periodDays	number	No	Time period for usage metrics (days) (default : 30)
staleDaysThreshold	number	No	Days after which a document is considered stale (default : 90)

Response

Field	Type	Description
overview	object
totalDocuments	number
totalChunks	number
totalSources	number
storageUsedMb	number
lastUpdated	string
usage	object
searchQueries	number
uniqueUsers	number
topQueries	array
query	string
count	number
zeroResultsRate	number
periodDays	number
coverage	object
bySource	array
source	string
sourceType	string
documents	number
byCategory	array
category	string
unmappedDocuments	number
staleDocuments	number
quality	object
avgChunksPerDocument	number
documentsWithoutChunks	number
lowQualityDocuments	number
duplicateCandidates	number

Related tools : knowledge_search, knowledge_list_sources

---

knowledge_ask

[RAG Q&A] Ask a question and get an answer with citations from the Knowledge Library.

USE WHEN:

• User asks a question that should be answered from documents
• Need synthesized answer with source citations
• Want to query the knowledge base in natural language RETURNS: Answer text, confidence score, citations with page numbers, follow-up questions REQUIRES: query (2-2000 chars); optional: sourceIds, maxSources, answerLength RELATED:
• knowledge_search: Raw document search without synthesis
• knowledge_get_document: Get full document after finding via ask
• knowledge_entity_docs: Find documents linked to a specific entity EXAMPLE: "What is our KYC onboarding process?" -> this tool

Scope	mcp.read
Read-only	Yes
Destructive	No
Idempotent	Yes

Parameters

Parameter	Type	Required	Description
query	string	Yes	The question to answer
answerLength	choice (concise, detailed, comprehensive)	No	Preferred answer length (default : "detailed")
maxSources	number	No	Maximum number of source documents to retrieve (default : 5)
sourceIds	array	No	Filter to specific knowledge source IDs

Response

Field	Type	Description
answer	string
confidence	number
citations	array
followUpQuestions	array
metadata	object
sourcesUsed	number
totalTimeMs	number

Example

{
  "name": "knowledge_ask",
  "arguments": {
    "question": "Quelle est la politique de retour ?"
  }
}

Related tools : knowledge_search, knowledge_get_document, knowledge_entity_docs

---

knowledge_bulk_reprocess

[ADMIN] Trigger bulk reprocessing of knowledge documents.

USE WHEN:

• Documents failed initial processing and need retry
• Embeddings need to be regenerated (model changed)
• Need to reindex stale documents RETURNS: Job ID, status, affected document count NOTE: Default is dryRun=true (preview mode). Set dryRun=false to execute. REQUIRES: none required; optional: scopeType, sourceIds, documentIds, stages, dryRun, priority RELATED:
• knowledge_analytics: Check which documents need reprocessing
• knowledge_list_sources: List sources for scopeType=source EXAMPLE: "Reprocess all failed documents" -> this tool

Scope	mcp.write
Read-only	No
Destructive	Yes
Idempotent	No

Parameters

Parameter	Type	Required	Description
stages	array	Yes	Processing stages to re-run
documentIds	array	No	Document IDs (required when scopeType=documents)
dryRun	boolean	No	Preview mode: show what would be reprocessed without executing (default : true)
priority	choice (low, normal, high)	No	Job priority (default : "normal")
scopeType	choice (all, source, documents, query)	No	Scope type: all documents, by source, specific documents, or by query (default : "all")
sourceIds	array	No	Source IDs (required when scopeType=source)

Response

Field	Type	Description
jobId	string
status	string
dryRun	boolean
affectedDocuments	number
stages	array
scopeType	string

Related tools : knowledge_analytics, knowledge_list_sources

---

knowledge_create_source

[WRITE] Create a new document source in the knowledge library.

USE WHEN: • Need to create a folder/source to organize documents • Setting up a new integration (GitHub, Confluence, etc.) • Creating an upload target for manual document ingestion RETURNS: Source ID, name, type, status SOURCE TYPES: • upload: Manual document uploads • github: GitHub repository • confluence: Confluence space • notion: Notion workspace • url_crawl: Web URL crawler • s3: S3 bucket REQUIRES: name, type; optional: description, config, syncFrequencyMinutes RELATED: • knowledge_upload_document: Upload documents to this source • knowledge_list_sources: List existing sources • knowledge_sync_source: Trigger source synchronization EXAMPLE: "Create a source for technical docs" → this tool

Scope	mcp.write
Read-only	No
Destructive	No
Idempotent	No

Parameters

Parameter	Type	Required	Description
config	record	Yes	Source-specific configuration (e.g., URL for url_crawl, repo for github)
name	string	Yes	Source name
type	choice (upload, github, confluence, notion, drive, s3, url_crawl, connecteur, livedata)	Yes	Source type
description	string	No	Source description
espaceId	string (uuid)	No	Espace ID to associate source with
syncFrequencyMinutes	number	No	Automatic sync frequency in minutes (0 = manual only, max 10080 = 1 week)

Response

Field	Type	Description
sourceId	string
name	string
type	string
status	string
description	string
createdAt	string

Related tools : knowledge_upload_document, knowledge_list_sources, knowledge_sync_source

---

knowledge_delete_document

[WRITE] Delete a document from the knowledge library.

USE WHEN: • Need to remove an obsolete or incorrect document • Cleaning up duplicate documents • Decommissioning outdated content RETURNS: Confirmation of deletion NOTE: This is a soft delete. The document is marked as deleted but can be recovered. Associated chunks, embeddings, and graph entities are cascaded. REQUIRES: documentId RELATED: • knowledge_list_documents: Find document IDs to delete • knowledge_get_document: Verify document before deleting EXAMPLE: "Delete document abc-123" → this tool

Scope	mcp.write
Read-only	No
Destructive	Yes
Idempotent	Yes

Parameters

Parameter	Type	Required	Description
documentId	string (uuid)	Yes	ID of the document to delete

Response

Field	Type	Description
documentId	string
deleted	boolean
message	string

Related tools : knowledge_list_documents, knowledge_get_document

---

knowledge_entity_docs

[BRIDGE] Find documents linked to an ontology entity.

USE WHEN:

• Need to find documentation about a specific entity
• Bridge between ontology entities and knowledge documents
• "What documents mention Customer?" type queries RETURNS: Entity info, linked documents with relevance, related entities REQUIRES: entityId (UUID); optional: entityType, includeLinkedEntities, limit RELATED:
• knowledge_search: General document search
• knowledge_ask: Ask questions about an entity
• ontology_get_node: Get entity details EXAMPLE: "What documents are linked to entity X?" -> this tool

Scope	mcp.read
Read-only	Yes
Destructive	No
Idempotent	Yes

Parameters

Parameter	Type	Required	Description
entityId	string (uuid)	Yes	Entity ID to find documents for
entityType	string	No	Entity type filter (e.g., ObjectType, LinkType)
includeLinkedEntities	boolean	No	Include documents from related entities (1-hop) (default : false)
limit	number	No	Maximum documents to return (default : 20)

Response

Field	Type	Description
documents	array
documentCount	number
relatedEntities	array

Related tools : knowledge_search, knowledge_ask, ontology_get_node

---

knowledge_explain_results

[ANALYSIS] Understand why documents ranked as they did in search.

USE WHEN: • Debugging search quality or unexpected rankings • Refining queries based on score breakdown • Understanding relevance signals for each result RETURNS: Per-stage scores (vector, lexical, graph, fusion), matched terms, ranking factors, query analysis FEATURES: • Score breakdown: vector (semantic), lexical (keywords), graph (entity links), fusion (RRF) • Query analysis: detected language, entity mentions, refinement suggestions • Per-chunk score details with match type classification • Human-readable ranking factor explanations REQUIRES: query (min 2 chars); optional: topK (default 3), espaceId, includeChunkDetails RELATED: • knowledge_search: Perform the actual search • knowledge_get_document: Get full document after analysis • knowledge_get_related: Explore related content EXAMPLE: "Why did doc X rank higher?" or "Explain search relevance" → this tool

Scope	mcp.read
Read-only	Yes
Destructive	No
Idempotent	Yes

Parameters

Parameter	Type	Required	Description
query	string	Yes	The search query to analyze
espaceId	string (uuid)	No	Filter by espace ID
includeChunkDetails	boolean	No	Include per-chunk score breakdown (default : true)
topK	number	No	Number of top results to explain (default: 3) (default : 3)

Response

Field	Type	Description
query	string
explanations	array
searchConfig	object
vectorEnabled	boolean
lexicalEnabled	boolean
graphBoostEnabled	boolean
rerankEnabled	boolean
minScoreThreshold	number

Related tools : knowledge_search, knowledge_get_document, knowledge_get_related

---

knowledge_get_chunks

[QUERY] Retrieve specific chunks from a document with hierarchical context.

USE WHEN: • Need specific sections of a document after search • Want to expand context around cited chunks • Building citations for referenced content RETURNS: Chunk text with page numbers, section titles, parent context, entity mentions FEATURES: • Hierarchical parent chunk context for better understanding • Flexible retrieval: by chunk IDs or index range • Entity mention extraction from chunks • Section-level metadata (title, level) REQUIRES: documentId; optional: chunkIds (specific), startIndex/endIndex (range), includeParent, limit RELATED: • knowledge_get_document: Get full document metadata • knowledge_search: Find chunks first with relevance scores • knowledge_explain_results: Understand why chunks matched EXAMPLE: "Get chunks 5-10 from the compliance doc" or "Retrieve specific chunk by ID" → this tool

Scope	mcp.read
Read-only	Yes
Destructive	No
Idempotent	Yes

Parameters

Parameter	Type	Required	Description
documentId	string (uuid)	Yes	Document ID
chunkIds	array	No	Specific chunk IDs to retrieve
endIndex	number	No	End chunk index for range query
includeParent	boolean	No	Include parent chunk text for hierarchical context (default : true)
limit	number	No	Maximum chunks to return (default : 10)
startIndex	number	No	Start chunk index for range query

Response

Field	Type	Description
documentId	string
documentTitle	string
chunks	array
totalChunks	number

Related tools : knowledge_get_document, knowledge_search, knowledge_explain_results

---

knowledge_get_document

[QUERY] Retrieve a document with full metadata and optional chunks.

USE WHEN: • Need full document details after finding it in search • Want to examine document metadata, status, or processing info RETURNS: Document metadata (title, status, tags, pageCount, chunkCount), optionally with text chunks FEATURES: • Optional chunk inclusion with configurable limit • Full processing status visibility (indexed, failed, pending) • Token counts for context budget planning REQUIRES: documentId (from knowledge_search or knowledge_list_sources); optional: includeChunks, chunkLimit RELATED: • knowledge_get_chunks: Get specific chunks with hierarchical context • knowledge_get_related: Find similar documents • knowledge_search: Find documents first EXAMPLE: "Get the full compliance policy document" or "Show document details with chunks" → this tool

Scope	mcp.read
Read-only	Yes
Destructive	No
Idempotent	Yes

Parameters

Parameter	Type	Required	Description
documentId	string (uuid)	Yes	Document ID to retrieve
chunkLimit	number	No	Maximum chunks to return if includeChunks is true (default : 20)
includeChunks	boolean	No	Include document chunks in response (default : false)

Response

Field	Type	Description
document	object
id	string
title	string
filename	string
mimeType	string
status	string
category	string
tags	array
description	string
pageCount	number
chunkCount	number
totalTokens	number
createdAt	string
updatedAt	string
effectiveDate	string
authority	string
language	string
isSuperseded	boolean
imageCount	number
canonicalId	string
chunks	array
index	number
text	string
tokenCount	number
pageStart	number
pageEnd	number
sectionTitle	string

Related tools : knowledge_search, knowledge_get_chunks, knowledge_get_related

---

knowledge_get_related

[DISCOVERY] Find documents related to a given document.

USE WHEN: • Exploring connected knowledge from a starting document • Finding similar documents for cross-referencing • Discovering entity-linked content across the library RETURNS: Related docs by type (semantic, entity, metadata), shared entities, similarity scores FEATURES: • Multi-relationship discovery: semantic similarity, shared entities, metadata overlap • Graph-based entity link traversal • Configurable similarity threshold and relation types • Deduplication across relationship types REQUIRES: documentId; optional: relationTypes (semantic|entity|metadata), minSimilarity, limit, excludeIds RELATED: • knowledge_search: Find initial document to explore from • knowledge_get_document: Get full details of related docs • knowledge_explain_results: Understand why documents are related EXAMPLE: "What docs are similar to this?" or "Find related policies" → this tool

Scope	mcp.read
Read-only	Yes
Destructive	No
Idempotent	Yes

Parameters

Parameter	Type	Required	Description
documentId	string (uuid)	Yes	Source document ID to find related content for
minSimilarity	number	Yes	Minimum similarity threshold (0-1)
relationTypes	array	Yes	Types of relationships to consider
excludeIds	array	No	Document IDs to exclude from results
limit	number	No	Maximum number of related documents per type (default : 5)

Response

Field	Type	Description
sourceDocument	object
id	string
title	string
category	string
tags	array
relatedDocuments	array
byType	object
semantic	array
entity	array
metadata	array
sharedEntities	array
stats	object
totalFound	number
semanticCount	number
entityCount	number
metadataCount	number

Related tools : knowledge_search, knowledge_get_document, knowledge_explain_results

---

knowledge_list_documents

[DISCOVERY] List documents in the knowledge library with filtering and pagination.

USE WHEN: • Need to see all documents in the workspace • Want to browse documents without a specific search query • Filtering documents by status, tags, category, or source • Checking processing status of documents RETURNS: Paginated document list with metadata (title, status, tags, counts) FEATURES: • Filter by status: pending, ready, error, etc. • Filter by tags, category, sourceId, espaceId • Simple text search in title/filename (use knowledge_search for semantic search) • Pagination with configurable page size • Sorting by creation date, title, or file size REQUIRES: All parameters optional; page/limit for pagination RELATED: • knowledge_search: For semantic/hybrid search with query • knowledge_get_document: Get full details of a specific document • knowledge_list_sources: List document sources EXAMPLE: "List all documents" or "Show documents with status error" or "What documents are in the library?" → this tool

Scope	mcp.read
Read-only	Yes
Destructive	No
Idempotent	Yes

Parameters

Parameter	Type	Required	Description
category	string	No	Filter by document category
espaceId	string (uuid)	No	Filter by espace ID
limit	number	No	Maximum documents per page (default : 20)
page	number	No	Page number for pagination (default : 1)
search	string	No	Search in title and filename (simple text match, not semantic search)
sortBy	choice (created_at, updated_at, title, file_size_bytes)	No	Sort field (default : "created_at")
sortOrder	choice (asc, desc)	No	Sort order (default : "desc")
sourceId	string (uuid)	No	Filter by source ID
status	choice (pending, extracting, chunking, embedding, tree_building, entity_linking, graph_upserting, ready, error)	No	Filter by document status (e.g., "ready", "pending", "error")
tags	array	No	Filter by document tags (documents must have ALL specified tags)

Response

Field	Type	Description
documents	array
total	number
page	number
limit	number
hasMore	boolean

Related tools : knowledge_search, knowledge_get_document, knowledge_list_sources

---

knowledge_list_sources

[DISCOVERY] List document sources in the knowledge library.

USE WHEN: • Need to see where documents come from • Checking sync status of external sources • Auditing configured integrations (GitHub, Confluence, etc.) RETURNS: Sources list with type, status, sync timestamps, document counts FEATURES: • Multiple source types: upload, github, confluence, notion, drive, s3, url_crawl, airbyte • Sync status tracking (active, paused, error) • Last/next sync timestamps for monitoring REQUIRES: optional: type, status, espaceId, limit (default 50) RELATED: • knowledge_search: Find documents from specific sources • knowledge_get_document: Get document details including source EXAMPLE: "What sources are configured?" or "Show GitHub connections" → this tool

Scope	mcp.read
Read-only	Yes
Destructive	No
Idempotent	Yes

Parameters

Parameter	Type	Required	Description
espaceId	string (uuid)	No	Filter by espace ID
limit	number	No	Maximum sources to return (default : 50)
offset	number	No	Number of sources to skip for pagination (default : 0)
status	choice (active, paused, error)	No	Filter by source status
type	choice (upload, github, confluence, notion, drive, s3, url_crawl, connecteur)	No	Filter by source type

Response

Field	Type	Description
sources	array
total	number

Example

{
  "name": "knowledge_list_sources",
  "arguments": {}
}

Related tools : knowledge_search, knowledge_get_document, knowledge_list_documents

---

knowledge_search

[SEARCH] Perform hybrid search across knowledge library (vector + lexical + graph).

USE WHEN: • Looking for information, documents, or answers in the knowledge base • Need semantic + keyword search combined RETURNS: Ranked results with document IDs, titles, relevant chunks, citations FEATURES: • Hybrid search: vector (semantic) + lexical (keyword) + graph (entity-linked) • RRF fusion for optimal ranking • Optional reranking with cross-encoder • Graph boost for entity-linked content REQUIRES: query (min 2 chars); optional: limit, tags, category, espaceId, graphBoost, useReranking RELATED: • knowledge_get_document: Get full doc after finding in search • knowledge_explain_results: Understand why results ranked as they did • knowledge_get_related: Find similar documents EXAMPLE: "Search for compliance policy" or "Find information about KYC" → this tool

Scope	mcp.read
Read-only	Yes
Destructive	No
Idempotent	Yes

Parameters

Parameter	Type	Required	Description
query	string	Yes	Search query text
category	string	No	Filter by document category
dateFrom	string	No	Filter docs effective after this date (ISO 8601)
dateTo	string	No	Filter docs effective before this date (ISO 8601)
docAuthority	array	No	Filter by document authority level
espaceId	string (uuid)	No	Filter by espace ID
excludeSuperseded	boolean	No	Exclude superseded/outdated documents
graphBoost	boolean	No	Enable graph-based boosting for entity-linked content (default : true)
language	string	No	Filter by detected language (fr, en, de, es, it, pt, nl)
limit	number	No	Maximum number of results (default : 10)
tags	array	No	Filter by document tags
useReranking	boolean	No	Enable cross-encoder reranking for improved relevance (default : true)

Response

Field	Type	Description
results	array
total	number
query	string

Example

{
  "name": "knowledge_search",
  "arguments": {
    "query": "politique de remboursement",
    "maxResults": 5
  }
}

Related tools : knowledge_get_document, knowledge_explain_results, knowledge_get_related

---

knowledge_similar

[SIMILARITY] Find documents similar to a given document.

USE WHEN:

• Want to find related/similar documents
• Need to discover document clusters
• Looking for potential duplicates RETURNS: List of similar documents with scores and overlapping topics REQUIRES: documentId (UUID); optional: limit, minScore RELATED:
• knowledge_search: Search by text query
• knowledge_get_related: Get related via graph EXAMPLE: "Find documents similar to this one" -> this tool

Scope	mcp.read
Read-only	Yes
Destructive	No
Idempotent	Yes

Parameters

Parameter	Type	Required	Description
documentId	string (uuid)	Yes	Source document ID to find similar documents for
minScore	number	Yes	Minimum similarity score threshold
limit	number	No	Maximum number of similar documents to return (default : 10)

Response

Field	Type	Description
similar	array
id	string
title	string
score	number
overlapTopics	array
summary	string
totalFound	number

Related tools : knowledge_search, knowledge_get_related, knowledge_get_document

---

knowledge_sync_source

[WRITE] Trigger synchronization of a knowledge source.

USE WHEN: • Need to refresh documents from an external source • Re-ingesting after source configuration changes • Forcing a full re-sync to catch missed updates RETURNS: Ingestion run ID, status, run type RUN TYPES: • delta: Only process new/changed documents (faster, default) • full: Re-ingest all documents from source (slower, thorough) REQUIRES: sourceId; optional: runType (default: delta) RELATED: • knowledge_list_sources: Find source IDs • knowledge_create_source: Create a new source first EXAMPLE: "Sync the technical docs source" → this tool

Scope	mcp.write
Read-only	No
Destructive	Yes
Idempotent	No

Parameters

Parameter	Type	Required	Description
sourceId	string (uuid)	Yes	ID of the knowledge source to synchronize
runType	choice (full, delta)	No	Sync type: "full" re-ingests everything, "delta" only processes changes (default : "delta")

Response

Field	Type	Description
runId	string
sourceId	string
runType	string
status	string
message	string

Related tools : knowledge_list_sources, knowledge_create_source

---

knowledge_upload_document

[WRITE] Upload and ingest a document into the knowledge library.

USE WHEN: • Need to add a new document to the knowledge base • Ingesting text content for RAG search • Uploading a file (PDF, Markdown, etc.) encoded as base64 RETURNS: Document ID, status (pending → processing pipeline), filename NOTE: Document goes through 6-stage processing pipeline (extract → chunk → embed → tree → entity link → graph). Use knowledge_get_document to check processing status. REQUIRES: title, content; optional: contentType, mimeType, description, tags, category, sourceId RELATED: • knowledge_get_document: Check document processing status • knowledge_create_source: Create a source to organize documents • knowledge_search: Search uploaded documents EXAMPLE: "Upload this text as a knowledge document" → this tool

Scope	mcp.write
Read-only	No
Destructive	No
Idempotent	No

Parameters

Parameter	Type	Required	Description
content	string	Yes	Document content: either base64-encoded file data or plain text content
title	string	Yes	Document title
category	string	No	Document category
contentType	choice (base64, text)	No	Content encoding: "base64" for binary files, "text" for plain text (default : "text")
description	string	No	Document description
espaceId	string (uuid)	No	Espace ID to associate document with
filename	string	No	Original filename (auto-generated from title if not provided)
mimeType	string	No	MIME type of the document (e.g., "application/pdf", "text/markdown", "text/plain") (default : "text/plain")
sourceId	string (uuid)	No	Source ID to associate document with (uses default upload source if not provided)
tags	array	No	Document tags for categorization

Response

Field	Type	Description
documentId	string
title	string
filename	string
mimeType	string
fileSizeBytes	number
status	string
sourceId	string
message	string

Related tools : knowledge_get_document, knowledge_create_source, knowledge_search

---

Troubleshooting

Code	Message	Solution
404	Document not found	Check the ID with knowledge_list_sources.

Need help?

Contact us: Support and contact.