Setting up collections

Purpose

The settings of the collection determine which language model answers, how it answers, what visitors may do and which paths lead out of the chat. Changes affect all channels of the collection: chat, search, assistants, agents.

Where

Open Edit in the collection rail. The page is divided into eight sections; on the first visit only “Basics” is open. All changes take effect after Save at the end of the page.

Changes to languages and language models affect subsequent indexing runs. Content that is already indexed is retained.

Basics

Field	Meaning
Title	Name of the collection in lists and the header.
Languages	Language codes separated by commas, for example `de,en`. Determines the language selection in the chat, the answer language and the language tabs of all texts.
Type (frontend behaviour)	Default behaviour of the embeds: chatbot or website search. The type of an assistant overrides this setting.
Schema	The technical name of the collection, derived from the title on creation, cannot be changed.
Chat / answer LLM	The language model that answers chat requests and generates texts. The choice comes from the LLM configurations (Chapter 24).
Embedding LLM	The model that creates the vectors for the index. Locked after indexing because the vector dimension is fixed.

If the collection is still on the global default without stored access credentials, a notice appears here: the chat then does not answer, and no index is created.

Additional language models and processing region

Field	Meaning
Additional LLM configurations	Additional models that can be chosen per embed: for model comparisons or a model of its own per page. A model with a different embedding gets an index of its own, which is built in the background after saving.
Processing region	“EU only” or “Germany only”: the chat widget then shows the attest “AI processing only in the EU” or “in Germany”, but only in embeds whose language model, search index and voice process in this region.

The table below shows, for each purpose (chat, embeddings, voice, speech recognition), the configuration used and its region. If the main configuration does not meet the region, the attest only appears in embeds with a suitable model. If no model meets the region, dAI Pro resets it to “No restriction” on saving and reports this. A link leads to the machine-readable attest. Chapter 20 goes deeper into the topic.

Answer behaviour

Field	Meaning
Main prompt	Prepended to every conversation: role, tone, company context, no-gos. One tab per language.
Link files in the source reference	File documents such as PDFs appear in the source reference with their title and are linked via an internal endpoint; path and file name stay hidden. Off = generic mention “our knowledge base”.
Source links jump to the passage	Web sources get a text fragment appended: supporting browsers jump directly to the quoted passage when opening. PDF sources jump to the page.
Offer answer mode “Plain language”	Shows a toggle in the chat for answers in plain language. Only the wording changes; content and sources stay the same.
Answer rating	Thumbs below every answer. Ratings land in the log and in its evaluation (Chapter 16).
Topic filter threshold	From which dissimilarity a question counts as off-topic and is politely declined. Higher = more tolerant.

Chapter 7 explains how these fields together determine the quality of the answers.

CAG configuration

dAI Pro knows two ways to an answer. RAG searches the index for the matching sections for each question. CAG passes the entire content to the language model, which delivers more precise answers for small collections and stays inexpensive thanks to the provider's cache.

Field	Meaning
CAG/RAG classifier per question	A small model decides per question between broad (CAG) and specific (RAG). Off = simple rule: if the content fits within the limit, CAG, otherwise RAG.
Default mode in the test chat	Auto, Classic (RAG) or Specialist (CAG) as the preset of the toggle.
Show toggle in the chat	Whether visitors may choose the mode themselves. Usually off for demo instances.
CAG token limit	Up to which size of the content CAG is used; above it, dAI Pro falls back to RAG. 0 = the provider's default.

Access and visitor features

Field	Meaning
Allowed websites	One address per line. Empty means: any website may embed the chat. As soon as an address is entered, only these work. Applies to chat, search, upload, handover and MCP.
Attach a file or photo	Paper clip in the chat: images and PDFs up to 10 MB are converted to text at the moment of the question and used as context; nothing is stored.
Identity secret	Lets your website tell the chat who is signed in, verified via a signed token. Chapter 13.
Answer endpoint for AI agents (MCP)	Provides the collection's chat as an endpoint for AI agents. Chapter 14.

Human handover

Shows the link “Talk to a human” in the chat and offers the handover when the chat could not help. Request and conversation history go by e-mail to the recipients; with “Offer a callback request”, visitors can leave a phone number and preferred time. Chapter 12 describes the process from the visitor's and the team's point of view.

Actions from the chat

Callback, ticket and appointment are confirmed actions: the chat prepares a card with all fields; nothing is sent until the visitor confirms. Here you connect CRM, ticket system, support mailbox and appointment calendar. Chapter 12 explains every connection.

Voice chat

Enables the microphone button, chooses voice and speech recognition and maintains the vocabulary for brand and product names. Chapter 11.

Frequently asked questions

Why can I no longer change the embedding LLM? The index is fixed to this model's vector dimension. For a different embedding model, create a new collection or enter it under “Additional language models”; a second index is then created.

I set the processing region, and after saving it says “No restriction”. None of the collection's language models processes in this region. Choose a configuration with a suitable region (Chapter 24) and set the region again.

Do the allowed websites also apply to the search? Yes, to all endpoints of the collection.