> ## Documentation Index > Fetch the complete documentation index at: https://docs.delphina.ai/llms.txt > Use this file to discover all available pages before exploring further. # Search > Find past chats and data apps in a workspace by keyword Search finds chats and data apps in your current workspace by keyword: what was asked, what the agent answered, the artifacts and files it produced, and what your data apps contain. Results from chats and data apps appear together in one list, best match first. Search is rolling out workspace by workspace. If you don't see **Search** in the sidebar, it isn't on for your workspace yet. [Let us know](#feedback) if you'd like it. ## Open search * Click **Search** at the top of the sidebar, or * Press ⌘ K on Mac or Ctrl K on Windows and Linux. On the Inbox page, that shortcut opens the Inbox's own command menu instead. Type at least two characters. Results update as you type. Use ↑ and ↓ to move through them, Enter to open one, and Esc to close search. * A **chat** result opens the chat at the turn that matched. A match on the chat's title opens the chat itself. * A **data app** result opens the app. A match in an unpublished draft, marked **Draft**, opens the app's editor. ## Tips for better results * **Use specific words.** Search matches words, not meaning: "net revenue retention" finds those words, not every chat about retention metrics. * **Word forms match.** "churned" also finds "churn" and "churning". Search understands English, Spanish, and Portuguese word forms, and ignores accents. * **The last word matches as a prefix,** so "cohort ret" finds "cohort retention". * **Put a phrase in quotes** to match it exactly: `"weekly active users"`. * **Exclude a word** with a minus sign: `revenue -forecast`. * **Search for table and column names** with **Include SQL & code** on (see below). `fct_orders` matches exactly. * **Names and titles rank first.** A data app whose name matches, or a chat whose title matches, appears above matches in the body of a chat. Among other results, more relevant and more recent ones rank higher. ## Filters The row under the search box narrows results: | Filter | What it does | | - | - | | **All / Mine** | **Mine** shows only chats and data apps you own. | | **All / Chats / Data apps** | Search one kind, or both. | | **Include SQL & code** | Also search the queries and code the agent ran in chats, and data apps' source code. Off by default. Turn it on to find where a table or column was used. | ## What's searchable **In chats:** * Your prompts and the agent's answers * Chat titles and the chat owner's name * The text of the agent's reports and notes, and the column names of tables it showed (not the rows) * The names of files you attached and files the agent created * With **Include SQL & code**: the SQL, Python, and other tool calls the agent ran **In data apps:** * The app's name, description, and owner's name * The steps' prompts and descriptions * Report templates, such as the Markdown a report is built from * With **Include SQL & code**: the app's source files, such as SQL, Python, `App.tsx`, and pipeline configuration. Chart titles and labels written in the app's code are found this way too. Both the published version of a data app and its unpublished draft are searchable. Which drafts you see depends on your role (see [Who sees what](#who-sees-what)). ## What's not searchable * **Query results and data:** the rows a query returned, tables, and datasets * **Rendered reports and dashboards,** including titles or numbers computed when the app runs. These change on every refresh. * **The contents of files.** File names are searchable, but not what's inside them. * **Images** * **Chats and data apps in other workspaces.** Search covers the workspace you're in. * **Deleted chats and data apps** * **Background data app refreshes and branch builds.** These chats are never indexed. * **A turn that's still running.** It becomes searchable when it finishes. * **Very old code matches:** with **Include SQL & code** on, only the most recent 2,000 matching tool calls in a workspace are ranked. Use a more specific query to find an older one. ## How quickly new work shows up * **Chats:** a turn is usually searchable within a few seconds of finishing. Renaming a chat updates its title in search just as quickly. In the rare case indexing fails, the turn is picked up overnight. * **Data apps:** a new app, a release, and changes to its name, description, or sharing are usually searchable within a few seconds. Edits to an unpublished draft can take until the next day to show up. * **Chats from before search was turned on** are added in the background when search is enabled for a workspace. A large workspace can take several hours to finish. * **Deleting a chat or data app, or making it private,** removes it from other people's results right away. If something you expect still doesn't appear after a few minutes, [tell us](#feedback). ## Who sees what Search shows you the same chats and data apps you can browse, never more. What that includes depends on your role in the workspace: | | Developer | Explorer or Viewer | | - | - | - | | **Chats** | Every chat in the workspace, except other people's private chats and chats run by Delphina's automated agents | Your own chats | | **Data apps** | Every app except private apps you don't own or edit (published private apps are included) | Published apps, plus apps you own, edit, or starred | | **Unpublished drafts** | Drafts of non-private apps, and of apps you own or edit | Drafts of apps you own or edit | Private chats are only ever shown to their owner. Making a chat or data app private, or deleting it, removes it from other people's results immediately. If your role changes, your results change with it the next time you search. ## Feedback Search is new, and we're tuning what it finds and how it ranks results. If a search didn't find what you expected, ranked something oddly, or was slow, email [support@delphina.ai](mailto:support@delphina.ai) with: * what you searched for, * what you expected to find, such as a link to the chat or data app, * which filters were on. We read every report, and it directly shapes how search works. This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.