Skip to main content

Vector Databases

An AI Scope is a knowledge base: you add documents, MonoChat splits them into pieces, turns each piece into a vector with an embedding model, and stores the vectors in a vector database. When a customer asks something, the AI searches the vector database for the passages closest in meaning and answers from them.

Settings › Automation › AI Vector DB Clients lists the vector databases your AI scopes can use.

Vector databases MonoChat provides and your own​

Like AI Models, the list can contain vector databases provided by MonoChat, offered to every workspace and read-only for you, and the ones you connect yourself. If you are happy with the one MonoChat provides, you do not need to add your own: just pick it when you create an AI scope.

Connect your own database when you want your vectors stored on infrastructure you control.

Supported provider​

MonoChat currently connects to Chroma (self-hosted or hosted; the server must be reachable from MonoChat).

Connect a vector database​

  1. Go to Settings › Automation › AI Vector DB Clients and click + (Create).
  2. Enter a Name and, optionally, a Description.
  3. Choose the Provider.
  4. Fill in the connection fields and click Save.
FieldDescription
PathThe URL of the Chroma server, for example https://chromadb.example.com.
Chroma Token (x-chroma-token) - OptionalThe token, if your server requires one. It is sent in the x-chroma-token header.
Database NameThe Chroma database to use.
Tenant NameThe Chroma tenant to use.

If the tenant or the database does not exist yet, MonoChat creates it the first time it writes to the server. Inside the database, MonoChat keeps one collection per workspace and embedding model.

Use it in an AI scope​

The vector database is chosen when you create an AI scope:

  1. Go to Settings › Automation › AI Scopes and create a scope.
  2. In AI Vector DB Clients, choose the vector database.
  3. In AI Models, choose the embedding model (a model with the Embed feature).
  4. Optionally choose a Speech to Text AI Models model for audio documents, and set the Distance Type and Distance Threshold used when searching.
warning

The vector database and the embedding model cannot be changed after the scope is created: the stored vectors only make sense for that model in that database. To move a knowledge base, create a new scope and add its documents again.

AI scopes are then used by the Create AI Chat Completion node (AI Scope Settings) and by the AI Agent node, where they become a knowledge-base search tool the agent can call. See AI Agents.

Permissions​

Access is controlled by the role permissions Read / Create / Update / Delete AI vector database clients in automation. See Roles.