Maharshi — Voice Chat

Status: Active product surface (voice sage).
Engineering living plan (Vāgdhenu / uccāraṇa roadmap): monorepo
docs/MAHARSHI_VAGDHENU.md — update that file as work progresses.

Note: Public docs may lag code; the living plan is the source of truth for implementation.

Speak with Maharshi, a wise Vedic sage who listens and responds with a spoken voice. The experience includes an ambient Omkaaram drone, cosmic-themed visuals, and curated side imagery.

Go to: veda-guru.com/maharshi

How it works

  1. Tap the microphone and speak your question
  2. Browser speech recognition converts your voice to text
  3. Text is sent to the Maharshi persona LLM (Sarvam AI)
  4. The sage’s response is spoken aloud via Sarvam AI Text-to-Speech (bulbul:v3)
  5. Side image panels display relevant Vedic imagery from Wikipedia
  6. A subtle Omkaaram drone plays in the background

Features

Feature Detail
Omkaaram Real CC0 Om chant (Freesound, 13.6s looped) with mute toggle
Voice meter Real-time volume bar shows mic sensitivity
Speech-to-text Browser Web Speech API with continuous listening
Text-to-speech Sarvam AI Bulbul v3 (shubh voice, en-IN) with browser fallback
LLM Sarvam AI sarvam-105b with Maharshi persona
Type input Text input also available below the mic
Side imagery Curated Wikipedia thumbnails matching Vedic concepts (50+ mappings)
Markdown Response rendered with marked.js for rich formatting

Design

Full-screen immersive experience:

Tech