CLEAR, SIMPLE & HONEST

Privacy & entertainment terms

Traditional interpretation, not certainty

Baba, the One Rupee Fortune Teller, is a fictional, lovable North Indian village elder with a playful, respectful manner. Readings include gentle everyday humour, never jokes at a customer’s expense. Sensitive questions receive a compassionate response. Exact-time readings use Prokerala’s calculated Vedic sidereal placements. Date-only readings use Western tropical zodiac and simplified digit-sum numerology. These are traditional interpretive frameworks, not scientifically reliable methods for predicting future events. No outcomes are guaranteed, and readings are not medical, legal, financial or other professional advice. This is a development preview. The illustrated character is 2.5D artwork, not a fully rigged 3D model; mouth motion is audio-reactive, not phoneme-accurate lip-sync.

Payments

Payment availability is shown in checkout. ₹1 is the minimum/default price for a digital-entertainment visit. A visitor may freely choose a higher amount for the same visit; it does not add time, credits, priority or other benefits. Each visit ends after 10 minutes of inactivity. Refreshing keeps a currently active visit. There is no donation, subscription or automatic charge. The verified administrator has free access. UPI and cards are collected by Razorpay’s hosted checkout. This app does not collect or store raw card numbers or CVV. Order/payment references, amount, mode, status, visit activity time and timestamps are stored for verification. Fees are borne by the merchant; the visitor pays only the amount explicitly selected and confirmed. See payment, refund and support policy. Support: rigeshj@gmail.com.

OpenAI readings and Sarvam audio

Character replies and speech are AI-generated, not a real person or cloned voice. The welcomes are approved recordings generated with Sarvam. English and Hindi use Girish devotional voices; Malayalam uses the approved Ashutosh voice. Tapping Hear my reading submits the form and requests a reading and spoken audio. In Talk to Baba, tapping Hear Baba’s reply sends your typed question, selected language and recent saved conversation context to OpenAI to create the reply. The generated reply text is sent to Sarvam to make speech. Do not include private or sensitive details. No microphone or live voice recording occurs. No checkbox is required; notices beside the form explain memory and submission.

Conversation memory on this browser

Starting with this update, the last 20 successful questions and Baba replies are saved in this browser’s local storage for up to 30 days from each exchange. Memory includes your typed question/situation, generated answer, selected language, reading type and time, but not the birth-date form field, payment details or audio/session identifiers. Old questions from before this feature was added cannot be recovered. The newest complete exchanges that fit a 12,000-character context budget accompany a later question so Baba can follow the conversation; older saved exchanges may not be included in each reply. This budget is not a daily reading limit.

Memory works across refreshes, visits and language changes in this browser and website only, not across devices, private-browsing sessions or different site addresses. Anyone sharing this browser profile can view its saved chats. Use Clear conversation before sharing the browser or starting with another person. This removes its saved chat memory and resets the on-screen exchange; it does not delete existing server-side response/speech records described below. Expired exchanges are excluded when memory is read, and pruned from storage when the page is opened or a new exchange is saved; this is not a background erasure service. If browser storage is blocked or full, a notice explains that memory may last only for the current page. Clearing browser data also removes local memory.

Birth time, birthplace and Prokerala

For an exact-time reading, submitting the form sends your birth date, local birth time and the coordinates of your selected birthplace to Prokerala’s astrology API. The server uses the selected city’s IANA time zone and historical clock rules, not your device’s current time zone. Times in ambiguous or nonexistent clock-change intervals are rejected rather than guessed. Place searches run against a bundled GeoNames city directory on this website; they are not sent to an external search provider. Smaller villages may be missing; a nearby selected city gives an approximation. Place data is from GeoNames, licensed under CC BY 4.0, without a guarantee of accuracy.

Prokerala calculates Vedic sidereal planetary positions using Lahiri ayanamsa, including Sun, Moon and ascendant. Only an allowlisted set of calculated placements, your situation/question, selected language and recent conversation context goes to OpenAI for interpretation; the app does not include the exact birth date, time or coordinates in that OpenAI request. Prokerala’s own processing policies apply. A failed or quota-limited chart request is reported honestly; no substitute chart is fabricated. No paid Prokerala plan or automatic credit purchase is enabled.

If you choose “I don’t know my birth time,” no Prokerala chart request is made. The date is used temporarily to calculate the simpler Western zodiac sign and digit-sum number. That result is labelled date-only, with no ascendant, houses, dashas or planetary chart claimed. The date, time and birthplace fields are not saved in the app database or browser conversation memory and are cleared after success. Do not type birth details into the free-text question field, which is saved in browser conversation memory.

Calculated chart facts personalize the traditional interpretation; they do not establish scientifically reliable predictive accuracy. Past and present are based only on what you share. One-week, two-week, one-month and one-year outlooks are interpretive reference windows, not calculated transits or dashas, promised event dates or guaranteed outcomes.

The app does not separately save your original question, situation text or exact birth date in its database. The generated answer may repeat details you supplied, so do not submit sensitive information. It saves the generated response, anonymous session, language and timestamp so only your browser can request speech for that reply. Text requests specify store: false, but OpenAI’s applicable processing and abuse-monitoring retention policies still apply. Selected response or show text is sent to Sarvam when you ask to hear it; generated speech may be cached server-side. A generated reply can repeat details from your question or previous chats, so those details can also reach Sarvam. Sarvam’s applicable processing and retention policies apply. No automatic paid Sarvam credit purchase is enabled; provider quotas can temporarily prevent speech without removing the written reply.

There are no app-imposed daily caps on readings or speech. Owner API charges apply, including conversation context sent with a question, and provider account quotas or service availability may still restrict requests. Voice quality varies across English, Hindi and Malayalam. No microphone recording occurs. Replies are entertainment, not professional advice; text remains available if speech fails.

Cookies and records

A random first-party browser cookie remembers your anonymous session and prevents immediate routine repetition. It expires after 30 days. The unique-IP counter is separate from this cookie and was restarted when IP-based counting was introduced. It counts distinct IP addresses using a keyed HMAC fingerprint; the app does not save raw IP addresses. Shared internet connections can count as one; changing networks or VPNs can count again. These fingerprints are pseudonymous records, not proof of distinct people. If the platform does not supply a usable client address, that request is not counted. Both unique-IP and total-visit statistics are visible only to the verified administrator, not public visitors.

Separate total-visit tracking includes repeat visits, starting with this update. The server stores the existing anonymous cookie identifier, last activity time and a random visit identifier to group a visit until 30 minutes without activity. On return after that gap, another visit is counted. Refreshes and navigation during an active visit do not add another visit. Parallel tabs using the same cookie share that visit; different browsers on one IP can have separate visits. Clearing cookies or using private browsing can start another visit. Real foreground interaction updates the activity record at most once a minute; background timers do not manufacture visits. Separate visit events store only a random visit identifier and start timestamp, not raw IP addresses, names, birth details or questions. Admin totals and daily counts use India time (IST). These are approximate session statistics, not verified people, paid visits or bot-filtered analytics. Historical repeat visits cannot be recovered from the old unique counter. Basic session, language, show, response, usage and payment records are stored server-side. No advertising trackers are included.

Question templates, replay and refinement

Choosing a template and filling its two optional details only prepares an editable question. Nothing is sent until you ask Baba. Avoid names and sensitive details. Previous chats have Hear again and Refine this answer controls. Hear again sends the saved answer to this website for a private playback reference and to Sarvam for speech in its original language; it does not request a new OpenAI answer or add a duplicate browser-memory entry. Refine puts that selected earlier exchange last in the supplied context and lets you edit a follow-up before submitting it to OpenAI. The new exchange follows the same 20-chat, 30-day browser-memory rule. Replay and refinement follow the same visit-access rules as new questions. Provider usage charges apply to the owner.

Faster replies and temporary chart reuse

New replies are delivered in completed text/speech sections while the rest is generated. Each section is stored under the browser-owned reply so only that browser can request its audio. Audio for the next section may be prepared during playback; stopping speech cancels the audio queue, but text can continue. Partial answers are not added to browser conversation memory until generation finishes successfully. Interrupted answers are marked incomplete, not silently accepted as finished; request a new reply if this happens. Original voice models and full-answer instructions remain unchanged.

To avoid recalculating an unchanged chart, calculated chart facts may be reused in one server instance for up to 30 minutes under a browser-scoped, secret-keyed HMAC lookup. The bounded cache is memory-only, not a persistent database or browser cache, and contains chart facts and an opaque key, not raw birth date, time or location fields. It is separate from payment eligibility and does not extend a paid visit. Changed input or another browser has a different lookup; failed calculations are not cached. Cache availability varies between server instances. Timings for chart calculation, first reply section, full reply and speech generation may be logged as durations and cache hit/miss labels, without question text, birth details, identity or credentials.

Marriage Matching

This separate tab compares two adults’ Vedic/Lahiri natal planetary placements calculated by Prokerala, followed by an AI-generated interpretation. Each person must be 18 or over and give permission to use their details. Both dates, local times and selected birthplace coordinates are processed by Prokerala; only calculated chart placements and the optional relationship question are sent to OpenAI. The generated answer goes to Sarvam for requested speech. Exact birth fields are request-local, not saved in the database or browser memory; the generated reply is stored for authorized playback but is not added to browser conversation memory. Avoid names and sensitive information in the optional question. No Ashtakoota/Guna score, dosha calculation or guaranteed relationship verdict is claimed. A failed chart request is reported without inventing a substitute. The separate ₹15 visit and 10-minute inactivity terms are explained in the payment policy; checkout shows current payment availability.

Optional account sign-in

Visitors may sign in through ChatGPT or create a ChatGPT account in that hosted flow. Baba does not collect an account password. The hosting platform supplies a site-specific identity and account email after sign-in. Signing in is optional and does not make a customer an administrator, unlock a paid visit, or sync browser conversation memory or payment access across devices.

Administrator access

Admin features require platform sign-in and an explicit owner email allowlist. Content changes are audited. No secrets are sent to the browser.

Before commercial launch

Public launch requires merchant identity, support/refund contacts, reviewed retention and refund policies, approved gateway setup, and genuine UPI/card test evidence. Do not accept live payments until complete. Hosted OpenAI configuration is separate from this computer’s local settings.