Write a Testimonial

Share your experience with KP Stellar Astro — an admin reviews it before it appears on the homepage.

Sign in to write a testimonial.

Birth Rectification — test a candidate birth time against known life events

Life Events

Add events with known dates (marriage, job change, accident, etc.). Category must be picked from the list below so it can be scored against KP house rules — free text can't be validated.

Marriage Compatibility — Ashtakoot Guna Milan (36-point match)

The classical Vedic marriage-matching system used across India — compares the boy's and girl's Moon sign & Nakshatra across 8 factors (Varna, Vashya, Tara, Yoni, Graha Maitri, Gana, Bhakoot, Nadi) for a total out of 36 points.

Fill in both birth details, then check compatibility. Use "💾 Save Chart" above to save it to your account.

Daily Panchang?

Rahu Kalam and muhurta timings for any place and date, computed from local sunrise.

Home Muhurta — auspicious date finder for personal events

Uses the same KP promise & dasha-timing engine as Predictive Timing to find upcoming well-matched windows for a personal event, with a traditional Panchanga (Tithi/Yoga/Karana) cross-check on the best day within each window.

This finds windows already promised in your chart for this event, refined to the best matching day. It doesn't override a chart where the event isn't structurally promised.

TSP — Theory of Short Prediction — a quick-question KP check

A faster, narrower alternative to full Horary judgment: give any number from 1 to 2193 and a moment/place, and this checks the number's Sub-Sub Lord against 3 classical KP techniques.

Cuspal Interlinks — S.P. Khullar's Kalamsa & Cuspal Interlinks Theory

A standalone technique, separate from Horary and TSP. For each of the 12 house cusps, its Sub-Sub Lord's Star lord and Sub lord are traced to every other cusp they touch (as Sign, Star, Sub or Sub-Sub lord) — those cusps are "linked", and graded favourable (houses 1,3,5,9,11 from the cusp), neutral (2,6,10) or adverse (4,7,8,12). This covers the book's core cross-cusp linkage mechanism (Chapters 4-5); it does not implement the book's later house-by-house interpretive chapters.

Natal ↔ Horary Check PREMIUM

Judges the SAME question from both your birth chart and a horary chart cast for it, then reports whether the two agree — on the outcome, and (if both promise it) on timing. Classical KP prefers the horary result when they disagree, since its casting moment is known exactly while a birth time can carry some uncertainty.

Question — asked of both charts

Account & Settings

Admin Manual

Step-by-step instructions for every tool in the Admin Panel — click a tool's name in the Admin Panel to jump straight to its page, then come back here anytime for the how-to.

🛡️ Identity & Access

Every account in one place — who they are, what they paid for, and what they may administer. This replaced the separate "User Lookup" tool.

1. Open Admin Panel → Identity & Access. Three tabs.
2. Users — search by name, phone or email, or filter to Staff, Admins, Paying customers or Suspended. Filtering is how you answer "who has admin?", which searching cannot: you can't search for a property without already knowing who holds it.
3. Open any account to manage it: grant credits, change plan, enable 7-Step KP, grant a desktop licence, sign in as them, suspend, or delete.
4. Staff roles on that same card grants operational access. Tick a role, then narrow it with the Scope boxes — a Support Staff agent scoped to desktop sees and answers only desktop tickets. Leave "all" ticked for the whole area.
5. Access Review tab — everyone holding staff access, with risk flags: Never reviewed, Review overdue (90 days), Unscoped, and Suspended but still granted. Click ✓ Certify to record that you have re-checked someone's access. Certifying NEVER changes what they can do — it only records the review. Revoke from the Users tab.
6. Access History tab — every grant and revoke, who did it and when, filtered out of the wider audit log so "how did they get admin?" is answerable.

What staff roles can never do: view logs, change settings or pricing, modify accounts, or grant roles. Those stay admin-only and are enforced in code, not by convention.

🔐 Two-Factor Authentication (staff & admin)

For any account that can administer the app. End users are unaffected — they already sign in without a password.

1. Go to Account → Settings. A Two-Factor Authentication section appears only if your account is staff or admin.
2. Click Set up, scan the QR with any authenticator app (Google Authenticator, Authy, 1Password…), or type the key shown.
3. Enter the 6-digit code to confirm. Save the backup codes it shows you — they appear once and cannot be recovered. Each works a single time.
4. From then on, signing in asks for a code after your usual method. A backup code works instead if you lose the phone.
5. Turn off requires a current code, so a stolen session cannot remove the protection.

Granting someone a role does not lock them out. A newly-promoted admin is prompted to enrol but is not blocked until they do — otherwise they could not reach the screen where they set it up. The static admin password login is exempt: it is the break-glass route, and an authenticator that can be lost must not be able to seal it.

📈 Operations Report

Errors, sign-ins, rate limits, payments and chart volume over time — with CSV export.

1. Open Admin Panel → Operations Report. Pick 7, 30 or 90 days.
2. The top strip totals each series. A number only turns amber or red when it is non-zero — a permanently coloured report stops meaning anything, so quiet is the healthy state.
3. The daily table is newest-first. Each column's bars scale against that column's own peak, so a spike of 8 failed sign-ins stays visible next to 58 charts.
4. ⬇ CSV exports the visible range.

This counts the rare-but-important events. Request-level digging (one user's trace, a specific stack) lives in Google Cloud Logging, which is better at it — see observability/README.md. Counters began when the feature shipped, so earlier history is empty by design, not broken.

🎧 Support Portal (admin side)

Tracked tickets from Premium/Advanced, API and desktop-licence customers.

1. Open the Support Portal from the footer, then the Admin Queue tab (admins only).
2. Filter by status. Open a ticket to read the thread and reply — your reply shows to the customer as "Support Team".
3. Replying to an Open ticket moves it to In Progress automatically. Set Resolved or Closed from the dropdown when done.
4. If a customer replies to a resolved or closed ticket it reopens automatically — that reply is the signal it was not actually fixed.

A Support Staff role holder can work this queue without being an admin, and a scoped one sees only their categories. See Identity & Access above.

🤖 AI Settings

Controls the Claude/OpenAI key used for EVERY user's AI Explanation feature.

1. Open Admin Panel → AI Settings.
2. Toggle AI Explanation feature on — while off, the feature is completely hidden from end users, not just blocked.
3. Decide who pays: leave Free for all users off so users spend their purchased/free credits, or turn it on so every signed-in user gets unlimited AI Explanations funded entirely by your API key.
4. Pick Claude or OpenAI, paste that provider's API key.
5. Choose a model from the dropdown, or click ⟳ Refresh to pull the provider's current live model list (useful whenever a new model is released — no code update needed).
6. Click Save. Takes effect immediately for every user, no redeploy.
7. Clear saved key removes it entirely (AI Explanation becomes unavailable until a new key is saved).

🧩 Model per question pack, and AI models (how the two relate)

Two screens, both under AI Settings, both pick which AI model answers a question — they are NOT two competing systems, one sits on top of the other.

1. Model per question pack (the "Tier models" table) is the FOUNDATION. It sets the DEFAULT model for each price tier (Basic/Standard/Premium/Advanced) — this is what makes a Premium pack answer with a stronger model than a Basic one.
2. AI models is an OVERRIDE layer on top. It lets you pin one specific AI surface — the question classifier (matter/intent), the independent frame judge, AI Explanation, Life Events, or Natal↔Horary Check, each per tier where relevant — to a DIFFERENT model than whatever Tier models already says for that tier, without touching the tier table itself.
3. Every cell on the AI models screen starts on Inherit. Inherit means "use whatever Tier models already says" — an untouched AI models screen changes nothing.
4. You only open AI models when you want ONE specific surface to diverge from its tier's default — e.g. testing whether OpenAI answers Life Events better than Claude for Premium, without changing what Premium's AI Explanation or Natal↔Horary Check use.
5. The question classifier (matter/intent) is the one surface with an extra safety step: a newly picked model cannot answer real customers until it passes a "golden run" check (or you use the explicit, audited override) — see the note on the AI models screen itself.

Short version: Model per question pack answers "what does each tier get by default?" AI models answers "should this ONE surface do something different?" — change the tier table for a broad pricing decision, use AI models for a narrow, surface-specific one.

💰 Pricing settings

Everything money-related: the Premium price, the Buy Questions pack ladder (India + international), and donation settings.

1. Open Admin Panel → Pricing settings — five tabs across the top.
2. Premium Subscription tab: set the yearly ₹ price, Save. This is what unlocks unlimited saved reports.
3. Question Packs tab: edit/add/remove the ₹ credit packs shown to Indian users. Each row needs an id, a label, a quantity, and a ₹ price. Save Packs when done, or Reset to default to go back to the built-in ladder.
4. International Currencies tab: the exchange-rate table ("1 USD = 83 ₹" etc.) used to auto-detect a visitor's currency. Only edit this if a rate has genuinely moved — it's not the source of the actual foreign prices (see next tab).
5. International Packs tab: the EXPLICIT price per currency per tier (not auto-converted) — every cell must clear Razorpay Checkout's real-world minimum charge for that currency (roughly $1/£1/A$1). If you enter something below that floor, Save will reject it with a clear error instead of letting it fail later at Checkout.
6. Payment Method tab: (a) toggle PayPal only for international payments once PayPal is confirmed live on your Razorpay account — until then leave it off so foreign Checkout still shows whatever methods Razorpay currently offers; (b) the Donation QRs section — separately enable/disable the free UPI QR and the Razorpay donation QR, and paste your Razorpay Payment Page link (e.g. razorpay.me/@yourhandle) here.

📊 Analytics

Signups, revenue, feature usage, saved content, donations and traffic — organized into tabs.

1. Open Admin Panel → Analytics. Pick a reporting window (7/30/90 days) from the dropdown top-right — this only affects the "in range" numbers, not the all-time totals.
2. Overview tab: the four headline numbers at a glance.
3. Site Hits / Traffic tab: page-view counts, daily unique visitors, and a Top Locations table (from GeoIP) — paginated, use the Rows/page selector and ‹Prev/Next› to page through it.
4. Users & Signups tab: signup trend chart and a Recent Signups table, showing who signed up and how. It is read-only — an ADMIN badge marks existing admins, but rights are granted and removed from User Lookup, on an account you have deliberately searched for.
5. Revenue tab: total revenue is an INR-equivalent across every currency (converted through the rate table in Pricing), a Revenue by Currency breakdown, a trend chart, and Recent Transactions (each row shows its own native currency, not just ₹).
6. Feature Usage tab: how many AI Explanations have been generated, in-range vs all-time.
7. Saved Content tab: counts of saved Birth/Horary/Financial charts, rectification sessions, and clients.
8. Donations tab: totals and a log of every Razorpay donation-link payment, recorded automatically by the payment webhook — no manual entry.

📬 Feedback

Read Contact Us / feedback form submissions.

1. Open Admin Panel → Feedback — a simple table of every submission, most recent first.
2. Each submission also emails your admin inbox the moment it's sent, so this page is for catching up on history, not the primary notification — you don't need to keep it open.

🛡️ Moderation

Approve/reject proposed Communities, and remove Community/Forum posts that violate guidelines.

1. Open Admin Panel → Moderation. Three tabs: Pending (awaiting your decision), Auto-rejected (the system already filtered these — review in case one was wrongly caught), All Communities (everything live).
2. On a Pending item, click Approve to make it public, or Reject to deny it.
3. On any live post/comment/community, use Delete to remove content that violates guidelines — this is permanent.

🔎 User Lookup

Find one specific account and act on it directly.

1. Open Admin Panel → User Lookup, type a phone number, email, or name, press Enter or click Search.
2. + Credits — manually grant AI-question credits to that account (e.g. as a goodwill gesture or to fix a billing issue).
3. Suspend / Unsuspend — immediately blocks (or restores) every authenticated feature for that account. Reversible, and their data is kept. This is the right control for almost every case: abuse, a payment dispute, an investigation. Prefer it whenever there is any chance you will want the account back.
4. Delete account permanently — erases the account and everything held about it: saved charts, clients, rectification sessions, usage history, AI history, and the questions they asked. Payments are kept but anonymised, because financial records must be retained. This cannot be undone — there is no restore, and the counts shown afterwards are the only record of what was removed. You must type DELETE to confirm. It runs exactly the same erasure as a customer's own "Delete my account", so it is the correct way to action an emailed erasure request. Two accounts it refuses: the one you are signed in as, and any other admin — revoke their admin rights first, which makes removing a privileged account two deliberate steps. Every deletion is written to the Audit Log before it runs, so even a deletion that fails halfway leaves a trace.
5. Enable/Revoke 7-Step KP — grants (or removes) just that one account's access to the Seven Fundamental KP breakdown tab, without making them a full admin. To turn this on for a whole GROUP of users at once instead of one at a time, use Feature Flags → Seven Fundamental KP Visibility instead.
6. Make admin / Revoke admin — grants (or removes) full admin rights for that account: unlimited AI questions and the whole admin panel, including the ability to grant admin to others. This is the only place it can be done. The static admin-password login always remains as a fallback, so revoking admin from an account can never lock you out.
7. Impersonate — opens a 10-minute session logged in AS that account (never elevated to admin, even if the target account happens to be an admin), so you can see exactly what they see when debugging a support issue. A banner stays on screen the whole time with a Return to admin button. Every impersonation is recorded in the Audit Log with both your account and the target account.

📜 Audit Log

A read-only trail of every admin action.

1. Open Admin Panel → Audit Log to see who changed pricing/AI keys/admin rights/bans — and when — most recent first.
2. Nothing here can be edited or deleted from the app itself; only the scheduled retention purge trims very old entries automatically.

💬 Questions Asked

Every real question, and the KP domain it resolved to — anonymous.

1. Open Admin Panel → Questions Asked. Filter by mode (Birth/Horary), by domain, or search the text; the chips above the table show how the current results split across domains.
2. Nobody is identified. The stored row carries a one-way tag instead of an account, and this screen never returns even that — it exists to find questions the engine reads wrongly, not to see what a person asked. Kept 180 days, then purged; erased at once if that person deletes their account. It is disclosed in the Privacy Policy.
3. Read it for three things: a large General Success share means a domain has no rule yet (the same rows Unclassified Questions lists); a question sitting under an obviously wrong domain means a keyword or frame is missing; and a Cusp that looks wrong for the domain is the most serious case — the right rule read from the wrong houses.
4. The Cusp shown is the one actually judged, which is not always the rule's default: a question about someone else rotates the chart onto them, so "will my son get admission?" is Education but judged on the 8th, not the 4th. A cusp shown as e.g. 7? is a row recorded before cusps were logged — that is the rule's default, not a measurement.
5. To act on one person instead — credits, suspension, admin — use User Lookup, not this screen.

❓ Unclassified Questions

Real user questions the classifier couldn't confidently categorize.

1. Whenever a real Horary/AI-Explanation question matches no curated event rule, it falls back to the generic "General Success" domain and emails you an alert immediately.
2. Open Admin Panel → Unclassified Questions to review that same list without digging through your inbox.
3. For a recurring pattern, add a dedicated event rule (or keywords to an existing one, in the underlying event_rules config) so future questions like it route to the correct houses.

🩺 System Health

A live checklist of every optional integration.

1. Open Admin Panel → System Health to see configured/not-configured status for AI, SMTP email, Razorpay, GeoIP, Firebase Auth, and the scheduled purge secret, plus which database backend is active.
2. Click ✉️ Send test email to confirm SMTP actually delivers end-to-end — a green "configured" status only means the env vars are set, not that email is actually arriving.

🔒 Locked-out / Throttled

Manually clear a rate limit or failed-login lockout.

1. Open Admin Panel → Locked-out / Throttled to see every currently-active limit, by IP or account, most active first.
2. "Locked-out" means someone entered the wrong OTP/password too many times on one account; "Throttled" means an IP/account hit a request-frequency cap (e.g. too many OTP sends in a short time).
3. Click Clear on an entry to lift it immediately — useful when a shared office IP or a legitimate user got stuck.

📢 Announcement

A site-wide banner, no deploy required.

1. Open Admin Panel → Announcement, type your message (e.g. "Scheduled maintenance tonight 11pm-12am IST").
2. Pick a Level (Info/Warning/Critical — controls the background color) and, optionally, a font size and custom text color.
3. Check the live Preview box, then click Post banner — it appears immediately at the top of the app for every visitor.
4. Click Clear banner once the issue is resolved. Note: if you change the message text, visitors who already dismissed the old one will see the new one again.

🚦 Feature Flags

Pause an individual tool, or change who sees a gated panel — three tabs.

1. Open Admin Panel → Feature Flags.
2. Tool Kill-Switches tab: turn Birth Chart, Horary, Financial Astrology, Birth Rectification, TSP, or Cuspal Interlinks off for everyone — e.g. during an incident with one tool's dependency. Users see a clean "temporarily unavailable" message instead of a broken page. Flip it back on the moment it's resolved.
3. KP Practitioner Toolkit Visibility tab: check every group (Admin/Astrologers/Basic/Standard/Premium/Advanced) that should see the diagnostics panel on a generated Birth Chart — Admin/Premium/Advanced by default, enforced server-side on every endpoint.
4. Seven Fundamental KP Visibility tab: same idea for the Seven Fundamental KP breakdown tab — Admin only (default) / Admin + Astrologers / Everyone. (For one single account instead of a whole group, use User Lookup's Enable/Revoke 7-Step KP button instead.)

🔮 Manage Astrologers

Approve/reject Personal Astrologer registrations, and control listing visibility.

1. Open Admin Panel → Manage Astrologers — three tabs: Pending, Approved, Rejected.
2. New registrations start Pending and are never publicly visible until you click Approve.
3. Click Reject for submissions that don't belong — an optional reason is kept for your own records only, never shown publicly.
4. Once approved, use the Visible toggle to temporarily hide a listing from the public directory without deleting their registration (e.g. while they're unavailable, or a complaint is being looked into).
5. Delete permanently removes the registration — this cannot be undone.

🕉️ KP Engine

Everything about the engine's behaviour that you are allowed to choose — in four tabs, because these are four unrelated decisions.

1. Open Admin Panel → KP Engine. Every setting shows what it changes, the KP position behind it, and the exact screens it moves. A tab showing a number has that many settings changed from their default.
2. Timing — two settings. Require a Fruitful Star. OFF (default) keeps the significator-based dasha window as the rule, with the Fruitful Star as a tiebreaker and a badge. ON applies the strict school: a sub-period is offered as a window only when its Pratyantar lord is also a Fruitful Star. This hides windows; it does not add promises. Require a Moon-transit confirmation day decides whether a dasha window must also contain a peak transit day before it is offered as a date — dasha says the period is capable, the transit says which day inside it is lit, and requiring both is a 2-of-2 confirmation. Relaxing it surfaces more windows, each resting on one testimony.
3. Chart calculation — the ayanamsa, node and combustion mark a new chart starts from. A user who chooses their own still gets theirs, and a saved chart keeps whatever it was calculated with, so nothing here re-reads an existing chart. Leave it on Krishnamurti + Mean node unless you have a specific reason: KP is worked in that ayanamsa, and the sub boundaries move with it.
4. Horary — the orb within which a Moon close to a sign boundary raises the "question may not be settled" caution. It is a practical caution, labelled as such, and never withholds or changes a verdict. Set it to 0 to stop showing it.
5. Rectification & Muhurta — whether Ruling Planets are cast as a rectification clue (advisory: it ranks candidate birth times, it never picks one), and how many of an elected day's Ruling Planets must also signify the event before Home Muhurta marks the day endorsed. Leave that at 2: one shared planet between any two charts is common enough to mean nothing, and lowering it endorses far more days on thinner evidence.
6. Reference layers — whether each judgment is cross-checked against the indexed KP literature. Advisory only; it never reaches a verdict, and no book, author or passage ever reaches the model or the reader. Switch it off to take the layer out of the path during an incident, without a redeploy.
7. Descriptive layer — the master switch for Descriptive Indications and its ask panel (the Premium tier gate applies separately), and whether a direction is given only when both testimonies agree. Leave the direction setting ON: on the built-in tables 93 of 108 sign × planet combinations disagree, so the relaxed setting is right far less often.
8. Changes take effect on the next chart, and every change is written to the Audit Log with the old panel's settings key kept in step.

Not configurable, deliberately: the cuspal sub lord, the significator tables, the dasha arithmetic and the ephemeris. Those are the validated layer — they are not opinions, and no panel will ever offer to change them.

User Guide

A complete, step-by-step walkthrough of every tool on this site — written for someone who has never used it before. Shown in your selected site language where translated.

Free Public Tools

Standalone tools and pages, free and open to everyone.

Daily breakdown

Serving (step 2)

Decision matrix (step 3)

Latest disagreements

Golden-set candidates

Approving here does not write to tests/golden — that's a separate, explicit commit step.