Gurukula — Learn roadmap
Updated: 2026-07-26
Live surface: /learn → learning.html
Hero products: Samvāda (/chat) for depth · Maharshi (/maharshi) for voice
Related: Feature Roadmap · Maharshi Voice · monorepo docs/LEARN_ROADMAP.md (canonical engineering copy)
Product principle
Learn is the structured path, not a third hero product.
- Oral-first, progressive — sound and short chunks before walls of text (matches Vedic pāṭha pedagogy).
- Short daily habits beat binge sessions (streaks, 3–21 day micro-paths, paced steps).
- Honest handoffs — never label text chat as Maharshi or voice as Samvāda.
- Tools assist; ācārya / paramparā remain the authority for ritual and initiation.
Samvāda → understanding, Q&A, sources, “I’m stuck on this lesson”
Maharshi → hear, chant, pronunciation, presence
Learn → courses / adhyayas / daily step / progress
Upakaraṇam → short utilities (Lipi, Tithi, Sankalpam, …)
Routing rule (per activity)
| User need |
Route |
| Meaning, clarify quiz miss, “what does this mean?” |
Samvāda with rich lesson context |
| Hear / repeat / japa / mantra practice |
Maharshi (?practice= deep-link) |
| Complete today’s lesson, XP, streak |
Learn |
| Script convert / tithi / sankalpam |
Tools hub |
Ideal adhyaya loop:
Hear → repeat (Maharshi) → read / meaning → quiz → optional Samvāda depth.
Research summary (why this shape)
- Vedic transmission is traditionally oral: progressive pāṭhas (saṃhitā → pada → krama → vikṛti), not print-first.
- Multi-modal practice (audio model + short text + retrieval with feedback) maps well to e-learning.
- Habit evidence (language / scripture apps): 7-day streak strongly predicts completion; short paced plans outperform binge; audio lowers friction.
- Sacred-text plans work best when daily units are finishable (roughly 3–21 day windows for themed paths).
Implications for Veda Guru: fix trust and wiring first; then habit sync; then higher-quality audio—not a Learn rewrite or React migration.
Shipped (2026-07-26)
Phase 0 — Trust and honesty
| Item |
Status |
Notes |
| Relabel Ask Maharshi → Ask in Samvāda |
Done |
Prefilled /chat?ask= with lesson context |
| Dual CTA Hear with Maharshi |
Done |
/maharshi?practice=… + sessionStorage seed |
| Quiz scoring after option shuffle |
Done |
mcq() computes correct index after shuffle |
| Daily reading API wired |
Done |
GET /api/learning/daily-reading (verse + reflection + XP) |
Phase 1 — Habit loop that syncs
| Item |
Status |
Notes |
| Today’s step strip |
Done |
Next incomplete adhyaya + streak + open / Samvāda / Maharshi |
| Soft streak freeze |
Done |
Up to 2 freezes if one day missed |
| Server XP / streak / badges blob |
Done |
GET/PUT /api/learning/gamification (merge local ∪ remote) |
| Adhyaya progress (existing) |
Done |
GET/PUT /api/learning/progress |
Phase 2 — Multi-modal handoffs (light)
| Item |
Status |
Notes |
| Verse-level Hear with Maharshi |
Done |
Links on corpus verse blocks |
| Maharshi practice deep-link |
Done |
Prefill type prompt from ?practice= |
| Web Speech / Sarvam term TTS |
Existing |
In-lesson 🔊; not full Vāgdhenu |
Primary code:
- UI:
veda-guru-ai-ui/learning.html, maharshi.js
- API:
veda-guru-ai-api/main.py — /api/learning/*, learning_gamification table
Commits (approx): UI ddf1e10 · API 16c889c (and follow-ups on main).
Not yet shipped (backlog)
Phase 3 — Path quality
| Item |
Priority |
Notes |
| Curated quiz bank per adhyaya |
High |
Prefer hand-written / server objectives over pure client auto-gen |
| Authored 3–21 day margs (micro-paths) |
High |
Paced daily steps; not only full course lists |
| Notifications / gentle reminders |
Medium |
PWA or email; optional |
| Light social (study buddy / shared plan) |
Low |
After solo streaks prove sticky |
Phase 4 — Voice depth (with Maharshi / Vāgdhenu)
| Item |
Priority |
Notes |
| Mantra practice tool (tools roadmap Phase 2) |
High |
Counter + bank + “hear in Maharshi” |
| Lesson audio models (key ślokas) |
High |
Prefer Maharshi / Vāgdhenu over Web Speech alone |
| Memorisation aids |
Medium |
Spaced recall; still not a Learn platform rewrite |
| Swara-guided recitation |
Later |
Research / Phase E in Vāgdhenu plan; don’t block habit work |
Explicitly out of scope
- Full kundli / commercial jyotiṣa suite
- Replacing guru / initiation with app-only “certification”
- Rewriting Learn in React without an explicit product request
Metrics to watch
| Signal |
Why |
| Daily reading completion rate |
Habit loop health |
| 7-day streak share of active learners |
Retention predictor |
| CTR: lesson → Samvāda vs Maharshi |
Handoff clarity |
| Quiz abandon / perfect-quiz rate |
Assessment quality |
| Cross-device XP continuity (logged-in) |
Gamification API health |
Decision log
| Date |
Decision |
| 2026-07-26 |
Learn stays structured path; Samvāda + Maharshi remain dual heroes |
| 2026-07-26 |
Text clarity → Samvāda; oral practice → Maharshi (honest labels) |
| 2026-07-26 |
Ship trust + habit wiring before large voice/memorisation bets |
| 2026-07-26 |
Gamification stored as per-user JSON blob (merge by max XP / union sets) |
| 2026-07-26 |
Roadmap living doc: this file (docs/LEARN_ROADMAP.md) |
Implementation checklist (for agents)
When changing Learn:
- Keep script order and shared shell/auth conventions (
veda-guru-ai-ui/AGENTS.md).
- Prefer same-origin
/api/* for learning endpoints.
- Any new CTA must state Samvāda or Maharshi correctly—no brand conflation.
- Quizzes must track correct answers by content after shuffle, not hardcoded index
0.
- Update this file when phases ship or priorities change.
- Cross-link tools work in monorepo
docs/TOOLS_ROADMAP.md; chant/uccāraṇa depth in monorepo docs/MAHARSHI_VAGDHENU.md.
