Skip to main content
Back

Knowledge base synchronisation

Understand the process of synchronising your documents to our vector search engine: chunking, embeddings, and indexing.

WHY SYNCHRONISE?

When you add or modify documents in your knowledge base, these changes are not immediately available to your bot. Synchronisation is the step that makes your documents usable by the AI by indexing them in our vector search engine.

Each venue has its own dedicated search space, guaranteeing complete data isolation between different clients.

THE CHUNKING PROCESS

During synchronisation, your documents are automatically split into pieces (chunks) of optimal size for search:

Chunk size: Each piece contains approximately 400 tokens, with an overlap of 50 tokens between consecutive pieces.

This overlap ensures that no information is lost at split points and that context is preserved between adjacent pieces.

EMBEDDING GENERATION

Each text chunk is converted into a numerical vector (embedding) by an artificial intelligence model. This vector captures the semantic meaning of the text, allowing the bot to retrieve relevant information even if the client's question does not use exactly the same words as the document.

Embeddings are then stored in the search space dedicated to your venue.

LAUNCHING A SYNCHRONISATION

To synchronise your knowledge base, click the 'Synchronise' button in the Knowledge base section of your dashboard. The process generally takes a few seconds to a few minutes depending on the volume of documents.

A progress indicator keeps you informed of progress. Once synchronisation is complete, your bot immediately uses the updated data.

Frequently asked questions

DO I NEED TO SYNCHRONISE AFTER EVERY MODIFICATION?

Yes, synchronisation is required after each document addition, modification, or deletion for the changes to be taken into account by your bot during calls.

DOES SYNCHRONISATION INTERRUPT THE SERVICE?

No, synchronisation runs in the background and does not interrupt the service. Your bot continues to function normally with the previous data until the new synchronisation is complete.