Project3000

🧠 Voice CRM – Voice-to-Structured CRM

Hack-Nation’s 3rd Global AI Hackathon Project Submission

📹 Demonstration and Technical Videos

demo.mp4

tech.mp4

🚀 Overview

Voice CRM eliminates manual data entry in CRM platforms like HubSpot.

Our prototype uses Google Cloud Speech-to-Text and Gemini (Vertex AI) to turn voice notes and recorded calls into structured CRM entries, automatically creating contacts, notes, deals, and tasks.

⚙️ Architecture

Input → Processing → Output

Stage	Description	Technology
🎙️ Input	User provides an audio file (voice note, call, memo)	—
🧩 Speech Recognition	Audio is transcribed into text	Google Cloud Speech-to-Text V2
🧠 Understanding & Automation	Text is semantically parsed to detect intent (create contact, note, deal, etc.)	Gemini 2.5 Flash (Vertex AI)
🔗 CRM Integration	Relevant HubSpot API endpoints are called automatically	HubSpot CRM API
✅ Output	Contact, Note, Deal, or Task created automatically	—

---

🧩 Project Structure

Path	Description
audio_to_transcript/	Handles transcription of remote audio using Google Cloud Speech-to-Text
front/	Frontend interface for audio upload and CRM visualization
server/	Flask backend handling AI inference and CRM automation
.gitignore	Git ignore configuration
LICENSE	Project license
README.md	This documentation
requirements.txt	Python dependencies list

---

🧠 Main Workflow (Voice → CRM)

1. User records or uploads an audio file (.wav or .mp3)
2. The audio_to_transcript module sends it to Google Cloud Speech-to-Text for transcription
3. The transcript is processed by Gemini, which determines the CRM intent and triggers HubSpot API calls to:
   • Create or update a contact
   • Add a note
   • Create a deal
   • Create a task
4. The system returns a concise confirmation message.

---

🧰 Tech Stack

Category	Tool
AI / NLP	Google Gemini 2.5 Flash (Vertex AI)
Speech Recognition	Google Cloud Speech-to-Text V2
Backend	Flask (Python)
Frontend	React + Vite (JavaScript)
CRM	HubSpot API v3 / v4
Cloud	Google Cloud Run / Cloud Functions

---

▶️ Run Locally

The project has four main parts:

1. Speech-to-text API (Google Cloud)
2. Generative AI agentic API for text to HubSpot commands (Gemini)
3. The backend (Flask + Python) inside the server/ folder
4. The frontend (React + Vite) inside the front/ folder
5. HubSpot API

Obs.: For parts 1, 2, and 5, provide the required API credentials using your preferred method (environment variables, external JSON file, etc.).
Follow the steps below to prepare your environment and run both components:

---

🧩 1. Backend Setup (Flask API)

# Install Python dependencies (from the root folder)
pip install -r requirements.txt

# Start the Flask backend
python -m server.app

💻 2. Frontend Setup (React + Vite)

# Navigate to the frontend folder
cd front

# 1. Install Node.js dependencies
npm install      # or simply: npm i

# 2. Start the local development server
npm run dev

By default, Vite will start the frontend at http://localhost:8080/. Make sure the Flask backend is running before testing the interface.

---

👥 Team

Project developed for Hack-Nation AI MIT — Voice-to-Structured CRM. Built with ❤️ using Google Cloud and HubSpot APIs.