Mind & knowledge: model, documents, memory and tools

Choose the AI model behind a digital employee, give it the documents its role needs, decide what it remembers, and enable only the connected tools that role may use.

These settings are what separate a chatbot from something useful. A character with no documents can only improvise; one with your price list, your policies and permission to call your booking system can finish a job.

In the character editor, the right dock's Mind & knowledge group holds Core AI, Knowledge Base, Personality, Memory, Mindview and Narrative. The Tool Registry tab sits in the Behavior group. Identity, voice, guardrails and emotion are covered in Identity, model & voice. Press Save in the toolbar (Ctrl/Cmd+S) after changing a tab; the file switches on Knowledge Base and the tool switches on Tool Registry take effect as soon as you flip them.

Core AI — the model

LLM Provider picks the family; Model picks the model within it. Choosing a provider resets the model to that provider's first entry, so check the Model list after switching.

Models marked Premium consume extra quota (a plan multiplier). See Plans & usage.

Choosing a model

Start with a mini or flash class model. They are fast and cheap, and for a kiosk or an NPC the quality difference is far smaller than the latency difference. Move up to a flagship model only when you can point at answers that were not good enough.

File & image attachments lets visitors attach one image or document per message from the chat box. Switch Images (JPG, PNG, WebP, GIF) and Documents (PDF, TXT, CSV, MD, JSON, DOCX) separately and set Max size (MB); the hard limits are 5 MB for images and 10 MB for documents. Turned off, the attach button is hidden everywhere.

Knowledge Base — what it can look up

Documents live in one library shared by your whole workspace (Knowledge › Knowledge Base in the top bar). Each character is then given permission to read specific ones on its own Knowledge Base tab, titled "Knowledge base for this character". A price list can serve five characters without being uploaded five times, and your internal HR policy can stay away from the character on your public website.

1. Drop the files onto the tab. Files you upload here are enabled for this character straight away; they also join the workspace library. Files count against your plan's Knowledge storage; when it is full, or your plan includes none (Free), the upload is refused.
2. Wait for Ready. Each file shows Processing while it is read and indexed. The list refreshes by itself; a file that cannot be read shows Failed.
3. Check the switches under Library. Every file in the workspace is listed with a switch ("Use … for this character"). Turn off anything this character has no business quoting. A disabled file stays in the library.
4. Test it. Type a realistic customer question into Search Test and press Search (or Enter). You see the passages the character would find, each with a similarity score.

More on formats, storage and writing documents that retrieve well: Knowledge Base.

Personality

Six Preset temperaments — Custom, Friendly, Professional, Creative, Witty and Empathetic — set the five Big-Five Traits at once: Openness, Conscientiousness, Extraversion, Agreeableness and Sensitivity, each 0–100. Pick the preset closest to your intent — Friendly for a greeter, Professional for banking or support, Empathetic for care and health — or choose Custom and move the sliders.

Memory — who it remembers

With Long-term Memory enabled, the character carries what it learned about a person into their next conversation — a returning customer is greeted with real recall and does not have to explain their situation twice.

While memory is on, each completed exchange is added to that visitor's record, from every client — website embed, kiosk, Unreal or phone. A background job then condenses the exchanges into a short summary and a few facts (a name, preferences, decisions, plans). Current Records lists each visitor with their summary, recent exchanges and facts; it refreshes by itself, or press Refresh.

Mindview — see what the character is told

Mindview answers the question "why is it behaving like that?". Type a test message, press Inspect, and the character's instructions are laid out by origin — Persona, Guardrails, Personality, State of Mind, Long-term Memory, Knowledge (RAG), Current Narrative Section and Tool Manifest — each with a character count, so you can see what dominates.

It is the fastest way to catch the common mistakes: a guardrail that contradicts the persona, an example dialogue that never got saved, or a tool you thought was enabled and isn't.

Narrative — guide the conversation

The Narrative Graph maps a conversation as connected nodes: add Start, Section, Trigger and End nodes, drag them into place and connect them, then select a node to edit its Name and Content. Narrative is a plan feature. See Narrative for how sections and triggers work.

Tool Registry — what it can do

A tool is a real action: check a balance, look up an order, call your own API. You define it once for the workspace under Knowledge › Tool Registry, then decide here which characters may use it. When the conversation calls for it, the character decides to call the tool itself, uses the result, and answers in its own words.

1. Open Behavior › Tool Registry. Every workspace tool is listed with its category badge; tools marked Identity required only run for a verified person.
2. Switch on the tools this character may use. A tool that is not switched on is invisible to the character. New tool and each tool's edit icon open the tool form.
3. Start a fresh conversation and try it. The tool list is handed to the character when a conversation begins, so one already in progress will not see a newly enabled tool.
4. Check Knowledge › Tool Registry › Execution Log for the audit trail.

Where to go next