DeepSearch AI — AI-Powered Document Intelligence

DeepSearch AI is a full-stack platform that lets users upload documents, search them using natural language, and get accurate insights powered by Google Gemini’s File Search API. The system includes authentication, secure file storage, multi-store organization, and a clean modern UI, making it ideal for research, knowledge management, and internal document search.

🚀 Features

🔐 User Authentication (Supabase)

• Email + password sign-up & login
• Secure session management
• Protected routes & user-scoped data isolation
• Logout and user profile display (shows username extracted from email)

📄 Document File Search (Gemini API)

DeepSearch AI uses:

• Gemini File Storage for uploading PDFs/TXT/DOCX
• Gemini File Search Tool-Calling to query files with structured semantic search
• Citations and context-aware answers returned from Gemini

More info: https://ai.google.dev/gemini-api/docs/file-search

🗂️ Organized Document Stores

• Each user can create multiple “stores”
• Stores help organize documents by project/topic
• Upload/delete documents inside each store
• Query a store and get AI answers with citations

⚡ Modern, Clean UI

• Fully responsive landing page
• Login & signup pages with modern card UI
• Dashboard inspired by SaaS document-search apps
• Dark mode toggle
• Smooth interaction flow: Landing → Auth → Dashboard → Store → Upload → Ask AI

🧪 AI-Powered Search

Ask questions in natural language:

• “Summarize the main findings of this research paper.”
• “Compare methods used across all stored documents.”
• “What does section 4.2 state about methodology?”

Gemini File Search returns precise, contextual answers.

---

📁 Repository Structure

README.md
requirements.txt
app/
    main.py                # FastAPI entrypoint
    deps.py                # Superbase
    gemini_client.py       # Gemini file search + model wrapper

static/
    index.html             # UI

---

🔧 Setup Instructions

1️⃣ Clone the repo

git clone https://github.com/<your-username>/DeepSearch-AI.git
cd DeepSearch-AI

---

2️⃣ Backend Setup (FastAPI)

Create a virtual environment:

python3 -m venv .venv
source .venv/bin/activate

Install dependencies:

pip install -r requirements.txt

Environment variables:

Create a .env file:

# Gemini
GEMINI_API_KEY=

# Supabase Auth
SUPABASE_URL=
SUPABASE_ANON_KEY=
SUPABASE_SERVICE_ROLE_KEY=
SUPABASE_JWT_SECRET=

---

3️⃣ Run Backend

python app/main.py

Backend will start at:

http://localhost:8000

---

🎨 **Frontend

Frontend available at:

http://localhost:8000

---

🔥 User Flow

1. Landing Page

Large hero section, marketing text, CTA button → “Get Started Free”.

2. Sign Up / Login

Clean authentication card with:

• Email input
• Password input
• Sign-in / Sign-up toggle

3. Dashboard

After login:

• Sidebar with list of stores
• Add new store
• User name displayed (first part of email)

4. Store View

Inside each store:

• Upload PDF/DOC/TXT
• View list of uploaded documents
• Delete docs
• Query input (Ask AI)

5. AI-Powered Search

Backend sends query → Gemini File Search → returns answer with citations.

---

📬 Support

Open an issue for bugs, feature requests, or integration questions.