Cipher

Overview

Cipher is an advanced AI Research Assistant designed to provide accurate, context-aware answers by combining real-time web search with Large Language Model (LLM) capabilities. It utilizes a Retrieval-Augmented Generation (RAG) architecture to ground its responses in verifiable data, ensuring high reliability for research tasks.

Key Features

• Hybrid Retrieval System: Combines semantic vector search with keyword-based search to retrieve the most relevant information from the web.
• Contextual Intelligence: Maintains full conversation history to understand follow-up questions and resolve coreferences naturally.
• Real-Time Streaming: Delivers responses token-by-token for an immediate and interactive user experience.
• Source Citations: Automatically cites sources for every claim, allowing users to verify information directly.
• Persistent History: Saves all conversations to a database, enabling users to revisit and manage past research sessions.
• Premium Interface: Features a modern, responsive dark-mode UI with a focus on readability and aesthetics.

Architecture

graph TD
    %% Styles
    classDef frontend fill:#e3f2fd,stroke:#1565c0,stroke-width:2px,color:#0d47a1
    classDef backend fill:#e8f5e9,stroke:#2e7d32,stroke-width:2px,color:#1b5e20
    classDef database fill:#fff3e0,stroke:#ef6c00,stroke-width:2px,color:#e65100
    classDef external fill:#f3e5f5,stroke:#7b1fa2,stroke-width:2px,color:#4a148c
    classDef user fill:#eceff1,stroke:#455a64,stroke-width:2px,color:#263238

    User((User)):::user
    
    subgraph "Frontend Layer"
        UI[React UI]:::frontend
        State[Redux Store]:::frontend
        StreamHandler[Stream Handler]:::frontend
    end

    subgraph "Backend Layer"
        API[FastAPI Server]:::backend
        Orchestrator[Chat Orchestrator]:::backend
        Embedder[FastEmbed Model]:::backend
        Summarizer[Background Summarizer]:::backend
    end

    subgraph "Data Layer"
        MongoDocs[(MongoDB Documents)]:::database
        MongoVector[(Atlas Vector Search)]:::database
    end

    subgraph "External Services"
        Parallel[Parallel Systems API]:::external
        OpenRouter[OpenRouter LLM]:::external
    end

    %% Flows
    User <-->|Interaction| UI
    UI <-->|Dispatch/Select| State
    UI <-->|SSE Stream| StreamHandler
    
    StreamHandler <-->|POST /chat| API
    API <-->|Process| Orchestrator
    
    Orchestrator -->|1. Search Query| Parallel
    Parallel -->|2. Search Results| Orchestrator
    
    Orchestrator -->|3. Embed Query| Embedder
    Embedder -->|Vectors| Orchestrator
    
    Orchestrator <-->|4. Retrieve Context| MongoVector
    Orchestrator -->|5. Generate Response| OpenRouter
    
    OpenRouter -->|6. Stream Tokens| Orchestrator
    Orchestrator -->|7. Persist Chat| MongoDocs
    
    Summarizer -.->|Async Summary| OpenRouter
    Summarizer -.->|Update Metadata| MongoDocs

Loading

Interface Preview

Technology Stack

Backend

• Framework: FastAPI (Python)
• Database: MongoDB Atlas (Vector Store & Document Storage)
• LLM Integration: OpenRouter API (Streaming Support)
• Search Engine: Parallel Systems API
• Embedding Model: FastEmbed (On-device embedding generation)

Frontend

• Framework: React (Vite)
• State Management: Redux Toolkit
• Styling: Tailwind CSS
• Icons: Lucide React

Setup Instructions

Prerequisites

• Python 3.8+
• Node.js 16+
• MongoDB Atlas Account

Installation

1. Clone the Repository

   git clone <repository-url>
   cd Cipher

2. Backend Setup Navigate to the server directory and install dependencies:

   cd server
   pip install -r requirements.txt

   Create a .env file in the root directory with the following variables:

   MONGODB_URI=your_mongodb_connection_string
   OPENROUTER_API_KEY=your_openrouter_key
   PARALLEL_API_KEY=your_parallel_api_key

3. Frontend Setup Navigate to the client directory and install dependencies:

   cd ../client
   npm install

Running the Application

1. Start the Backend Server From the server directory:

   python -m uvicorn main:app --reload --port 8000

2. Start the Frontend Development Server From the client directory:

   npm run dev

   The application will be accessible at http://localhost:5173.

License

Proprietary software. All rights reserved.