Ask one question; three answers run side by side. For each, choose the model (Sol, Luna or Gemini 3.8 Flash), the length, how much the model reasons before answering, how many manual passages each search may return (1–18), and how many searches it may run. Each model searches the manual itself and may search several times; with no limit, Gemini stops at 5. Any panel can also search with FinestIQ’s own search instead of OpenAI’s. Each panel shows an estimated cost.
Each answer runs its own manual searches with its own settings, so retrieved passages and answers can differ. Each run costs one answer from the chosen model plus its manual searches (see the estimate under each answer). Results are saved under the administrator account and may contain errors; verify against the original PDFs.
Upload a complete replacement set. The active edition remains available until you publish the new one.
Download the latest PDFs from the Department Manual page on nyc.gov ↗
When nyc.gov released these PDFs. Officers see this date, so they know how current the answers are.
Publication requires a complete search index AND validated private PDF viewing copies for every file. Older editions are excluded immediately and automatically deleted after seven days. The upload tab should remain open during file transfer; search indexing continues server-side.
Saved on this device.
To ask by voice: tap Tap to talk, speak, then tap Tap when done. Recording stops by itself after one minute.
Review every saved question and full AI answer, including responses without a rating.
Activity is stored privately in Netlify Blobs. The dashboard shows up to the latest 500 records. These are retained until a separate retention or deletion policy is configured.
Estimated from recorded usage and the prices below. Your OpenAI billing page shows the exact amounts.
Officers’ questions, by what answered them: model, search, length, and whether the backup took over. Counted from this update on (older questions weren’t tagged); costs use today’s prices below.
Try a plan: what an officer who uses every question would cost you, at today’s average cost per question.
Only officers’ answers, read-aloud and voice grow with more officers; the Test page and background work (procedure library, example questions, search checks) don’t.
Each Test answers panel, by model, search and reasoning: compare cost and time side by side before changing Answer models & limits.
Enter prices from each provider’s pricing page (OpenAI, Google). Model prices are per 1 million tokens; leave a box empty if unknown. Voice and search defaults are OpenAI’s list prices when this was built; check them.
Everything that applies to the whole app. Settings (gear) holds only your own preferences.
Changes take effect within a minute.
Each model has its own settings. More passages, more searches and more reasoning each cost more and take longer; 10 passages, 2 searches and Medium reasoning gave the best value in tests. OpenAI index: about 0.25¢ per search. FinestIQ search: our own search, no search fee and shorter passages; switch only after comparing both in Browse cards › Search check and on the Test page. Extra high and Max reasoning are on the Test page only (they can run longer than a live answer is allowed).
Each account’s limits per day (they reset at midnight UTC). Leave a box empty to use the default shown.
How FinestIQ works and why. Search, or pick a topic.
Errors officers’ phones hit and errors on FinestIQ’s server (marked Server), reported automatically (never questions or answers). New kinds are also emailed to you daily. Kept 30 days.
Every day FinestIQ copies what can’t rebuild itself: accounts, passkeys, settings, manual editions, example questions, conversations and pins, prices, and saved answers. The last 14 days are kept. Download one now and then to keep a copy outside Netlify; it holds password fingerprints (not passwords), so store it safely.
Officers’ questions the Department Manual didn’t answer well, grouped by why. The same question asked again shows once, with a count. Test page answers and saved-answer matches aren’t included. No AI cost.
For each new question, routing records which kind of answer it would use: Focused (one procedure), Combined (2–3) or Deep (today’s answer). Officers’ answers don’t change unless routing is live, and live is only for your account.
What the routing index knows about each procedure. Cards are for finding procedures only; answers always come from the manual pages.
Type a question the way an officer would. Both searches run on it and show their top 10 passages side by side: the procedure, the piece, a snippet and a score. One OpenAI search costs about 0.25¢; FinestIQ’s costs a fraction of that. The Test page can then answer the same question both ways.
Complete, cited answers are saved automatically. When an officer asks the same question again, or a close rewording, the saved answer is shown instantly with no AI cost. A thumbs-down removes it. Answers that don’t depend on rank are shared by all ranks. When a new manual edition is published, answers whose cited procedures didn’t change are kept; the rest start fresh.
Sol writes questions, each tied to a real procedure in the published edition (a few cents per 10). New questions are drafts: officers only see the ones you approve. Try, edit, replace or remove any that don’t fit, or add your own.
Everyone who can sign in: plans, devices, invite codes and what you changed.
People enter a code under “Have an invite code?” on the sign-in screen to create their own account with the code’s plan. Leave Code empty for a random one.
Suspending and removing take effect at once. Removing an account does not delete previously recorded answers or feedback. Each account can use 2 devices, one signed in at a time; the administrator is never limited.
Official page from the Department Manual, as published on nyc.gov. FinestIQ is an independent reference tool, not affiliated with or endorsed by the City of New York or the NYPD.
This is a thank-you account with extras: every day you get 2 premium answers from our best model, to use whenever you choose, and 20 fast answers.
One look instead of typing your password. Your password keeps working, and you can turn it off in Settings › Sign-in.
By continuing you agree to the Terms and Privacy Policy.
Answers from the Patrol Guide and Administrative Guide, with their pages. Invited testers only; no public sign-ups.
Your questions, answers and feedback are saved for administrator review during this private test. Don’t include personal or case details.
Based on the publicly available Department Manual · nyc.gov
FinestIQ is an independent reference tool, not affiliated with or endorsed by the City of New York or the NYPD.