SFT Brain

Get statistics about your flashcard deck and learning progress

Get the forgetting/memory curve for a single flashcard: the predicted retention (recall probability) over the next ~30 days based on the adaptive (FSRS-lite) memory model. Used by the review widget to draw an inline retention chart for a card.

Get current readiness for one learning target (goal): overall readiness, coverage, mastery, per-area breakdown, and honest pending/stale flags. Returns cached readiness; if any scores are pending or stale, background re-scoring is kicked off.

Get recently added flashcards sorted by creation date

Get review flashcards for a specific deck (or all decks) and the user's deck list for the widget's deck selector. Optionally limit to cards due for review (spaced repetition).

Get your past review history (scores and feedback over time), optionally filtered to a single flashcard. Read-only.

Get a rich learning statistics overview: totals, streaks, average score, due counts, plus optional daily activity, retention curve, and tag performance. Read-only.

List your decks with basic info (id, name, category, description, card count).

Retrieve flashcards from your study deck

List your learning targets (goals) with their latest readiness snapshot (if any) and a breakdown of how many of your cards are aligned (fresh), stale, or still pending. Read-only; does not compute readiness or score cards.

Prepare draft flashcards for the user to review before adding. Returns guidance for drafting high-quality Q&A flashcards; does NOT save anything. After the user approves the drafts, the model should call add_flashcard.

Save a flashcard review with score and feedback. ChatGPT evaluates the user answer and provides score (0-100) and feedback, then this tool saves it to the database and updates the spaced repetition schedule.

Refine an existing learning target (goal) by merging the capabilities found in a new source document (e.g. another job description) into its competency areas. Each capability the document evidences strengthens its area by one; genuinely new capabilities become new areas; near-duplicates fold into existing areas. Does not score cards immediately; existing alignment scores are marked stale for background re-scoring.

Remove a flashcard from a named deck. This only removes the association between the card and the deck; the flashcard itself is NOT deleted and remains in other decks and your library.

Save visualization HTML to an existing flashcard so it can be rendered later in SFT Brain. Replaces any existing visualization for that flashcard.

Return a list of relevant search results for a query

Set the sharing/visibility of one of your decks (by deck id or deck name). Modes: "private" (not shared; clears the share link), "view_only" (shared via link, others can view but NOT clone), "cloneable" (shared via link, others can view AND clone a copy), "public_view_only" (publicly discoverable, view only, not cloneable), "public_cloneable" (publicly discoverable, view and cloneable). For any shared mode a share link is ensured and the full share URL is returned so you can give it to the user.

Set a learning target (goal) to align your flashcards against — a company, role, skill, position (job posting), job description, or free-text intent. Creates the target, extracts its competency profile, and queues your cards for background alignment scoring. Does not score cards immediately; use get_progress and get_aligned_cards afterwards.

Record an organic encounter with an existing flashcard (card_id) or a free-text concept you bumped into while studying something else, raising its organic importance and spreading a similarity-weighted boost to the most similar cards. Does NOT create/merge cards, grade a review, or change the review schedule/FSRS.

Update an existing flashcard