Options

Connecting to the API

The widget talks to two Verbatim API domains: Thread (/v1/thread/*) to open a conversation, and Post
(/v1/post/*) to load history, send messages, and fetch source attachments. Every request carries your
X-Access-Token.

ChatbotWidget.mountChatbotWidget("#verbatim-chatbot", {
    accessToken: "YOUR_ACCESS_TOKEN",
    apiBaseUrl: "https://api.verbatim-ai.com", // optional; this is the default
    corpusIds: ["YOUR_CORPUS_ID"],             // the documents to answer from
    agentId: "YOUR_AGENT_ID",                  // optional; omit for the default agent
    lang: "en"                                 // optional
});

• accessToken — required. Without a valid token the API returns 401 and the widget cannot load history or send
  messages.
• corpusIds — required (at least one). The assistant answers from these corpora. Provide the IDs configured in
  your Verbatim account. An empty array is rejected at mount time.
• agentId — optional. Sends every question to a specific agent. See Choosing an agent.
• lang — hints the answer language for each message.

model is deprecated and ignored. A conversation no longer carries a model: how a question is answered is
decided per question by an agent. Use agentId instead. The option is still accepted so existing snippets keep
working, it simply has no effect.

Choosing an agent

An agent is the configuration on the Verbatim side that decides how a question gets answered — which model,
which prompt, which retrieval settings. Your organization has a default one, and can have several.

agentId points a widget at one of them:

ChatbotWidget.mountChatbotWidget("#verbatim-chatbot", {
    accessToken: "YOUR_ACCESS_TOKEN",
    corpusIds: ["YOUR_CORPUS_ID"],
    agentId: "11111111-2222-3333-4444-555555555555"
});

It is entirely optional. Leave it out and the widget behaves exactly as it always has: the question is sent
without an agent and the platform answers with your organization's default agent. Existing embed snippets need no
change.

Two things are worth knowing before you set it:

• Changing the agent starts a fresh conversation. A visitor's conversation is tied to the agent that answered
  it, so pointing the widget at a different agent gives them a clean one instead of a history written by someone
  else. Widgets without agentId keep the conversation they already had.
• If the id is not one of your agents, questions fail. The visitor sees the widget's error bubble. Check the id
  against GET /v1/agent/, or remove the option to fall back to the default agent.

To offer a choice of agents on one page, mount a widget per agent, each on its own container — they keep separate
conversations.

Where to get your Agent ID?

By API

Use GET /v1/agent/ to list the agents available to your organization:

curl -X 'GET' \
  'https://api.verbatim-ai.com/v1/agent/?pageSize=25&pageIndex=0' \
  -H 'accept: application/json' \
  -H 'Authorization: Bearer MY_JWT_TOKEN'

The id of each item is what you pass as agentId.

Using your back office

Open your Agents page and select your agent — the id is on the chip under its title, same as for a corpus.

The mount function

ChatbotWidget.mountChatbotWidget(target, options, mountOptions ?)

Calling mountChatbotWidget again on the same target updates the widget in place with the new options (it does not
create a second instance).

All options live in the second argument to mountChatbotWidget.

Connection

Content & branding

Appearance

Behavior & layout