Web search restrictions after internal content (#1074)

Co-authored-by: Thorsten Sommer <SommerEngineering@users.noreply.github.com>

app/MindWork AI Studio/Plugins/configuration/plugin.lua

@@ -987,7 +987,30 @@ CONFIG["SETTINGS"] = {}
 --                         separately. Allowed private hosts require a provider with HIGH
 --                         confidence. AI Studio only tries the current user's operating-system
 --                         sign-in for explicitly allowed HTTPS targets when that provider
---                         requirement is met, and it never reuses browser cookies.
+--                         requirement is met, and it never reuses browser cookies. Allowed
+--                         private hosts count as services configured in AI Studio: a chat which
+--                         read internal content may still read them, see below.
+--   internalPageOutboundDataRestriction
+--                         Where a chat may send data once it read internal content: a page with
+--                         a private or VPN address, a page of the Confluence wiki configured for
+--                         search_confluence, whatever its address, or a search of that wiki. Only
+--                         the page reached after all redirects counts. A web search or an address
+--                         the AI writes could otherwise carry internal content to third parties.
+--                         The same level applies to the search_confluence tool. A chat keeps the
+--                         level it had when it read the content. Allowed values are:
+--                           ONLY_CONFIGURED_SERVICES -> the chat reads only pages of the wiki and of
+--                                                       the allowed private hosts, and no web
+--                                                       search runs.
+--                           ONLY_LINKS_FROM_CHAT     -> the chat also reads pages whose addresses
+--                                                       stand in the chat, written by the user or
+--                                                       returned by a tool, exactly as they stand
+--                                                       there. No web search runs, and the AI
+--                                                       cannot choose addresses outside the wiki
+--                                                       and the allowed private hosts. This is the
+--                                                       default.
+--                           UNRESTRICTED             -> no restriction beyond the ones of each tool.
+--                         Lock it to ONLY_CONFIGURED_SERVICES when the AI must not even choose which
+--                         of the given links to follow.
 --
 -- Field names of the Search Confluence tool, which supports Confluence Data Center. Confluence
 -- Cloud is not supported yet.
@@ -1000,6 +1023,7 @@ CONFIG["SETTINGS"] = {}
 --   timeoutSeconds   Search request timeout in seconds, at most 120. Default: 30.
 -- Selecting search_confluence also selects read_web_page, which opens the pages found. If your
 -- wiki has a private or VPN address, add its host to read_web_page.allowedPrivateHosts as well.
+-- Where a chat may send data after a search is set by read_web_page.internalPageOutboundDataRestriction.
 --
 -- CONFIG["SETTINGS"]["DataTools.LockedToolSettings"] = {
 --     ["web_search.searxng.baseUrl"] = "https://searxng.example.org/",
@@ -1008,6 +1032,7 @@ CONFIG["SETTINGS"] = {}
 --     ["web_search.tavily.apiKey"] = "ENC:v2:<base64-encoded encrypted data>",
 --     ["read_web_page.freeAddressChoice"] = "OFF",
 --     ["read_web_page.allowedPrivateHosts"] = "example.org, *.example.org",
+--     ["read_web_page.internalPageOutboundDataRestriction"] = "ONLY_LINKS_FROM_CHAT",
 --     ["search_confluence.baseUrl"] = "https://wiki.example.org/confluence/"
 -- }
 --