InFlow
Skill pack

Docs

Search & recall

The point of a second brain is getting the thing back. InFlow search is in the app. An AI pass is optional, and you run it on a file you already have.

In the app

Search is a primary tab. Filters for type, reading status, and tags live in one sheet — not a permanent chip row. A badge shows when a filter is on.

Reading states on clips: Unread, Reading, Done, Reference. Related entries surface from relatedEntryIds. Soft-deleted items are hidden from lists and search until you restore them from Trash.

Saved views as recall

A rule view is a standing question: “unread articles in Reading”, “tasks due this week”. You are not copying those entries into a new pile. See organize.

AI recall (optional)

InFlow does not embed your vault or call a model. If you want a natural-language pass:

  1. Export or copy inflow-db.json from Drive or a backup ZIP. That file is notes metadata — not attachment bytes.
  2. Load the skill pack in the AI tool you already use.
  3. Attach the JSON and ask. The recall prompt searches titles, content, tags, and source titles in the file. It should not guess from general knowledge.

Treat the JSON as sensitive if it contains personal notes. You choose which model sees it. We do not receive that copy.

What recall is not

Not semantic search inside the app. Not a hosted index. Reserved fields aiSummary, aiEntities, embeddingId, lastAiRunAt exist so a future partner layer can land without redesigning the vault. This release does not fill them. Do not store vectors in inflow-db.json.