SourceLace Docs
FAQOpen the app

Searching everything

Search all your connected systems at once for a few words, such as "Corvanta renewal", when you do not know where something is: a document, a chat message, an email, a ticket or a customer record. You get one list of the best results, each with a link that opens it where it lives.

Status: Pilot. Ask your SourceLace contact to turn it on for your organization, and tell support@sourcelace.com how the results look.

What it does

When you ask the assistant (or your own AI app, or an agent) to find something, SourceLace:

  • Searches each system live, with that system's own search. The results are as fresh as the system itself.
  • Uses your own access. Each system is searched with your own sign-in, so you only ever see what that system already lets you see. Systems your groups in SourceLace do not let you use are never searched.
  • Asks every system at the same time. One search takes about as long as the slowest system, up to a few seconds. A system that takes too long is left out of that answer and named, and the answer says it is incomplete, so a slow system never holds up the others.
  • Puts the best results first, and says briefly why each one ranks where it does (such as "#1 in Slack, 2 of 2 words, changed 3 days ago"). You can also ask for the AI to put the top results in a better order; that counts as AI use.
  • Applies your organization's data protection rules to every result before the AI sees it, as for any other answer.

Which systems are searched

Systems that have a search of their own:

System What is searched
Google Drive Files, by name and by the text inside them
SharePoint and OneDrive Files, by name and by the text inside them
Slack and Microsoft Teams Messages in the channels, chats and direct messages you are in
Gmail and Outlook Your mail
Jira Issues
ServiceNow Tickets (incidents, requests, problems and changes)
Zendesk Tickets
Salesforce Accounts, contacts, opportunities, leads and cases
Amazon S3, Azure Storage and Google Cloud Storage Files in the folders your admin allowed, by name and by the text inside them, where the admin chose Names and file contents (how buckets are searched)
Your own search index (OpenSearch, Elasticsearch, Azure AI Search) Whatever your index holds, with links back to the files (Search index)

Other systems (such as databases, data platforms, and buckets searched by names only) are not part of Search everything: ask them directly, or ask a question across systems. A bucket searched by file contents reads only as many files as fit in the time each system has; when it stops early, its results say so and the answer is marked incomplete.

A system is left out of a search when you have not connected it yet (the answer says so: connect it to include it), when your groups give you access to only part of it, or when your admin left it out.

You can narrow a search to some systems, to the days something changed (such as "since 1 September"), or to some kinds of things (files, messages, emails, tickets, issues or records).

Choosing which systems are searched

Admins choose per source. On Connections, open a source and turn Include in Search everything on or off; the Search everything filter there lists the sources that are included, left out or cannot be searched. It is on for every source whose system can be searched. A source left out can still be asked directly. Access by group applies as always: people only search the sources their groups let them use.

Your organization also has limits for Search everything on the Limits page: how long it waits for each system (8 seconds), how many results it asks each system for (10), and how many systems one search asks at the same time (10). Admins may lower them.

In the app

When Search everything is on for your organization, the SourceLace app has a search box at the top of every page.

  1. Click the box, or press / anywhere outside a text field, type a few words (such as "Corvanta renewal") and press Enter.
  2. The results page lists what was found, grouped by system. Each result shows its title, a short extract, the system it comes from, the kind of thing it is (a file, message, email, ticket, issue or record) and the day it changed. Click the title to open it where it lives; it opens in a new tab.
  3. To narrow the results, use the buttons above the list to show one system only, or set Changed on or after, Changed before or Kind and search again.
  4. If a system did not answer in time, the page says so, for example "Showing results from 6 of 7 systems", names the system and offers Try it again. Trying again asks only that system; the results you already have stay. Systems you have not connected yet are listed too, with a link to My connections.
  5. To ask the assistant about some results, tick up to ten of them, type your question (or leave it empty for a summary) and choose Ask about these results. A new chat opens and starts with the results you chose.

Before you search, and when nothing is found, the page lists the systems that are searched for you and why any is not (not connected yet, or access to part of it only), and links back to this page. The results are shown to you only and are not saved; closing the page forgets them.

Admins also see a Search everything column on Connections, which says whether each source is included, left out or cannot be searched, and a filter to list only those. A bucket searched by names only counts as one that cannot be searched; choosing Names and file contents includes it.

Why nothing is copied

SourceLace does not build an index of your systems and keeps no copy of what it finds. Each search asks your systems at the moment you search, with your own access, so:

  • results always follow each system's current permissions: a file someone stopped sharing with you is not found;
  • there is no second copy of your data to protect, keep in step or delete;
  • the results are held only like any other answer, for 30 minutes at most (or less if your admin chose less), and then deleted.

Every system searched is recorded in your organization's audit trail: who searched, which system, the words, and how many results came back. The results themselves are never written to the audit trail.

In your own AI app

The tool is search_everything. It takes the words, and optionally the systems, the days and the kinds. It returns the results, which systems answered, and which did not and why. When the answer is incomplete, your AI app should say so.