A E G I S

⚡ The Autonomous Deep Research & Local-First Document Intelligence Nexus ⚡

"Why rent intelligence by the token when you can own the entire research lab on your GPU?"

---

🌌 The Vision: Sovereign Intelligence, Zero Compromise

No GPT-4 subscriptions. No monthly API bills. Zero private bytes leaving your machine.

Most developers underestimate the sheer power of quantized local LLMs when paired with agentic loops and high-fidelity retrieval pipelines.

Aegis proves that consumer hardware (e.g. RTX 4060 8GB / 24GB RAM) running open-weight GGUF models (Gemma 4 e4b, Llama 3.1 8B, Qwen 2.5 14B) can outperform monolithic cloud APIs by delegating the heavy lifting to autonomous multi-agent state machines, headless browser scrapers, and dense vector indexes.

Two mission-critical tools. One unified glassmorphic launchpad. Zero cloud tethering.

---

🎬 Visual Walkthrough & Core Engines

🛸 1. WAYFARER: Autonomous Deep Research Engine

"An AI research vessel surveying the live web while you sleep."

Instead of relying on single-shot LLM hallucinations, Wayfarer orchestrates an autonomous LangGraph multi-agent loop that plans, crawls, cross-examines, and synthesizes cited whitepapers in real time.

🔥 Wayfarer Superpowers:

• 🗺️ Autonomous Agent Swarm:
  • Planner Agent: Deconstructs broad prompts into tactical, multi-hop sub-questions.
  • Researcher Agent: Harvests live web sources with headless Playwright Chromium and DuckDuckGo.
  • Critic Agent: Audits incoming findings for factual depth, detects knowledge gaps, and triggers follow-up crawl rounds.
  • Writer Agent: Compiles comprehensive, academic-grade markdown reports with strict inline source citations.
• 🪐 3D Celestial Reconnaissance HUD: A cinematic Three.js space visualizer that maps active web probes, round progress, live search queries, and scraped URLs in real time.
• ⚡ Interactive Section-Level Refinement: Don't like a specific section? Click and refine it with custom directives (/api/refine-section) in real time without restarting the entire research mission.

---

📄 2. SYNTHESIZED REPORT & SECTION REFINEMENT

"Publication-ready research with in-place iterative refinement."

• 📝 Live Section-Level Rewriter: Select any header (## Summary, ### Market Dynamics) and command the agent to expand, simplify, or re-verify that specific topic.
• 💾 Instant Multi-Format Export: One-click downloads for Markdown (.md), formatted HTML (.html), Raw Plaintext (.txt), and Word Documents (.doc).

---

🧠 3. DOCKMIND: Local-First Document Intelligence & RAG

"Chat with your most confidential PDFs with zero cloud leaks."

💎 DockMind Superpowers:

• 📑 Docling Parser Integration: Industry-leading document understanding that extracts nested tables, LaTeX mathematical equations, and document hierarchies with extreme fidelity.
• 🧬 Dense Vector Memory: Powered by ChromaDB with nomic-embed-text-v1.5 embeddings and 1,500-character semantic chunk boundaries.
• 💬 Multi-Session Chat: SQLite-backed persistent memory across server reboots.

---

🌉 4. THE NEURAL BRIDGE: Zero-Friction Intelligence Handoff

"From raw web research to queryable vector memory in a single click."

   [ WAYFARER ] ────( One-Click Handoff )────> [ DOCKMIND ]
  Synthesized Report                         ChromaDB Vector Index
  + Source Citations                          + Interactive RAG Chat

Finished generating an extensive research report in Wayfarer? Simply click "Chat with this Research":

1. The report is instantly parsed and vector-embedded into ChromaDB.
2. A dedicated session is named and initialized in DockMind.
3. Omni Shell switches tabs in zero milliseconds—ready for you to interrogate your research.

---

🕹️ User Flow & Experience

                                    +-----------------------+
                                    |     AEGIS NEXUS       |
                                    |  (Initial Launchpad)  |
                                    +-----------+-----------+
                                                |
                       +------------------------+------------------------+
                       |                                                 |
                       v                                                 v
           [ KEY: 1 ] Click                                  [ KEY: 2 ] Click
        +----------------------+                          +----------------------+
        | <img src="assets/wayfarer-logo.svg" width="16" /> WAYFARER          |                          | <img src="assets/dockmind-logo.svg" width="16" /> DOCKMIND          |
        | Deep Research Engine |                          | Document RAG Chat    |
        +----------+-----------+                          +----------+-----------+
                   |                                                 |
                   +------------------------+------------------------+
                                            |
                                            v
                               +-------------------------+
                               | In-App Slide-Out Drawer |
                               |  - [1] Switch Wayfarer  |
                               |  - [2] Switch DockMind  |
                               |  - [H] Return Launchpad |
                               |  - [Esc] Fullscreen     |
                               +-------------------------+

1. Initial Portal Launchpad: Clean 2-option welcome gate on boot.
2. Dedicated Fullscreen Canvas: 100% full-bleed 100vw x 100vh workspace without static sidebar clutter.
3. Slide-Out Nav Drawer: Subtle [ ✦ AEGIS / AppName ▼ ] trigger on the top-left edge slides out a sleek drawer on demand.

---

🎮 Global Keyboard Shortcuts

Key	Action
1	Instantly launch / switch to Wayfarer Deep Research
2	Instantly launch / switch to DockMind Document Chat
H	Return to the Portal Launchpad (Home Hub)
Esc	Dismiss the slide drawer and return to 100% fullscreen

---

🏛️ System Architecture

graph TD
    User([User Browser]) <-->|Port 3050| OmniShell[Omni App Shell: Quantum Launchpad & Slide Drawer]
    
    subgraph OmniShell [Omni Shell Host]
        direction TB
        Launchpad[2-Option Portal Launchpad]
        SlideDrawer[Slide-Out Nav Drawer]
        TabResearch[Wayfarer Fullscreen Frame]
        TabChat[DockMind Fullscreen Frame]
    end

    subgraph WayfarerSystem [Wayfarer Subsystem]
        UI_Wayfarer[Wayfarer UI - Port 3000<br/>3D Space Visualizer & Recon HUD]
        BE_Wayfarer[Wayfarer Backend - Port 8000<br/>LangGraph Multi-Agent Engine]
        Playwright[Playwright Turbo Scraper & DuckDuckGo Search]
        DB_Wayfarer[(SQLite Research History)]
        
        UI_Wayfarer <-->|WebSocket / REST| BE_Wayfarer
        BE_Wayfarer <--> Playwright
        BE_Wayfarer <--> DB_Wayfarer
    end

    subgraph DockMindSystem [DockMind Subsystem]
        UI_DockMind[DockMind UI - Port 5173<br/>Multi-Session RAG Workbench]
        BE_DockMind[DockMind Backend - Port 8001<br/>Docling Parser & ChromaDB RAG]
        VectorDB[(ChromaDB: nomic-embed-text-v1.5)]
        DB_DockMind[(SQLite Chat History & Sessions)]
        
        UI_DockMind <-->|REST & WebSockets| BE_DockMind
        BE_DockMind <--> VectorDB
        BE_DockMind <--> DB_DockMind
    end

    subgraph LLM_Engine [Local Inference Engine]
        LlamaServer[llama.cpp / Local GGUF Endpoint - Port 8085<br/>e.g. Gemma 4 E4B / Llama 3 / Qwen 2.5]
    end

    OmniShell -->|Preloaded Iframe| UI_Wayfarer
    OmniShell -->|Preloaded Iframe| UI_DockMind
    
    %% Neural Bridge Handoff
    UI_Wayfarer -.->|1-Click 'Chat with Research' POST /ingest| BE_DockMind
    UI_Wayfarer -.->|postMessage SWITCH_TAB| OmniShell
    OmniShell -.->|postMessage SELECT_SESSION| UI_DockMind

    BE_Wayfarer <-->|Completions| LlamaServer
    BE_DockMind <-->|Completions| LlamaServer

Loading

---

🔌 Microservice Port Matrix

Service	Port	Description	Technology
Omni App Shell	3050	Quantum Launchpad & Slide Drawer	Vite + React
Wayfarer Frontend	3000	3D Celestial HUD & Research Console	Vite + React + Three.js
Wayfarer Backend	8000	LangGraph Multi-Agent Graph	FastAPI + LangGraph + Playwright
DockMind Frontend	5173	Document RAG Workbench	Vite + React + TypeScript + Tailwind
DockMind Backend	8001	Docling Parsing & Vector RAG	FastAPI + Docling + ChromaDB
LLM Server (Local)	8085	Local Inference Endpoint	llama-server / OpenAI API

---

🚀 Quick Start in 60 Seconds

Prerequisites

• Node.js v18+ & npm
• Python 3.10+
• Playwright Chromium:

  python -m playwright install chromium

• Local LLM Server (Optional if using NVIDIA NIM):
  • Run your favorite model via llama-server on http://localhost:8085/v1

---

Installation

1. Clone the Repository:

   git clone https://github.com/iam-tarun-86/Aegis.git
   cd Aegis

2. Install Frontend Dependencies:

   # Omni Shell
   cd omni_shell && npm install && cd ..
   
   # Wayfarer UI
   cd wayfarer/frontend && npm install && cd ..
   
   # DockMind UI
   cd dockmind/frontend && npm install && cd ..

3. Install Python Backend Dependencies:

   pip install fastapi uvicorn requests langchain-core langgraph duckduckgo-search playwright chromadb docling sentence-transformers pydantic

---

⚡ One-Click Startup (Windows Native)

Launch all 5 microservices in one command:

.\start_omni.ps1

💡 Developer Mode: To open visible terminal windows for every service during debugging, run:

.\start_omni.ps1 -ShowWindows

Open your browser and navigate to: 👉 http://localhost:3050

To stop all servers and free every port automatically, press Ctrl + C or run .\stop_omni.ps1.

---

📂 Project Structure

Aegis/
├── assets/                      # Vector logos, emblems & UI screenshots
│   ├── aegis-logo.svg
│   ├── wayfarer-logo.svg
│   ├── dockmind-logo.svg
│   └── screenshots/
│
├── start_omni.ps1               # Master one-click startup orchestrator
├── stop_omni.ps1                # Automatic port-killer & shutdown script
│
├── omni_shell/                  # Port 3050: Master App Shell
│   └── src/
│       ├── components/
│       │   ├── AegisLogo.jsx         # Quantum Singularity Nexus emblem
│       │   ├── PortalLaunchpad.jsx   # 2-option welcome dashboard
│       │   └── SlideNavDrawer.jsx    # Non-obstructive in-app slide drawer
│       ├── App.jsx
│       └── index.css
│
├── wayfarer/                    # Deep Research Subsystem
│   ├── frontend/                # Port 3000: React + Three.js UI
│   └── backend/                 # Port 8000: FastAPI + LangGraph
│       └── app/
│           ├── agents/          # Planner, Researcher, Critic, Writer
│           ├── tools/           # DuckDuckGo, Playwright, LLM Client
│           └── main.py          # WebSocket & Section Refinement APIs
│
└── dockmind/                    # Document RAG Subsystem
    ├── frontend/                # Port 5173: React + TS RAG Chat
    └── backend/                 # Port 8001: FastAPI + ChromaDB
        ├── db.py                # SQLite & ChromaDB configuration
        ├── ingest.py            # Docling high-fidelity document parser
        ├── llm.py               # Local LLM completion handler
        └── main.py              # Session & Ingest APIs

---

🛡️ License

Distributed under the MIT License. See LICENSE for more information.

---

_{⚡ Engineered with passion by Tarun | Powered by LangGraph, Docling & ChromaDB ⚡}