Intelligence Analysis — by Aryan is a fully offline, single-folder intelligence-analysis and tradecraft-education platform.
It combines 20 Structured Analytic Techniques (SATs), a Personality Lab, a searchable Intelligence Education Library, and a locally hosted Qwen2.5-7B-Instruct model through llama.cpp.
No cloud calls. No API keys. No external AI service. No data leaving the machine.
Once the required local engine and GGUF model are installed, the platform can operate without an Internet connection.
The application is designed around local processing:
User
↓
index.html
↓
Local Browser Interface
↓
llama-server
↓
Qwen2.5-7B-Instruct GGUF
↓
Local AI Response
The AI engine communicates through a local OpenAI-compatible REST API:
/v1/chat/completions
Streaming responses are supported so analysis appears progressively in the interface.
No information needs to leave the computer for AI analysis.
IntelligenceAnalysis-ByAryan/
│
├── index.html
│
├── start.sh
├── start.bat
│
├── assets/
│ ├── css/
│ │ └── style.css
│ │
│ └── js/
│ ├── data-methods.js
│ ├── data-personality.js
│ ├── data-library.js
│ ├── engine.js
│ └── app.js
│
├── engine/
│ └── llama-server(.exe)
│
├── models/
│ └── Qwen2.5-7B-Instruct-GGUF
│
└── docs/
└── SETUP_GUIDE.md
The platform contains 20 dedicated Structured Analytic Techniques divided into five major categories.
Each method has its own tailored prompt template rather than relying on one generic analysis prompt.
| Category | Techniques |
|---|---|
| Diagnostic & Bias Removal | ACH, Key Assumptions Check, Quality of Information Check, Structured Self-Critique |
| Contrarian & Challenge | Devil's Advocacy, Red Teaming, High-Impact/Low-Probability Analysis, What-If Analysis |
| Threat, Risk & Strategic | SWOT, PESTLE, Linchpin Analysis, Deception Detection |
| Forecasting & Scenario Planning | Alternative Futures, Indicators & Warning Tracking, Cone of Plausibility |
| Network, Cause & Link | Link/Network Analysis, Pattern of Life, Morphological Analysis, Chronology/Timeline Analysis, Event Tree Mapping |
The platform is designed to help users examine information from multiple analytical perspectives.
Structures evidence against competing explanations and helps identify which hypothesis is least disconfirmed by the available evidence.
Identifies assumptions underlying an assessment and records confidence and potential break conditions.
Examines information quality, reliability, relevance and limitations before allowing it to drive an assessment.
Provides a deliberate challenge to an existing assessment to identify overlooked evidence, assumptions and reasoning weaknesses.
Builds a structured argument against a prevailing assessment.
Challenges plans, assumptions and conclusions from an opposing perspective.
Examines events that may have low probability but potentially significant consequences.
Explores how an assessment could change if a key condition or assumption changed.
Structures Strengths, Weaknesses, Opportunities and Threats.
Examines Political, Economic, Social, Technological, Legal and Environmental factors.
Identifies assumptions, facts or conditions upon which an assessment heavily depends.
Provides a structured framework for examining inconsistencies, indicators and possible deceptive activity.
Develops multiple plausible future scenarios instead of relying on a single forecast.
Identifies observable indicators that could signal movement toward a particular scenario.
Explores a range of possible and plausible future outcomes.
Examines relationships between entities, events and information.
Structures recurring behavioral or activity patterns.
Breaks complex problems into dimensions and explores combinations of possible conditions.
Organizes events chronologically to expose sequence, timing and potential relationships.
Maps possible branches and consequences from a starting event.
The Personality Lab contains two separate frameworks:
The system can examine behavioral descriptions through the five-factor personality framework:
- Openness
- Conscientiousness
- Extraversion
- Agreeableness
- Neuroticism
A separate MBTI workbench provides a framework-based interpretation of behavioral descriptions.
Users can paste approximately 500–1000 words describing someone's behavior and run an analysis based on the supplied text.
Important: The Personality Lab is a reflective/developmental tool and not a clinical psychological instrument.
The platform also explicitly acknowledges known reliability limitations associated with MBTI.
The Education section provides searchable learning material covering core intelligence-analysis and tradecraft concepts.
Topics include:
- Confirmation Bias
- Mirror Imaging
- Linchpins
- Warning Intelligence
- Analytical Bias
- Alternative Analysis
- Structured Reasoning
- Indicators & Warnings
- Strategic Assessment
- Scenario Analysis
- Network Reasoning
- Intelligence Tradecraft
The library contains 12 original educational topic essays and a 19-title bibliography.
The system deliberately avoids inventing detailed content for references that have not been independently established.
The Intelligence Library distinguishes between:
Verified detailed source material
and
Reference-only bibliographic entries.
The implementation is specifically designed to avoid presenting invented chapter summaries or falsely claiming detailed knowledge of books that have only been included as references.
Heuer's Psychology of Intelligence Analysis is the primary verified source for the detailed chapter-by-chapter educational outline in this build.
The platform uses:
| Component | Technology |
|---|---|
| AI Engine | llama.cpp |
| Local Server | llama-server |
| AI Model | Qwen2.5-7B-Instruct GGUF |
| API | OpenAI-compatible local REST API |
| Streaming | Server-Sent Events |
| Frontend | HTML5 + CSS3 + JavaScript |
| Launcher | Windows BAT / Unix Shell |
| Operation | Local / Offline |
The model and llama.cpp engine are intentionally not bundled into the text-based project.
The setup guide provides the required installation instructions and verified download locations.
The GGUF model and operating-system-specific llama.cpp binaries are not included directly in the source package.
This is intentional because:
- Model weights can occupy multiple gigabytes.
- llama.cpp binaries differ by operating system and architecture.
- Keeping them separate makes the application portable.
- Users can select an appropriate GGUF model/quantization for their hardware.
After the one-time setup, the application can operate without Internet access.
The platform follows an offline-first design philosophy.
No cloud AI API is required.
No API key is required.
No user analysis needs to be transmitted to an external AI provider.
The local browser communicates with the local llama-server running on the same machine.
For sensitive analytical material, users should still follow their organization's own security classification, handling and authorization policies.
The project includes launcher logic designed around practical local deployment.
Features include:
- Startup checks
- Engine configuration
- Crash-recovery marker
- Local engine connectivity
- Cross-platform launcher scripts
- Configuration handling
The Windows launcher is:
start.bat
The Unix/macOS launcher is:
start.sh
The system is designed to support structured reasoning, not replace human analytical judgment.
AI-generated assessments should be treated as analytical assistance and examined against:
- Source quality
- Evidence
- Assumptions
- Alternative explanations
- Contradictory information
- Uncertainty
- Analyst judgment
The platform is particularly designed to encourage users to challenge their own conclusions rather than simply confirm them.
The current platform is focused on Intelligence Analysis and OSINT/tradecraft education, but its architecture can be expanded substantially.
Future versions can incorporate additional intelligence disciplines and specialized analytical modules, including:
- OSINT — Open Source Intelligence
- HUMINT — Human Intelligence
- SIGINT — Signals Intelligence
- IMINT — Imagery Intelligence
- GEOINT — Geospatial Intelligence
- MASINT — Measurement and Signature Intelligence
- TECHINT — Technical Intelligence
- FININT — Financial Intelligence
- CYBINT — Cyber Intelligence
- SOCMINT — Social Media Intelligence
- ELINT — Electronic Intelligence
- COMINT — Communications Intelligence
- FISINT — Foreign Instrumentation Signals Intelligence
- TELINT — Telemetry Intelligence
- RADINT — Radar Intelligence
- ACINT — Acoustic Intelligence
- MEDINT — Medical Intelligence
- CRIMINT — Criminal Intelligence
- DOMINT — Domestic Intelligence
- SCIINT — Scientific Intelligence
The long-term objective can therefore evolve from an Intelligence Analysis desk into a broader offline intelligence-fusion and analytical workstation, with each intelligence discipline contributing specialized collection, processing, visualization and assessment capabilities.
A future architecture could combine multiple intelligence streams:
OSINT ──────┐
HUMINT ─────┤
SIGINT ─────┤
IMINT ──────┤
GEOINT ─────┤
MASINT ─────┤
FININT ─────┤
CYBINT ─────┤
SOCMINT ────┤
TECHINT ────┤
MEDINT ─────┤
↓
INTELLIGENCE FUSION
↓
STRUCTURED ANALYSIS
↓
ASSESSMENT
↓
HUMAN ANALYST
This is an expansion roadmap, not a claim that all of these intelligence disciplines are implemented in v1.0.
A natural future extension is a mobile/tablet companion.
The static frontend could potentially be packaged using a WebView or Capacitor-based application and communicate with the local llama-server over a trusted local network.
This would allow:
Desktop AI Engine
↕
Local Wi-Fi
↕
Phone / Tablet
The desktop would remain the primary local inference engine while other authorized devices act as thin clients.
| Component | Status |
|---|---|
| Browser-based frontend | ✅ Implemented |
| 20 Structured Analytic Techniques | ✅ Implemented |
| Dedicated SAT prompts | ✅ Implemented |
| Big Five Personality Lab | ✅ Implemented |
| MBTI Personality Lab | ✅ Implemented |
| Intelligence Education Library | ✅ Implemented |
| Searchable educational topics | ✅ Implemented |
| Bibliography | ✅ Implemented |
| Local llama.cpp engine integration | ✅ Implemented |
| Qwen2.5-3B GGUF support | ✅ Implemented |
| Streaming responses | ✅ Implemented |
| Windows launcher | ✅ Implemented |
| macOS/Linux launcher | ✅ Implemented |
| Crash-recovery marker | ✅ Implemented |
| Cloud AI dependency | ❌ None |
| API-key requirement | ❌ None |
| Internet requirement at runtime | ❌ None |
| Session-history panel | 🔄 Future |
| PDF/Word export | 🔄 Future |
| Hardware-based quantization recommendation | 🔄 Future |
| Multi-device client | 🔄 Future |
| Multi-INT fusion architecture | 🔄 Future |
- Session history
- Save completed analyses
- JSON case storage
- PDF export
- Word export
- Hardware detection
- Automatic GGUF recommendation
- Expanded educational library
- Advanced link analysis
- Timeline visualization
- Indicator dashboards
- Scenario comparison
- Evidence management
- Analyst workspaces
- Cross-case analysis
Potential integration of:
OSINT + HUMINT + SIGINT + IMINT + GEOINT + MASINT + TECHINT + FININT + CYBINT + SOCMINT + COMINT + ELINT + MEDINT + CRIMINT + other specialized intelligence disciplines.
Intelligence Analysis — by Aryan is an analytical-support and tradecraft-education platform.
It does not independently establish facts, determine the truth of intelligence reporting, or replace qualified human analysts.
AI-generated outputs may contain errors or incorrect interpretations. Users should independently verify important claims against reliable sources and apply appropriate analytical standards.
Personality analysis is provided for reflective/developmental purposes and is not a clinical diagnosis.
Users remain responsible for ensuring that information entered into the system is handled according to applicable laws, organizational policies, classification requirements and authorization rules.
The project is built around five principles:
1. Offline First
Keep processing local wherever practical.
2. Structured Reasoning
Use established analytical frameworks instead of relying exclusively on free-form AI responses.
3. Transparency
Make the analytical method visible to the user.
4. Intellectual Challenge
Encourage alternative explanations and deliberate contradiction.
5. Human-in-the-Loop
AI assists the analyst; it does not become the analyst.
Project: Intelligence Analysis — by Aryan Version: v1.0 Architecture: Local Browser + llama.cpp Model: Qwen2.5-7B-Instruct GGUF Operation: Offline / Local Primary Purpose: Intelligence Analysis & Tradecraft Education
Built as a local analytical workstation concept focused on structured intelligence reasoning, tradecraft education, alternative analysis and future multi-INT expansion.