IntellyWeave
January 12, 2026 · View on GitHub
AI-powered OSINT platform that transforms document chaos into actionable intelligence through entity extraction, geospatial analysis, archive discovery, and hypothesis-driven investigation.
Quick Start | Installation | Feature Guides | Live Demo
What is IntellyWeave?
IntellyWeave is an Open-Source Intelligence (OSINT) analysis platform that automatically extracts entities from documents, visualizes relationships on maps and network graphs, and employs multiple AI agents that debate complex questions to deliver well-reasoned answers with source citations.
Upload your documents. Ask questions in natural language. Get intelligence.
Mission
IntellyWeave democratizes intelligence analysis by making professional-grade OSINT workflows accessible through AI automation. We eliminate the manual labor of entity extraction, relationship mapping, and geographic analysis—transforming months of work into minutes.
Who Should Use This?
IntellyWeave is for teams who need to:
- Intelligence Analysts — Conduct OSINT research, connect disparate sources, build comprehensive assessments
- Historical Researchers — Explore archives, declassified documents, and primary sources for patterns and narratives
- Investigators — Track entities across documents, piece together evidence, build case narratives
- Research Professionals — Extract structured data from unstructured sources, discover non-obvious relationships
Scope
In Scope
- Automatic entity extraction (persons, organizations, locations, dates, events, laws, cryptonyms)
- Geospatial visualization with interactive 3D maps
- Network relationship analysis and graph visualization
- Archive discovery and hypothesis-driven investigation (Quartermaster + Case Officer)
- Multi-agent reasoning for complex analytical questions
- Multi-format document processing (PDF, DOCX, TXT, Markdown)
- Multi-provider LLM support (OpenAI, Anthropic, Google, local models)
Out of Scope
- Legal case analysis (that's Spectre's domain)
- Real-time surveillance or monitoring
- Automated decision-making without human review
Key Features
Entity Extraction (GLiNER)
Automatic identification of 7 entity types from multilingual documents using zero-shot recognition. No training required.
Entity Extractor identifying key persons with 92% confidence, enriched with contextual analysis and follow-up suggestions.
| Entity Type | Examples |
|---|---|
| Persons | Klaus Barbie, Josef Mengele |
| Organizations | Vatican, CIA, ODESSA |
| Locations | Buenos Aires, Rome, Damascus |
| Dates | May 1945, 1960s |
| Events | Nuremberg Trials |
| Laws | Decreto-Lei 7967/1945 |
| Cryptonyms | Operation Paperclip |
Geospatial Intelligence
Interactive 3D maps powered by Mapbox GL. Plot extracted locations, visualize routes, explore geographic patterns.
Geospatial Analyst extracting operational locations with coordinates, route data, and strategic intelligence assessments.
Interactive Mapbox heatmap showing location intensity across Italy—key embarkation points for transatlantic escape routes.
Network Analysis
Relationship graphs with physics-based layouts using vis-network. Discover hidden connections between entities.
Force-directed network graph revealing the rat lines organization: persons (orange), organizations (green), and locations (blue) with documented relationship edges.
Archive Research
Two-agent investigative system for archival discovery and hypothesis-driven research:
| Agent | Role |
|---|---|
| Quartermaster | Maps the information landscape—discovers archives, classifies access levels, identifies digitization status |
| Case Officer | Conducts the investigation—tests hypotheses, gathers evidence, synthesizes reports with actionable next steps |
Case Officer generates structured hypotheses with evidence tracking: Confirmed (85%), Indeterminate (45%), Pending, or Refuted—each with supporting and contradicting evidence.
Key capabilities:
- Curated archive sources (30+ institutions across 10 geographic regions)
- Access level classification (Public, Subscription, Physical-Only, Restricted)
- Hypothesis generation with confidence scoring and evidence citations
- Step-by-step access instructions for restricted archives
- Intelligent PDF preview with AI-inferred schema
- Documents flagged for manual review with rich metadata for follow-up
Archive Research Guide | Configuration Guide
Intelligence Orchestrator
Automated 6-phase analysis pipeline that transforms uploaded documents into comprehensive intelligence assessments with actionable follow-up tasks:
| Phase | Agent | Output |
|---|---|---|
| 1. Extraction | Entity Extractor | Persons, organizations, locations, dates, events, laws, cryptonyms |
| 2. Mapping | Relationship Mapper | Entity connections with relationship types and confidence scores |
| 3. Geospatial | Geospatial Analyst | Location coordinates, routes, movement patterns |
| 4. Network | Network Analyst | Graph structures, centrality metrics, cluster detection |
| 5. Patterns | Pattern Detector | Cross-document correlations, temporal sequences, anomalies |
| 6. Synthesis | Synthesizer | Executive summary, key findings, unresolved questions |
Intelligence Orchestrator generates follow-up investigation tasks from each analysis phase—color-coded by originating agent with priority indicators (High, Medium) and status tracking.
Key capabilities:
- Automatic task generation from each analysis phase
- Priority-weighted investigation queue
- Provenance tracking (which agent generated each finding)
- Confidence accumulation across phases
- Executive summary with evidence citations
Intelligence Analysis Guide | Ticket Display Guide
LLM Support
Multi-provider support including OpenAI, Anthropic Claude, Google Gemini, and local models via Ollama. Configurable reasoning effort controls for complex analysis tasks.
IntellyWeave CLI
Operations backbone for IntellyWeave—manage Weaviate databases, migrate data, configure archives, and query with AI from the command line.
| Capability | Description |
|---|---|
| Weaviate Management | Collections, objects, hybrid search, stats |
| Data Migration | Export/import with mutation safety (--live flag) |
| Archive Configuration | Add sources for Quartermaster |
| AI Shell | Natural language queries with Claude + MCP integration |
# Clone and start interactive shell
git clone https://github.com/vericle/intellyweave-cli.git
cd intellyweave-cli && npm install && npm run dev shell
CLI Documentation | GitHub Repository
See It In Action
Nazi Rat Lines Demo
Explore how IntellyWeave analyzes 17 historical documents to uncover Nazi escape networks to South America (1945-1962).
Launch Interactive Demo — Click through a guided tour without installing anything.
What you'll discover:
- How a single name (Father Draganovic) unravels an entire network
- Three distinct escape routes from Europe to South America
- A courthouse debate on whether Brazilian immigration law was exploited
Full Demo Documentation | Step-by-Step Walkthrough
Quick Start
# 1. Start local Weaviate
docker compose up -d weaviate
# 2. Setup dependencies
scripts/setup.sh
# 3. Configure API keys
cp backend/.env.example backend/.env
# Edit backend/.env with your LLM provider key
# 4. Launch
cd backend && source .venv/bin/activate && elysia start
# 5. Open http://localhost:8000
Detailed Installation Guide | First Query Guide
Enable Entity Extraction
cd backend && source .venv/bin/activate
pip install torch --index-url https://download.pytorch.org/whl/cpu
pip install -e ".[ner]"
Documentation
Getting Started
| Guide | Description |
|---|---|
| Quick Start | 5-minute setup |
| Installation | Detailed configuration |
| First Query | Your first analysis |
Feature Guides
| Guide | Description |
|---|---|
| Entity Extraction | GLiNER (7 entity types) |
| Geospatial Mapping | Mapbox 3D maps |
| Network Analysis | vis-network graphs |
| Archive Research | Quartermaster + Case Officer |
| Document Processing | Pipeline & watchdog |
| Courthouse Debate | Multi-agent reasoning |
| Intelligence Analysis | 6-phase orchestrator |
| LLM Configuration | Multi-provider support |
| Agents | Domain routing, custom agents |
Reference
| Document | Description |
|---|---|
| Architecture | System design |
| API Endpoints | REST API |
| Environment Variables | Configuration |
| CLI Commands | Command line |
Technical Stack
| Layer | Technology |
|---|---|
| Backend | Python 3.12, FastAPI, Weaviate, DSPy, LiteLLM |
| Frontend | Next.js 15, React 18, TypeScript, Tailwind CSS |
| CLI | TypeScript, Commander.js, Vercel AI SDK, MCP |
| Entity Extraction | GLiNER multi-v2.1 (zero-shot NER) |
| Geospatial | Mapbox GL 3.16 with 3D controls |
| Network Graphs | vis-network 10.0.2 with ForceAtlas2 |
| LLM Providers | OpenAI, Anthropic, Google, Ollama |
Support
- GitHub Issues — Bug reports and feature requests
- GitHub Discussions — Questions and community discussion
Contributing & Governance
| Document | Description |
|---|---|
| Contributing Guide | How to contribute |
| Code of Conduct | Community standards |
| Security Policy | Vulnerability reporting |
| Governance | Decision-making process |
| Maintainers | Project maintainers |
| Roadmap | Project direction |
License
BSD 3-Clause License — see LICENSE for details.
Acknowledgments
IntellyWeave is built on:
- Weaviate Elysia — Agentic AI framework with decision tree architecture
- Weaviate — Vector database for semantic search
- Spectre — Agentic Legal AI system (based on Elysia)
- new/s/leak — Open-source investigative journalism platform (Hamburg University, TU Darmstadt, Der Spiegel) — the original inspiration and foundation for IntellyWeave's document analysis capabilities
- GLiNER — Zero-shot named entity recognition
- Mapbox GL — Interactive 3D mapping
- vis-network — Network graph visualization
IntellyWeave — Where intelligence meets insight.