the Geervanee Sanskrit AI tools suite · AI शिक्षकः is home
A split you're handed teaches nothing.
Ask any chat tool for a śloka's word-split and it obliges instantly — no reason to withhold it, no idea what you're studying. ai.geervanee.in withholds the answer on purpose and walks you to it, one verified hint at a time. AI should sharpen the mind, not spare it the work — why we built it this way.
Tools साधनानि
Four tools, one Geervanee login
AI शिक्षकः, the tutor, is where you start — a patient teacher who walks you to every answer instead of handing it over. Beside it, संशोधकः helps you actually speak Sanskrit. Two more are on the way.
शिक्षकः AI शिक्षकः — the tutor
Guided, never-answering practice across sandhi, vibhakti / vachana, śabda-rūpa, dhātu-rūpa, and the five kinds of complex-word formation — every fact checked against a real Pāṇinian derivation engine before it reaches you.
संशोधकः Samsodhaka — the sentence coach
Practice framing and speaking your own Sanskrit sentences. It never writes the sentence for you — it nudges, one hint at a time, and leans on शिक्षकः for every grammatical call so the correction is always sound.
वाचकः Vachaka — the reciter
Reads a śloka or passage aloud as a metrically-correct chant, built on the Vāgdhenu open chant system. Most of the library batch-rendered once, then served as static audio.
लेखकः Lekhaka — the scribe
Record or upload a Sanskrit recitation and get a clean Devanāgarī transcript — line breaks kept, daṇḍas at verse ends. Document export (PDF / Word / Kindle) is next.
Method रीतिः
Guided questioning, never a direct answer
The tutor never states a śloka's split or translation as its first response. It runs the same Socratic loop for every grammar concept — with one deliberate exception for dictionary lookups.
// generic chat tool
Instant answer
गजाननम् = गज + आननम्. Done in one turn. Nothing about the a + ā → ā rule sticks, and the next compound is just as opaque as this one was.
// ai.geervanee.in
Derived answer
A simpler parallel word, a question about where it joins, then your own re-application to the original. You leave able to split the next one unaided.
Present the śloka; you attempt a split
The exchange starts from your work, not the tutor's.
If correct — confirm, and say why
The rule that made it correct is named, so the win reinforces something reusable instead of just moving on.
If incorrect — offer a structurally similar word
Same underlying sandhi rule, simpler or more familiar context. You split that one first.
Re-apply to the original
Once the simpler example lands, you carry what you noticed back to the śloka that stumped you.
Last resort — name the specific rule or boundary
Only after repeated rounds, and only ever the rule to look at — never the finished answer, unless every rung above has failed.
Pattern-transfer by analogy
Form and derivation questions can't always use "here's a similar word to split". Asked for the चतुर्थी विभक्ति of गुरु, the tutor doesn't answer गुरवे. It shows that गुरु and शम्भु share one उ-stem paradigm, that शम्भु becomes शम्भवे, and asks what the same ending does to गुरु — so you derive गुरवे yourself.
Every fact — the paradigm match, the sound change — comes from the correctness layer, not the model's guess.
Scope व्याप्तिः
The full grammar a learner works through — not sandhi alone
Every concept below follows the same never-answering approach. Only the last one, a dictionary lookup, is allowed to give the meaning outright.
Where two sounds meet, what they become, and how to run it backwards.
Which case ending is doing the work in a form, singular through plural.
Noun and verb paradigms, reasoned about by class rather than memorised flat.
Splitting a word into base and affix, and what each part contributes to the meaning.
An unfamiliar word's meaning is given directly — then the tutor may invite you to name its grammatical category or attempt its derivation.
Correctness शुद्धिः
The model phrases the conversation. It doesn't decide what's true.
A generic LLM can state a wrong split with total confidence and no way for a learner to know. Every fact in a hint — the "correct" split, the example words — is drawn from a curated grammar reference first.
Pricing मूल्यम्
Bring your own key, and you're set.
A very nominal platform fee keeps the service running. Add your own free Google API key and that's it — no payment for tokens, no daily bill. Just open AI शिक्षकः and learn, every day.
Yearly platform fee — that's the whole cost
Very nominal, once a year. It covers the servers, the Pāṇinian grammar engine and upkeep, so the tutor stays available to every learner.
Add your API key — and you're good to go
Create a free key at Google AI Studio (we show you exactly how), paste it into your profile once, and your practice runs on Google's free tier. No token payments, ever. Encrypted here, one-click to remove.
Rather not manage a key? Metered plan
Practice runs on Geervanee's own AI, metered and capped so a normal day of learning stays affordable.
Try it before you decide
New here? You get a handful of free turns on us to feel how the tutor works — no key, no fee, nothing to set up.
Your key never leaves ai.geervanee.in. It's encrypted at rest, used only to run your own practice, and you can remove it in one click at any time.
One tutor एकः शिक्षकः
Every Geervanee platform, one AI शिक्षकः
The same tutor, the same verified grammar, wherever you learn with Geervanee — geervanee.in, geervanee.com, the Samsad app, the Geervanee mobile app, and Kshanasah. One Geervanee ID, one place the answers are checked.
AI शिक्षकः is a focused practice tutor and a small Sanskrit AI-tools suite — not a general-purpose chatbot or translator. It does one thing: guide you to the answer.
Roadmap आगामि
What deepens the gap over time
The MVP is the never-answering pedagogy, the verified correctness layer, and a plain data stance. These build on it.
later
Content-linked practice
A session picks up the exact śloka from the video or course you're already studying.
later
Persistent skill tracking
Which sandhi rules you've actually mastered, carried across sessions into Geervanee's performance tracking.
later
Purpose-built interaction
Devanagari + IAST rendering, audio pronunciation, tap-to-mark word boundaries — not a plain text box.
later
Institutional insight
An aggregated, non-individual view of which rules a class collectively struggles with.
Success लक्ष्यम्
What the platform is measured against
Unaided completion
Share of hint-ladder sessions where the student reaches a correct split without ever being shown the answer.
Cache-hit rate
Most repeat queries served with zero API call.
Real cost / student / day
Tracked average held at or under the affordability target.
Tutor retention
Day-over-day return to the tutor feature specifically, separate from overall app retention.