Releases: mantisfury/ArkhamMirror
ArkhamMirror v1.0.0 Initial Public Release
Local-first AI document intelligence for investigative journalists
Connect the dots without connecting to the cloud.
Headline Features
🤖 Offline RAG (Chat with Your Documents)
Ask questions, get answers - all running locally. No OpenAI, no cloud APIs, no data leaving your machine.
🕸️ Contradiction Detection Engine
Automatically finds conflicting statements across thousands of documents. Builds "contradiction chains" showing how lies reference other lies.
🔗 Knowledge Graph (34,000+ Relationships)
Visualize who-knows-who with automatic entity extraction and relationship mapping. Path finding shows hidden connections between people, organizations, and places.
👁️ Vision AI Table Extraction
Extract complex financial tables from PDFs and images using Qwen-VL. Gets data other tools miss.
⏱️ Timeline Reconstruction
Automatically finds every date mention and builds chronological event sequences. See what happened when.
Investigation Tools
- Geospatial Intelligence - Auto-map locations, heatmaps, geographic patterns
- Red Flag Detection - Pattern matching for suspicious content with customizable weights
- Regex Pattern Extraction - Built-in patterns for emails, phones, SSNs, credit cards, plus custom regex
- Author Unmasking - Metadata forensics to find hidden file history and true authorship
- Hidden Content Detection - Steganography detection, invisible text, doctored images
- Narrative Reconstruction - AI-powered gap analysis and speculation mode
Privacy & Architecture
- 100% Local - Air-gapped operation, zero cloud dependencies
- Data Isolation - All investigation data stays in local "DataSilo"
- No Subscriptions - MIT licensed, fully open source
- Hybrid OCR - PaddleOCR (fast/CPU) + Qwen-VL (smart/GPU)
Installation
Windows: setup.bat
Mac/Linux: chmod +x setup.sh && ./setup.sh
Requirements: Python 3.11+, Docker Desktop, 16GB RAM (8GB minimum)
Tech Stack: Reflex 0.7.14, PostgreSQL 15, Qdrant, Redis 7, spaCy, BAAI/bge-m3
See README for full documentation and CONTRIBUTING to get involved.
Perfect for: Investigative journalists, OSINT analysts, researchers, anyone who needs powerful document analysis without cloud surveillance. Intelligence for everyone.