Conversation metadata

Summaries and tags, written for you

Six months of chat history is only useful if you can find things in it. Nobody has ever gone back and tagged a year of transcripts by hand. ZChat's AI writes a short summary and a handful of tags for each conversation as it ends - so the archive becomes something you can filter and actually learn from.

Tags you can filter by

Each finished conversation gets a few short tags, shown in your session list. Filter by one and you have every chat about that topic - without anyone having categorised anything by hand.

A summary on the record

A couple of lines describing what the customer wanted and how it ended, stored with the conversation. Reviewing yesterday's chats stops meaning reading every transcript in full.

Written after the fact

Summarising happens once the chat is over, in the background. Your agents never wait for it, and a slow model can't hold up a live conversation.

What a tagged archive is actually good for

The value isn't tidiness - it's the questions you can suddenly answer.

The question What tags let you do
"What are people actually contacting us about?" Look at which tags dominate, rather than guessing from the chats you happen to remember.
"Did that release cause support pain?" Filter to the relevant tag and compare the weeks either side of the release.
"What should we document next?" The most common tags are your missing help articles, in priority order.
"Find that chat from last month" Narrow to a tag and skim summaries instead of opening transcripts one by one.

Those most-common tags are also the best candidates for your AI knowledge base.

A starting point, not a verdict

You can always overrule it

Tags are ordinary text on the conversation. Edit them, add your own vocabulary, or bulk-edit across many sessions at once. The AI is doing the work nobody was going to do by hand - it isn't claiming to know your business better than you do.

Fails quietly

If the model is off or unavailable the conversation is stored exactly as normal, just without a summary. Nothing is ever lost waiting on the AI.

Whole transcripts

This feature reads entire conversations, so the local-model option matters more here than anywhere else. See local or hosted.

When does the AI write the summary?
After the conversation ends. It runs as background work, so nobody waits on it - the agent has already moved on to the next chat by the time the summary is written to the record.
Does this slow down live chats?
No. Summarising happens after a chat is over, one at a time, and is deliberately capped so a backlog can never pile up and compete with live conversations for resources.
What happens if the AI is off or fails?
The conversation is stored exactly as it always would be, just without a summary or tags. Summarising is an enhancement to the record, never a condition of saving it - a failed summary is logged and otherwise ignored.
Can I edit or override the tags?
Yes. Tags are ordinary text on the conversation, so you can adjust them, add your own, or bulk-edit tags across many sessions from the dashboard. The AI gives you a starting point rather than the last word.
Does this send old transcripts to an AI vendor?
Only if you have chosen a hosted model. With a local Ollama model the summarising happens on your own server, so transcripts never leave your network - which matters more here than elsewhere, because this feature reads whole conversations.
Owned customer support software

Deploy live chat on your own terms, not on someone else's pricing model.

ZChat gives you the installable server, web dashboard, website widget, and desktop agent tools in one self-hosted product you buy once and keep. Run it on infrastructure you trust and connect AI only if and how you want it.

Deployment

Install on Windows or Linux, behind IIS or Nginx, in a VM, or in Docker if that fits your stack.

Commercial model

One-time purchase, perpetual license, and no monthly per-agent bill attached to growth.

AI Flexibility

Use Ollama locally or connect OpenAI and Anthropic with your own provider accounts.