The three-dot button in the panel header opens its options.
Guidance and tips
When it is on, opening a case file makes the panel tell you where things stand and suggest where to go next. When it is off, the panel only answers what you ask and suggests nothing on its own. It starts switched on.
Automatic steps
When they are on and papers come into a matter, the chat tells you what it sees and what it is doing, and asks for your go-ahead before splitting them: When papers come in. Telling you uses your key. Whatever uses more artificial intelligence (splitting, identifying, analysing) waits for you to click Go ahead, and the button says how much it will send, not what it will cost.
When they are off, the paper goes into the vault and nothing is touched until you ask. They start switched on, and the choice applies to this computer.
How I talk to you
The tone the panel uses to answer:
Casual: on first-name terms, like a trusted colleague.
Neutral: polite, friendly and direct, with no frills.
Formal: polite, in the register of the courts, with no familiarity.
It changes how the panel answers, not what it answers.
Watch the welcome again
Replays the JUDLEX introduction that appears the first time, when you connect the licence and the artificial intelligence. It uses no calls. It is only available with an active licence and a connected key.
The provider and the model
Below the typing box there are two drop-down lists: the artificial intelligence Provider and its Model. The list offers the providers on the firm's list, including any you have added yourself (Add an AI provider that is not on the list); to use another, add it first in Settings → Connections. Providers without a key appear as (not operational) and cannot be chosen.
On top of that, each conversation remembers what was chosen in it. If you switch to a more powerful model in one conversation and back to a cheaper one in another, when you return to the first the chat carries on with the powerful one, without you touching anything and without changing Settings → Connections. Below the drop-down lists it says so: Chosen for this conversation. A new conversation, or one in which you never chose anything, starts with whatever you chose last.
If what a conversation chose can no longer be used (for example, that provider's key was removed), the chat carries on with the one in Settings → Connections and tells you. The choice stays saved for when it can be used again.
If no key has been entered, instead of the drop-down lists you see No connection configured — Settings → Connections: Add your artificial intelligence key.
Reasoning
At the foot of the list of models is Reasoning: how much the model thinks before answering, from No reasoning to Maximum, depending on the levels each model allows. More reasoning usually gives better judgement, but it takes longer and spends more, because reasoning is billed. If a model does not let you choose, the row is greyed out and says Not adjustable. The chosen level also shows in the drop-down itself, next to the model's name.
Daily spend
At the very bottom, Daily spend shows what the firm has spent on artificial intelligence so far today. It is what has already been spent, never a forecast. It counts every call of the day, not just the chat's: what really costs money is reading and analysing exhibits. It resets to zero at midnight.
It is a guide figure: the provider's public rates applied to what has been used, with no mark-up. The provider is the one who charges you, on its invoice. Clicking it opens Settings → History.
Performance
On the right, a small coloured bar labelled Performance shows whether the provider's service is working right now, according to the provider's own status page: green if all is well, amber or red if it has incidents. It says nothing about how good the model is. Clicking it opens Settings → Performance.