Disha

Disha — Usage Guide

Installation & Setup
CLI Usage
AI Intelligence Platform
Decision Engine
Historical Strategy Module
Cyber Defense System
MCP Server Integration
Web Dashboard
Training AI Models
Continuous Learning
Sentinel Monitoring
Configuration
Troubleshooting

Installation & Setup

Prerequisites

Bun 1.1+ (CLI runtime)
Node.js 18+ (MCP server, web dashboard)
Python 3.11+ (AI platform, decision engine, cyber defense)
Docker (optional, for full stack)

Clone and Install

git clone https://github.com/Tashima-Tarsh/Disha.git
cd Disha
npm install

Environment Variables

Create a .env file in the root directory:

# .env
ANTHROPIC_API_KEY=your_anthropic_key     # For Claude LLM
OPENAI_API_KEY=your_openai_key           # For GPT-4o (optional)
NEO4J_URI=bolt://localhost:7687          # Graph DB (optional)
NEO4J_PASSWORD=your_password             # Graph DB (optional)

Note: Disha works with mock providers and open-source APIs — no paid keys required for core functionality.

CLI Usage

The Disha CLI is built with Bun + TypeScript + React/Ink:

# Start the CLI
bun run src/entrypoints/cli.tsx

# List available tools
bun run src/entrypoints/cli.tsx --tools

# List available commands
bun run src/entrypoints/cli.tsx --commands

AI Intelligence Platform

The AI platform runs 7 specialized agents orchestrated via FastAPI:

cd ai-platform/backend
pip install -r requirements.txt
uvicorn app.main:app --reload --port 8000

API Endpoints

Endpoint	Method	Description
`/api/v1/investigate`	POST	Launch multi-agent investigation
`/api/v1/agents/osint`	POST	OSINT data collection
`/api/v1/agents/crypto`	POST	Blockchain analysis
`/api/v1/agents/detection`	POST	Anomaly detection
`/api/v1/agents/graph`	POST	Knowledge graph queries
`/api/v1/agents/reasoning`	POST	LLM-powered reasoning
`/api/v1/multimodal/vision`	POST	Image analysis
`/api/v1/multimodal/audio`	POST	Audio transcription

Decision Engine

Multi-agent reasoning with political, legal, ideological, and security analysis:

cd decision-engine
pip install -r requirements.txt

# Run with mock LLM (no model download needed)
DISHA_MODEL_PROVIDER=mock python main_decision_engine.py

# Run tests
DISHA_MODEL_PROVIDER=mock python -m pytest tests/ -v

Optional: FAISS Retrieval

pip install faiss-cpu sentence-transformers

Optional: Local LLM

pip install llama-cpp-python
export DISHA_MODEL_PROVIDER=llamacpp
export DISHA_MODEL_PATH=/path/to/model.gguf

Historical Strategy Module

AI classifier and simulation engine for 32+ historical conflicts:

cd historical-strategy
pip install -r requirements.txt

# Start the API server
uvicorn api.main:app --reload --port 8001

# Train the classifier
python model/train.py

# Run a simulation
python simulation/engine.py

Cyber Defense System

AI-powered honeypot stack with threat classification:

# Using Docker (recommended)
cd cyber-defense
docker-compose up -d

# Train the threat classifier
cd model
pip install torch --index-url https://download.pytorch.org/whl/cpu
python train.py

MCP Server Integration

The Model Context Protocol server exposes Disha’s tools for AI assistants:

cd mcp-server
npm install
npm run dev    # Development mode
npm run build  # Build for production
npm start      # Production mode

Live deployment: disha.vercel.app

Web Dashboard

Next.js dashboard for threat intelligence visualization:

cd web
npm install
npm run dev    # http://localhost:3000

Training AI Models

Train All Components

python scripts/train_all.py

Individual Training

# Reinforcement Learning (PPO)
cd ai-platform/backend && python -m app.rl.train

# Graph Neural Networks
cd ai-platform/backend && python graph_ai/train.py

# Decision Engine
cd decision-engine && DISHA_MODEL_PROVIDER=mock python train.py

Continuous Learning

Disha supports continuous self-improvement:

# Offline mode (synthetic data)
python scripts/continuous_train.py --rounds 3 --offline

# Online mode (fetches from abuse.ch, arXiv, OEIS)
python scripts/continuous_train.py --rounds 3

# Single component
python scripts/continuous_train.py --rounds 3 --component rl

Sentinel Monitoring

Real-time threat monitoring and self-healing:

# Run the full sentinel system
python scripts/sentinel/guardian.py

# Run tests
python -m pytest scripts/sentinel/test_sentinel.py -v

Configuration

Biome (TypeScript Linting)

npx biome check src/                         # Check
npx biome check --write src/                  # Safe fixes
npx biome check --write --unsafe src/         # All auto-fixes

Python Linting

flake8 ai-platform/backend/ --max-line-length=120 --ignore=E501,W503,W504
flake8 decision-engine/ --max-line-length=120 --ignore=E501,W503,W504

Troubleshooting

Issue	Solution
`bun` command not found	Install Bun: `curl -fsSL https://bun.sh/install \\| bash`
Python import errors	Ensure you’re in the correct subdirectory and have installed requirements
PyTorch not found	Install CPU-only: `pip install torch --index-url https://download.pytorch.org/whl/cpu`
Neo4j connection refused	Start Neo4j or set `NEO4J_URI` environment variable
FAISS not available	Install: `pip install faiss-cpu` (optional dependency)
TypeScript errors	Some SDK types are generated at build time — run `bun run build` first

Full documentation: WIKI.md · Architecture · LEARNING_LOG.md

This site is open source. Improve this page.