Account and data
Upload reusable source files, follow parsing states, choose personal, team, or Brand scope, search grounded content, and maintain the memory graph.
3 min readUpdated
Something missing from this page? Email help@tap.co
Memories combines two reusable context types:
Both can use personal, team, or Brand scope. Scope controls where the context resurfaces; it does not publish the content to a supplier.
Memories opens as a full workspace with a graph and searchable list. Use the unified search to find a memory or Brand across both views. Select a graph node to inspect its note and relationships, then return to the graph without losing the broader context.
Graph links organize recall. They do not merge content, grant access, or prove that two facts are equivalent.
Open Memories and use the file panel, or attach a file in an Agent conversation. Choose the destination:
Tap checks whether an identical stored file can be reused before uploading another copy. A reused file keeps one authoritative parse instead of duplicating work.
A file moves through:
Do not assume the Agent read a file that is still uploading, parsing, or in error. Use the available retry or reprocess action. Legacy spreadsheet files can be recovered into a new parsing attempt when Tap detects the old state.
Open a file to review its title, summary, tags, scope, Brand, size, type, and processing details. Download the original when you need to compare Tap's summary with the source.
Ask the Agent to search saved files before relying on them. It should cite the source file and separate extracted facts from inference. A citation proves where the fact came from, not that the source is current or correct.
Changing a file scope changes who can retrieve it. Moving a file to a Brand requires choosing the exact Brand.
Open a memory to edit its title and tags. Link related memories to build the visible graph, and unlink a relationship that no longer helps. Links organize recall; they do not merge note content.
Ask the Agent to replace an obsolete active note instead of creating a duplicate. Forgetting a memory permanently deletes that saved note and requires an explicit request. It does not delete an uploaded source file with similar wording.