How to search your chat history in Whizi

The short answer

Yes, Whizi searches your chat history, in two layers from one box. The sidebar search box filters chat titles instantly as you type, and once your query reaches three characters it also searches your message text on the server, listing those chats under a Found in messages header with a snippet of the text around the match.

What the search box reaches:

What you search on	Matched
The chat title	Yes, instantly, filtered in your browser as you type
Text inside a message	Yes, from three characters, searched on the server
Chats filed inside a project	Yes, one search covers projects and the ungrouped list together
Chats the sidebar has not loaded yet	Yes, message search runs across your whole history

Message search needs a signed-in account and a connection. When the request cannot run, offline for example, the title filter keeps working on its own.

How the two layers divide the work

Title matching runs in your browser, which is why it is instant and works at any query length. Message search is a server query, and Whizi is deliberately sparing with it: nothing is sent until the query is at least three characters, the request waits out a 450 millisecond pause in your typing, and a query you have already run this session is answered from a cache rather than sent again.

A message search returns up to 20 matching chats. Each one renders under the Found in messages header with a snippet of the text around the match, so you can tell the results apart before opening any of them. While the request runs, the sidebar shows Searching messages…, and a query with no title or message matches shows No chats match your search.

Searching conversations is rate limited to 20 requests per minute. Normal typing never gets near that, because the pause before sending and the session cache keep the request count low.

One detail that matters on an old account: the sidebar loads your chat list in pages, but message search runs across your whole history on the server. A chat you have not scrolled to in months still comes back as a match.

One search covers every project at once

Opening a project shows only that project's chats. Typing a search removes that filter: matching cuts across projects and the ungrouped list together, so a chat is findable whichever folder it sits in, and you do not have to remember where you filed it.

Projects are otherwise a hard partition: a chat belongs to at most one project, which is what makes the "everything else" view in the sidebar work. Search is the one place where that partition is ignored. Renaming, deleting and the project structure around it are covered in organize your conversations.

Rename the chats you will want back

Titles are still the sharpest handle. A title match is instant, works offline, and works on a query too short for message search, so a chat named something you would actually type surfaces before anything else.

The rules the rename enforces, in the order you hit them:

• A blank name is refused with Chat name cannot be empty.
• Over 100 characters is refused in the browser with Chat name is too long (max 100 characters).
• The backend cap on the same field is 120 characters, and it answers Conversation titles must be between 1 and 120 characters.
• If the call itself fails you get Failed to rename chat. Please try again.

Two habits do most of the work. Put the nouns you would search for into the title, meaning the client, the file, the feature or the decision, not the verb. And rename at the end of a session rather than the start, because that is the point at which you know what the conversation turned out to be about.

Imported chats are searched like everything else

Whizi imports from ChatGPT and Claude, and an imported conversation carries its original title across, trimmed to 120 characters. A conversation with no usable title in the export becomes Imported chat. Each importer has its own page, import chats from ChatGPT and import chats from Claude, covering the export file it expects and what carries across.

Once imported, those are ordinary chats, so both layers reach them: their titles filter instantly, and their message text is searched from three characters like anything you wrote in Whizi. A large import that leaves a run of rows all named Imported chat is therefore still searchable by what is inside them, though renaming the handful you expect to want again remains the faster route back.

Imports are also idempotent, keyed by account plus source plus source id, so re-running the same export does not duplicate the rows you have already renamed. If the import itself is failing, the message you are seeing is covered in chat import errors.

When search is not the right tool

A query under three characters never leaves your browser, so a two character fragment, an initialism like Q3, only matches titles. If the fragment you remember is that short, search for a longer phrase near it instead, or rename the chat once you find it.

For searching outside the product, the account export is the route. Export Data sits in Settings under Account and returns your account as a single downloadable JSON file covering chats and messages, along with memory, projects and file metadata, media, share links, subscription and referral profile. Open that file in any editor and search it there. Export is rate limited to 5 requests per minute, because it reads everything the account owns. The mechanics are in exporting your chats and files.