🧠 Voice-to-Vector AI System

A Flask-based AI system that transforms voice notes into structured insights and stores them in a vector database for semantic retrieval.
This project demonstrates an end-to-end intelligence pipeline — from voice recording to actionable knowledge — integrating transcription, LLM reasoning, and vector storage.

---

🚀 Overview

The Voice-to-Vector system automates the process of turning spoken thoughts into searchable, structured data.

Workflow Summary:

1. Voice Note Input → uploaded or recorded audio
2. Transcription → processed via Whisper API
3. AI Enrichment → Gemini for intent recognition and tagging
4. Vector Embedding → stored in Qdrant using Sentence Transformers
5. Semantic Search → query system retrieves related insights by meaning

---

🧩 Architecture

Voice-to-Vector Architecture Diagram

🎙️Voice Input

 ⬇️️

💻[Flask API Server] — handles upload, processing, and response

 ⬇️

🧩[Whisper API] — transcription

 ⬇️

🤖[Gemini ] — enrichment + intent recognition

 ⬇️

🗄️[SentenceTransformer + Qdrant] — embeddings + vector storage

 ⬇️

🔍[Search Endpoint] — query for semantically similar notes

Key Components

• Flask: API backend handling requests and orchestration
• Qdrant: Vector database storing embeddings and metadata
• SentenceTransformer: Generates text embeddings
• LLM API (Gemini 2.5 Pro): AI reasoning layer for structuring and tagging insights
• Whisper API: Speech-to-text transcription

---

🧠 Features

• 🎙️ Transcribe voice notes automatically
• 🪄 AI-powered enrichment and sentiment tagging
• 🔎 Semantic search for finding related ideas and tasks
• 🧱 Modular components: replace or extend with your own APIs
• 📦 Local vector storage using Qdrant
• ⚙️ Easily deployable with Docker

---

🧰 Tech Stack

Layer	Tools Used
Frontend/Input	Local or browser-based audio recorder
Backend/API	Flask, Python
AI/ML	Whisper API, (Gemini 2.5 Pro)
Vector DB	Qdrant
Embeddings	SentenceTransformers (all-MiniLM-L6-v2)

---

🖥️ Installation & Setup

1️⃣ Clone the Repository

git clone https://github.com/98Devops/Voice-to-vector.git
cd voice-to-vector
Set Up Environment Variables

Create a .env file:

QDRANT_HOST=localhost
QDRANT_PORT=6333
GEMINI_API_KEY=your_api_key_here
WHISPER_API_KEY=your_whisper_api_key_here

3️⃣ Install Dependencies
pip install -r requirements.txt

4️⃣ Run Qdrant (Docker)
docker run -p 6333:6333 qdrant/qdrant

5️⃣ Start the Flask Server
python app.py


Flask will start on http://localhost:5000