Privacy-focused local RAG system with hybrid retrieval, reranking, document generation, and technical diagram support.
Hibrit retrieval, reranking, doküman üretimi ve teknik diyagram desteğine sahip gizlilik odaklı yerel RAG sistemi.
DeepResearch is a fully local AI-powered research assistant designed for secure and offline environments. The system allows users to upload documents, perform semantic search, generate grounded answers, and create structured outputs such as reports, presentations, diagrams, and audio summaries — all without cloud APIs.
DeepResearch, güvenli ve çevrimdışı ortamlar için geliştirilmiş tamamen yerel çalışan yapay zeka destekli bir araştırma asistanıdır. Sistem; kullanıcıların doküman yüklemesine, semantik arama yapmasına, kaynaklara dayalı cevaplar üretmesine ve rapor, sunum, diyagram, sesli çıktı gibi yapılar oluşturmasına olanak sağlar.
- PDF support | PDF desteği
- DOCX support | DOCX desteği
- TXT support | TXT desteği
- Automatic text extraction and chunking | Otomatik metin çıkarma ve chunking
- Metadata tracking | Metadata takibi
-
Dense vector search using FAISS
-
Sparse keyword search using BM25
-
Reciprocal Rank Fusion (RRF)
-
CrossEncoder reranking
-
FAISS tabanlı dense vector arama
-
BM25 tabanlı keyword arama
-
Reciprocal Rank Fusion (RRF)
-
CrossEncoder reranking desteği
- Fully offline inference | Tamamen çevrimdışı inference
- GGUF model support | GGUF model desteği
- llama.cpp integration | llama.cpp entegrasyonu
- Streaming responses | Streaming çıktı desteği
- Context-aware prompt generation | Bağlam farkındalıklı prompt üretimi
Supported models | Desteklenen modeller:
- Qwen
- Llama
- Phi
- Other GGUF-compatible models | GGUF uyumlu diğer modeller
Generate:
- Structured reports | Yapısal raporlar
- Summaries | Özetler
- PowerPoint presentations | PowerPoint sunumları
- Word documents | Word dokümanları
- Technical explanations | Teknik açıklamalar
Automatically generates:
- Flowcharts | Akış şemaları
- System architectures | Sistem mimarileri
- Pipeline diagrams | Pipeline diyagramları
- Technical process diagrams | Teknik süreç diyagramları
Example outputs | Örnek çıktılar:
- RAG pipelines
- AI workflows
- Software architectures
- Electronics/system diagrams
- Text-to-speech support | Metin seslendirme desteği
- Audio response generation | Sesli çıktı üretimi
- Chat history support | Sohbet geçmişi desteği
- Rolling conversation summaries | Özet tabanlı hafıza sistemi
- JSON memory structure | JSON tabanlı hafıza yapısı
- Reduced context window usage | Daha verimli context kullanımı
User Query / Kullanıcı Sorusu
↓
Hybrid Retrieval
(FAISS + BM25)
↓
RRF Fusion
↓
CrossEncoder Reranking
↓
Local LLM (GGUF)
↓
Grounded Response / Kaynak Destekli Cevap
- Python
- Flask
- SentenceTransformers
- FAISS
- BM25
- CrossEncoder
- llama.cpp
- HTML
- CSS
- JavaScript
- Mermaid.js
- pdfplumber
- python-docx
git clone https://github.com/haticenurcakr/DeepResearch.git
cd DeepResearchpython -m venv venv
venv\Scripts\activatepython3 -m venv venv
source venv/bin/activatepip install -r requirements.txtpython app.pyDeepResearch/
│
├── api/ # API routes
├── core/ # Core RAG logic
├── services/ # AI and document services
├── static/ # CSS / JS / assets
├── templates/ # HTML templates
├── utils/ # Utility functions
│
├── app.py # Main Flask application
├── config.py # Configuration file
├── requirements.txt
└── README.md
- Academic research assistant | Akademik araştırma asistanı
- Offline enterprise document analysis | Çevrimdışı kurumsal doküman analizi
- Secure/private AI assistant | Güvenli özel yapay zeka asistanı
- Technical report generation | Teknik rapor üretimi
- Research paper summarization | Makale özetleme
- Local AI chatbot systems | Yerel AI chatbot sistemleri
- Air-gapped network deployments | İzole ağ sistemleri
Many AI systems require cloud access and external APIs.
DeepResearch is designed for environments where privacy, local processing, and offline operation are critical.
Birçok yapay zeka sistemi bulut servislerine bağımlıdır. DeepResearch ise gizlilik, yerel işlem ve çevrimdışı kullanım gerektiren ortamlar için geliştirilmiştir.
Add screenshots here:
- Chat interface
- Retrieval results
- Mermaid diagram outputs
- Generated reports
- Document upload page
Buraya ekran görüntüleri ekleyebilirsiniz:
- Sohbet arayüzü
- Retrieval sonuçları
- Mermaid diyagram çıktıları
- Üretilen raporlar
- Doküman yükleme ekranı
- Multi-user support | Çok kullanıcılı yapı
- Agentic workflows | Agent tabanlı iş akışları
- OCR support | OCR desteği
- Docker deployment | Docker desteği
- Electron/Tauri desktop version | Masaüstü uygulama sürümü
- Advanced citation system | Gelişmiş kaynak sistemi
- Multi-modal support | Multi-modal destek
Contributions, issues, and feature requests are welcome.
Katkılar, öneriler ve hata bildirimleri memnuniyetle karşılanır.
This project is intended for educational and research purposes.
Bu proje eğitim ve araştırma amaçlı geliştirilmiştir.
Hatice Nur Çakır
Computer Engineering Student — Düzce University
Düzce Üniversitesi Bilgisayar Mühendisliği Öğrencisi
AI • RAG • Local LLM Systems • Research Tools