Struct KnowledgeApi 

pub struct KnowledgeApi { /* private fields */ }

Knowledge base management

Implementations

impl KnowledgeApi

pub async fn create_knowledge_base( &self, body: &KnowledgeBaseCreate, ) -> Result<KnowledgeBase>

Create a knowledge base

POST /api/v1/knowledge-bases

Required scopes: memory:write.

pub async fn delete_kb_document( &self, id: &str, doc_id: &str, ) -> Result<DeleteKbDocumentResponse>

Delete document

DELETE /api/v1/knowledge-bases/{id}/documents/{docId}

Required scopes: memory:write.

pub async fn delete_knowledge_base( &self, id: &str, ) -> Result<DeleteKnowledgeBaseResponse>

Delete knowledge base

DELETE /api/v1/knowledge-bases/{id}

Required scopes: memory:write.

pub async fn get_knowledge_base(&self, id: &str) -> Result<KnowledgeBase>

Get knowledge base

GET /api/v1/knowledge-bases/{id}

Required scopes: memory:read.

pub async fn ingest_kb_document( &self, kb_id: &str, body: &IngestKbDocumentRequest, ) -> Result<IngestKbDocumentResponse>

Ingest document into knowledge base

POST /api/v1/knowledge-bases/{kbId}/documents

Required scopes: memory:write.

pub async fn list_kb_documents( &self, kb_id: &str, ) -> Result<ListKbDocumentsResponse>

List documents in knowledge base

GET /api/v1/knowledge-bases/{kbId}/documents

Required scopes: memory:read.

pub async fn list_knowledge_bases(&self) -> Result<ListKnowledgeBasesResponse>

List knowledge bases

GET /api/v1/knowledge-bases

Required scopes: memory:read.

pub async fn reindex_knowledge_base( &self, kb_id: &str, ) -> Result<ReindexKnowledgeBaseResponse>

Re-embed every chunk with the current model

Recovers a knowledge base that was indexed without embeddings (keyword-only) and clears embedding drift after a model change. Requires an embeddings backend: without one the answer is 503 and nothing is written.

POST /api/v1/knowledge-bases/{kbId}/reindex

Required scopes: memory:write.

pub async fn search_knowledge_base( &self, kb_id: &str, body: &SearchKnowledgeBaseRequest, ) -> Result<KnowledgeBaseSearchResult>

Retrieve chunks from one knowledge base

Vector search when an embeddings backend is configured, keyword scoring when it is not — mode says which ran, and a vector pass that matches nothing falls back to keyword rather than answering empty. status distinguishes the three outcomes a caller must render differently: KB_EMPTY (nothing indexed yet), NO_MATCHES (indexed, nothing matched) and RESULTS_FOUND. All three are 200: an empty result is an answer here, not a failure.

POST /api/v1/knowledge-bases/{kbId}/search

Required scopes: memory:write.

pub async fn update_knowledge_base( &self, id: &str, body: &KnowledgeBaseUpdate, ) -> Result<KnowledgeBase>

Update knowledge base

WRITE SEMANTICS: merges. Measured 2026-08-31 on the wire: {name} left description intact and vice versa, and chunk_size and embedding_model survived both. Note this is the OPPOSITE of PUT /agents/{id}/schedule, which replaces — the verb decides nothing here, see docs/WRITE_SEMANTICS.md.

PUT /api/v1/knowledge-bases/{id}

Required scopes: memory:write.

Trait Implementations

impl Clone for KnowledgeApi

fn clone(&self) -> KnowledgeApi

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for KnowledgeApi

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.