{
  "id": 1077395,
  "title": "Saathi: building a Kirana store voice agent in 10 days with Murf Falcon published: true tags: voiceai, python, webdev, beginners",
  "url": "https://urgent.news/2026/08/15/saathi-building-a-kirana-store-voice-agent-in-10-days-with-murf",
  "topic": "ai",
  "section": "AI",
  "published": "2026-08-15T16:27:32.000Z",
  "source": {
    "name": "Dev.to",
    "slug": "dev-to",
    "url": "https://dev.to/greeva_jagani_c97a0f70e05/saathi-building-a-kirana-store-voice-agent-in-10-days-with-murf-falcon-published-true-tags-22bm"
  },
  "original_language": "en",
  "account": "A voice assistant was created to help local grocery stores in India called Saathi. These stores rely heavily on phone calls and memory, which can be unreliable for customers who prefer talking over typing. The assistant is designed to accommodate customers who are more comfortable speaking in Hinglish (a mix of Hindi and English) and for store owners who need help managing routine calls without pretending to be human.\n\nThe assistant can engage in natural conversations in English, Hindi, or Hinglish, greet returning customers by name, recall their last order, check live stock before promising anything, place orders, confirm them back to the customer, and escalate issues to the store owner when necessary. It also proactively calls customers to confirm orders, handles payment disputes and refunds, and logs outcomes for analytics.\n\nThe assistant's core functionality involves real-time speech-to-text conversion, reasoning, and text-to-speech conversion using Deepgram STT and Murf Falcon TTS. It uses Google Gemini LLM as its reasoning engine, equipped with persona, guardrails, and tool-calling logic. The assistant's data is stored in a SQLite database, and it can transfer conversations to a specialist agent when needed.\n\nThe assistant is built to work seamlessly over a LiveKit room, accommodating both browser and phone channel inputs. It prioritizes accuracy and transparency in its interactions, ensuring that customers are always aware of when they are talking to an automated system.",
  "summary": "How I built a Hinglish-speaking voice assistant for local grocery stores — and what ten days of #VoiceForBharat taught me about voice AI. The problem and the users Most local Kirana (neighbourhood grocery) stores in India still run on phone calls and memory. The shopkeeper remembers what Ramesh usually orders, but there's no system behind that memory — no record of preferences, no way to check…",
  "key_points": [
    "Saathi voice agent built in 10 days for Indian kirana stores",
    "Uses Deepgram STT, Murf Falcon TTS, and Gemini LLM for speech processing",
    "Handles conversations in Hinglish, English, Hindi; integrates with SQLite DB"
  ],
  "editors_take": null,
  "illustration": null,
  "coverage": {
    "outlets": 2,
    "also_reported_by": [
      {
        "outlet": "Dev.to",
        "title": "Building MoneyBuddy: From a Simple Voice Agent to a Multi-Agent AI System in 10 Days",
        "url": "https://urgent.news/2026/08/15/building-moneybuddy-from-a-simple-voice-agent-to-a-multi-agent-ai",
        "published": "2026-08-15T16:53:01.000Z"
      }
    ]
  },
  "ai_generated": true,
  "disclaimer": "Summaries, key points and the editor’s take are written by software from other outlets’ reporting and may contain errors — always check the linked original."
}