The KlicForge Inbox, and import a website as knowledge

New

The KlicForge Inbox — a new app for handling live conversations from your phone. It opens on three queues: Needs Attention for conversations where an agent has asked for a human, Live for everything currently running, and Mine for the ones you have taken over. Add it to your home screen from inbox.klicforge.ai and it behaves like any other app on the phone. See the KlicForge Inbox.
Take over a conversation, then hand it back — step in as a human, reply directly, then either return the conversation to the agent or close it. If a colleague is already handling one, you can see who, and take it over from them if you need to. See taking over a conversation.
Get told the moment an agent asks for a human — turn on notifications and the Inbox alerts you when a conversation needs a person. Tapping the alert opens that exact conversation, ready to take over. See getting alerted.
One phone, every workspace and account — tap the workspace name at the top of the queue to move between the workspaces you belong to, and sign in with a second account to swap between those too, without signing out; tapping an alert switches to the right account and workspace first, so a handover in one workspace never lands you in another. Signing out on a device signs out every account on it, so switch rather than sign out. See working across workspaces and accounts.
Import a website as knowledge — paste your website address and KlicForge reads it, then writes a set of tidy knowledge documents from what it found. You choose how many pages to import before anything runs, and you review every document — editing, keeping or discarding each one — before it becomes agent knowledge. Available on the Startup plan and above, from Knowledge Library → Add Source → Import a website. See importing a website.
Scanning a site is free — finding out how big your website is, and what importing it would cost, does not touch your allowance. Only pages actually imported are charged, at one interaction per page.
Generated images are kept — an image an agent creates is now saved to your workspace asset library instead of expiring, so the agent can send it again later in the same conversation and you can find it afterwards. See generated images.
A Generated view on the Assets page — the default Library view still shows the files you uploaded; switch to Generated or All to see what your agents produced. Generated images are kept out of the default view because they accumulate quickly. See managing assets.
Share with all conversations — owners and admins can promote a generated image, or a file a customer sent, into a reusable workspace file. See which conversations an asset appears in.

Fixed

Turning on browser notifications while in one workspace silently stopped them arriving from any other workspace on that same device. Notifications now cover every workspace you belong to; a device affected by this is re-registered for all of them the next time you turn notifications on, or when you sign in to the Inbox.
Switching workspaces kept the permissions of the previous one for the rest of the session, so an owner of one workspace could still see owner-only actions in a workspace where they are a member. Permissions are now read from the workspace you are actually in. The same applied to managing knowledge sources, where someone removed from a workspace could keep acting with their former permissions until they signed out.
A file a customer sent in chat was saved as a shared workspace file, so an agent could offer one customer's document — a payment proof or an ID — to another. Such files are now kept to the customer who sent them. Existing files stay shared; open any that should be private and delete or re-scope them.
Deleting or merging a contact failed if an agent had ever generated an image for them. Their files now stay in the library and simply stop being offered in chat.
An agent that could see a file but was not allowed to send it told the customer the file did not exist, then kept retrying. It now says it cannot send that one.
A website import could sit on Organising knowledge for twenty minutes and then fail without producing a single document. Imports now finish in the time they should, and one that genuinely cannot finish stops promptly instead of running on. If an import of yours ended this way, start it again — pages that failed were never charged.
Content from image-based PDFs — scanned documents, brochures, menus — could come through incomplete or as plain unformatted text, leaving an agent with a poorer version of a file you had added. Re-upload any such file added recently that an agent does not answer well from.

Changed

Generated images count toward your workspace storage. They previously expired on their own; now they stay until you delete them. Worth a look at Settings → Usage if you generate images heavily.
An agent only offers a customer the files from their own conversation. A generated image, or a document someone sent in chat, is no longer something the agent can surface to a different customer. Files you upload to the library are unaffected and remain available to everyone the agent talks to. See which conversations an asset appears in.
On the website widget, those files stay with the conversation rather than the person. On WhatsApp and Telegram the platform tells us who is writing, so a file follows that customer into later conversations; on the widget a visitor can type any email address, so a file is available for the rest of that chat and not afterwards. See which conversations an asset appears in.
A generated image is named for what it depicts, not for what the customer typed. The request itself stays in the conversation, where access is limited to owners and admins and every view is recorded.
Turning an agent off now removes its chat bubble from your website, rather than leaving a bubble that opens onto a maintenance message. An agent embedded directly inside a page still shows that message in the space it occupies, so your layout does not collapse.