Feature Roadmap
✅ Phase 1 — Foundation (MVP) — Live
- Conversational AI chat interface
- Web search (Tavily → Wikipedia API → DuckDuckGo → Google)
- Sarvam AI LLM integration (
sarvam-105b, 128K context) - Markdown rendering in chat (
**bold**,code, links) - Code block styling with warm Vedic theme (gold border, serif font)
- Source citations with expandable URLs
- Rate limiting and CORS security
- Vedic-themed UI with suggestion chips
- Sage voice mode (mic + TTS response)
- Omkaaram ambient drone (real CC0
om.mp3loop) - Speech-to-text via browser SpeechRecognition
- Sarvam TTS (
bulbul:v3) with browser fallback - Side imagery panels (curated Vedic Wikipedia thumbnails)
- Maharshi voice page with cosmic background, glassmorphism
- Auth system (email/password register + login)
- Session persistence (JWT tokens, 30-day expiry)
- Conversation history (auto-save for logged-in users)
- History panel (browse and reload past conversations)
- Guest mode (no login required to chat)
- Auto-resize textarea with Enter/Shift+Enter
- Stop generation button (AbortController)
- Smart scroll (auto-scroll only when near bottom)
- Follow-up suggestion chips after each reply
- Tavily search as primary web search engine
- Persistent database (Neon Postgres, survives restarts)
- User feedback loop (thumbs up/down)
- Landing page (beautiful dharmic entry with Om theme, CTAs, Vedic quote)
🔜 Phase 2 — Depth
- Custom Vedic knowledge base (BM25 over DharmicData + HuggingFace IndianVedasOriginal)
- Full Rig, Yajur (Shukla Madhyandina + Kanva), Atharva Veda corpus (DharmicData)
- Sama Veda coverage (HuggingFace IndianVedasOriginal, Griffith translation)
- Krishna Yajur Veda coverage (HuggingFace, Keith translation)
- Memory-optimized BM25 (no redundant document token list, precomputed
dl) - Entry caps (
MAX_CORPUS_ENTRIES=10K,HF_MAX_ENTRIES=3K) to stay within 512MB memory limit - Upanishad integration (Kena Upanishad from Project Gutenberg, with expandable downloader for future texts)
- Multiple translation comparisons (results grouped by verse reference across sources;
/api/vedic/translationsendpoint) - Mantra / keyword search (IAST diacritics normalization, Devanagari Unicode tokenization, exact BM25 keyword matching)
- Shruti / Smriti categorization (
categorymetadata on every entry; breakdown in/api/vedic/stats)
Note: Non-Vedic texts (Valmiki Ramayana, Mahabharata, Bhagavad Gita, Ramcharitmanas) were rolled back due to OOM on Render’s 512MB free tier. Vedas-only focus.
✅ Phase 3 — Learning — Live
Living plan: Learn Roadmap (pedagogy, handoffs, shipped phases, backlog).
Engineering canonical copy: monorepodocs/LEARN_ROADMAP.md.
- Guided learning paths with Dharmic terminology (Gurukula, Svadhyaya, Adhyaya, Marg)
- Learning page (
/learn) with path cards, progress tracking, expandable adhyayas - Honest handoffs: Ask in Samvāda (clarity) + Hear with Maharshi (voice practice)
- Landing page CTA: “Start Learning” button
- Learning progress API (GET/PUT
/api/learning/progress) for adhyaya completion - Gamification API (GET/PUT
/api/learning/gamification) for XP / streak / badges sync - Course content (beginner / intermediate / advanced adhyayas)
- Pronunciation guide (IAST + Devanagari terms; Web Speech / Sarvam where available)
- Quiz / self-assessment with correct scoring after option shuffle
- Daily reading via
/api/learning/daily-reading(corpus verse + reflection) - Today’s step strip + soft streak freezes
- Curated quiz bank / authored 3–21 day micro-margs (see Learn Roadmap Phase 3)
- Mantra practice tool + Vāgdhenu-grade lesson audio (Phase 4)
🔜 Phase 4 — Community & Scale
- 10+ Indian and world languages
- Community-contributed notes and commentaries
- Expert-reviewed content badge system
- API for third-party integrations
- Mobile apps (iOS / Android)
🔜 Phase 5 — Advanced
- Voice interaction (ask and receive spoken answers)
- Verse recitation with proper intonation (swara) guidance
- Comparative theology (cross-reference with other traditions)
- AI-generated verse memorisation aids (mnemonics, repetition)
- Offline mode for low-connectivity regions